diff --git a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetApplicationService.java b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetApplicationService.java
index c19a707..43ca54b 100644
--- a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetApplicationService.java
+++ b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/application/DatasetApplicationService.java
@@ -21,6 +21,7 @@ import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
 import org.apache.commons.collections4.CollectionUtils;
 import org.springframework.beans.factory.annotation.Value;
+import org.springframework.data.redis.core.StringRedisTemplate;
 import org.springframework.scheduling.annotation.Async;
 import org.springframework.stereotype.Service;
 import org.springframework.transaction.annotation.Transactional;
@@ -42,11 +43,13 @@ import java.util.stream.Stream;
 @Transactional
 @RequiredArgsConstructor
 public class DatasetApplicationService {
+    private static final String DATASET_PVC_NAME = "sys.management.dataset.pvc.name";
     private final DatasetRepository datasetRepository;
     private final TagMapper tagMapper;
     private final DatasetFileRepository datasetFileRepository;
     private final CollectionTaskClient collectionTaskClient;
     private final DatasetFileApplicationService datasetFileApplicationService;
+    private final StringRedisTemplate redisTemplate;
 
     @Value("${datamate.data-management.base-path:/dataset}")
     private String datasetBasePath;
@@ -76,6 +79,10 @@ public class DatasetApplicationService {
         return dataset;
     }
 
+    public String getDatasetPvcName() {
+        return redisTemplate.opsForValue().get(DATASET_PVC_NAME);
+    }
+
     public Dataset updateDataset(String datasetId, UpdateDatasetRequest updateDatasetRequest) {
         Dataset dataset = datasetRepository.getById(datasetId);
         BusinessAssert.notNull(dataset, DataManagementErrorCode.DATASET_NOT_FOUND);
@@ -130,7 +137,10 @@ public class DatasetApplicationService {
     public PagedResponse<DatasetResponse> getDatasets(DatasetPagingQuery query) {
         IPage<Dataset> page = new Page<>(query.getPage(), query.getSize());
         page = datasetRepository.findByCriteria(page, query);
-        return PagedResponse.of(DatasetConverter.INSTANCE.convertToResponse(page.getRecords()), page.getCurrent(), page.getTotal(), page.getPages());
+        String datasetPvcName = getDatasetPvcName();
+        List<DatasetResponse> datasetResponses = DatasetConverter.INSTANCE.convertToResponse(page.getRecords());
+        datasetResponses.forEach(dataset -> dataset.setPvcName(datasetPvcName));
+        return PagedResponse.of(datasetResponses, page.getCurrent(), page.getTotal(), page.getPages());
     }
 
     /**
diff --git a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/domain/model/dataset/FileTag.java b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/domain/model/dataset/FileTag.java
index 292aa08..b9f41da 100644
--- a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/domain/model/dataset/FileTag.java
+++ b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/domain/model/dataset/FileTag.java
@@ -28,7 +28,7 @@ public class FileTag {
 
     private String fromName;
 
-    public List<String> getTags() {
+    public List<String> getTagValues() {
         List<String> tags = new ArrayList<>();
         Object tagValues = values.get(type);
         if (tagValues instanceof List) {
@@ -40,9 +40,6 @@ public class FileTag {
         } else if (tagValues instanceof String) {
             tags.add((String) tagValues);
         }
-        if(StringUtils.isNotEmpty(fromName)) {
-            return tags.stream().map(tag -> fromName + " " + tag).toList();
-        }
         return tags;
     }
 }
diff --git a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/converter/DatasetConverter.java b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/converter/DatasetConverter.java
index 96c8de7..faa668f 100644
--- a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/converter/DatasetConverter.java
+++ b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/converter/DatasetConverter.java
@@ -70,8 +70,8 @@ public interface DatasetConverter {
      * @return 标签分布
      */
     @Named("getDistribution")
-    default Map<String, Long> getDistribution(List<DatasetFile> datasetFiles) {
-        Map<String, Long> distribution = new HashMap<>();
+    default Map<String, Map<String, Long>> getDistribution(List<DatasetFile> datasetFiles) {
+        Map<String, Map<String, Long>> distribution = new HashMap<>();
         if (CollectionUtils.isEmpty(datasetFiles)) {
             return distribution;
         }
@@ -81,7 +81,9 @@ public interface DatasetConverter {
                 return distribution;
             }
             for (FileTag tag : tags) {
-                tag.getTags().forEach(tagName -> distribution.put(tagName, distribution.getOrDefault(tagName, 0L) + 1));
+                Map<String, Long> tagValueMap = distribution.getOrDefault(tag.getFromName(), new HashMap<>());
+                tag.getTagValues().forEach(tagValue -> tagValueMap.put(tagValue, tagValueMap.getOrDefault(tagValue, 0L) + 1));
+                distribution.put(tag.getFromName(), tagValueMap);
             }
         }
         return distribution;
diff --git a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/dto/DatasetResponse.java b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/dto/DatasetResponse.java
index 7e43092..e039019 100644
--- a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/dto/DatasetResponse.java
+++ b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/dto/DatasetResponse.java
@@ -46,5 +46,7 @@ public class DatasetResponse {
     /** 更新者 */
     private String updatedBy;
     /** 分布 */
-    private Map<String, Long> distribution ;
+    private Map<String, Map<String, Long>> distribution;
+    /** 数据集pvc名称 */
+    private String pvcName;
 }
diff --git a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/rest/DatasetController.java b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/rest/DatasetController.java
index 173a4bb..aa05077 100644
--- a/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/rest/DatasetController.java
+++ b/backend/services/data-management-service/src/main/java/com/datamate/datamanagement/interfaces/rest/DatasetController.java
@@ -58,8 +58,9 @@ public class DatasetController {
      */
     @GetMapping("/{datasetId}")
     public DatasetResponse getDatasetById(@PathVariable("datasetId") String datasetId) {
-        Dataset dataset = datasetApplicationService.getDataset(datasetId);
-        return DatasetConverter.INSTANCE.convertToResponse(dataset);
+        DatasetResponse dataset = DatasetConverter.INSTANCE.convertToResponse(datasetApplicationService.getDataset(datasetId));
+        dataset.setPvcName(datasetApplicationService.getDatasetPvcName());
+        return dataset;
     }
 
     /**
diff --git a/deployment/docker/datamate/backend.conf b/deployment/docker/datamate/backend.conf
index 6b89b20..795c852 100644
--- a/deployment/docker/datamate/backend.conf
+++ b/deployment/docker/datamate/backend.conf
@@ -14,6 +14,13 @@ server {
         proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
     }
 
+    location /api/evaluation/ {
+        proxy_pass http://datamate-backend-python:18000/api/evaluation/;
+        proxy_set_header Host $host;
+        proxy_set_header X-Real-IP $remote_addr;
+        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+    }
+
     location /api/annotation/ {
         proxy_pass http://datamate-backend-python:18000/api/annotation/;
         proxy_set_header Host $host;
diff --git a/frontend/src/pages/DataEvaluation/Create/CreateTask.tsx b/frontend/src/pages/DataEvaluation/Create/CreateTask.tsx
index e98f246..4ad3823 100644
--- a/frontend/src/pages/DataEvaluation/Create/CreateTask.tsx
+++ b/frontend/src/pages/DataEvaluation/Create/CreateTask.tsx
@@ -1,574 +1,395 @@
-import { useState } from "react";
-import {
-  Button,
-  Card,
-  Badge,
-  Input,
-  Select,
-  Checkbox,
-  Form,
-  Typography,
-} from "antd";
-import {
-  PlusOutlined,
-  ArrowLeftOutlined,
-  EditOutlined,
-  SaveOutlined,
-  DeleteOutlined,
-} from "@ant-design/icons";
-import {
-  evaluationTemplates,
-  presetEvaluationDimensions,
-  sliceOperators,
-} from "@/mock/evaluation";
-import { useNavigate } from "react-router";
+import React, { useState, useEffect } from 'react';
+import { Button, Form, Input, Select, message, Modal, Row, Col, Table, Space } from 'antd';
+import { EyeOutlined } from '@ant-design/icons';
+import { queryDatasetsUsingGet } from "@/pages/DataManagement/dataset.api.ts";
+import { mapDataset } from "@/pages/DataManagement/dataset.const.tsx";
+import { queryModelListUsingGet } from "@/pages/SettingsPage/settings.apis.ts";
+import { ModelI } from "@/pages/SettingsPage/ModelAccess.tsx";
+import { createEvaluationTaskUsingPost } from "@/pages/DataEvaluation/evaluation.api.ts";
+import { queryPromptTemplatesUsingGet } from "@/pages/DataEvaluation/evaluation.api.ts";
+import PreviewPromptModal from "@/pages/DataEvaluation/Create/PreviewPrompt.tsx";
 
-const { Title, Paragraph } = Typography;
-const { Option } = Select;
+interface Dataset {
+  id: string;
+  name: string;
+  fileCount: number;
+  size: string;
+}
 
-const EvaluationTaskCreate = () => {
-  const navigate = useNavigate();
-  const [datasets, setDatasets] = useState([]);
-  const [selectedTemplate, setSelectedTemplate] =
-    useState<string>("dialogue_text");
-  const [allDimensions, setAllDimensions] = useState<EvaluationDimension[]>([
-    ...presetEvaluationDimensions,
-  ]);
-  const [editingDimension, setEditingDimension] = useState<string | null>(null);
-  const [newDimension, setNewDimension] = useState({
-    name: "",
-    description: "",
-  });
-  const [createForm, setCreateForm] = useState({
-    name: "",
-    datasetId: "",
-    evaluationType: "model" as "model" | "manual",
-    dimensions: [] as string[],
-    customDimensions: [] as EvaluationDimension[],
-    sliceConfig: {
-      threshold: 0.8,
-      sampleCount: 100,
-      method: "语义分割",
-    },
-    modelConfig: {
-      url: "",
-      apiKey: "",
-      prompt: "",
-      temperature: 0.3,
-      maxTokens: 2000,
-    },
+interface Dimension {
+  key: string;
+  dimension: string;
+  description: string;
+}
+
+interface PromptTemplate {
+  evalType: string;
+  prompt: string;
+  defaultDimensions: Dimension[];
+}
+
+interface CreateTaskModalProps {
+  visible: boolean;
+  onCancel: () => void;
+  onSuccess: () => void;
+}
+
+const TASK_TYPES = [
+  { label: 'QA评估', value: 'QA' },
+];
+
+const EVAL_METHODS = [
+  { label: '模型自动评估', value: 'AUTO' },
+];
+
+const DEFAULT_EVAL_METHOD = 'AUTO';
+const DEFAULT_TASK_TYPE = 'QA';
+
+const CreateTaskModal: React.FC<CreateTaskModalProps> = ({ visible, onCancel, onSuccess }) => {
+  const [form] = Form.useForm();
+  const [loading, setLoading] = useState(false);
+  const [datasets, setDatasets] = useState<Dataset[]>([]);
+  const [models, setModels] = useState<ModelI[]>([]);
+  const [dimensions, setDimensions] = useState<Dimension[]>([]);
+  const [newDimension, setNewDimension] = useState<Omit<Dimension, 'key'>>({
+    dimension: '',
+    description: ''
   });
+  const [taskType, setTaskType] = useState<string>("QA");
+  const [promptTemplates, setPromptTemplates] = useState<PromptTemplate[]>([]);
+  const [previewVisible, setPreviewVisible] = useState(false);
+  const [evaluationPrompt, setEvaluationPrompt] = useState('');
 
-  const handleTemplateChange = (templateKey: string) => {
-    setSelectedTemplate(templateKey);
-    const template =
-      evaluationTemplates[templateKey as keyof typeof evaluationTemplates];
-    if (template) {
-      const customDimensions = allDimensions.filter((d) => d.isCustom);
-      setAllDimensions([...template.dimensions, ...customDimensions]);
+  const handleAddDimension = () => {
+    if (!newDimension.dimension.trim()) {
+      message.warning('请输入维度名称');
+      return;
+    }
+    setDimensions([...dimensions, { ...newDimension, key: `dim-${Date.now()}` }]);
+    setNewDimension({ dimension: '', description: '' });
+  };
+
+  const handleDeleteDimension = (key: string) => {
+    if (dimensions.length <= 1) {
+      message.warning('至少需要保留一个评估维度');
+      return;
+    }
+    setDimensions(dimensions.filter(item => item.key !== key));
+  };
+
+  useEffect(() => {
+    if (visible) {
+      fetchDatasets().then();
+      fetchModels().then();
+      fetchPromptTemplates().then();
+    }
+  }, [visible]);
+
+  const fetchDatasets = async () => {
+    try {
+      const { data } = await queryDatasetsUsingGet({ page: 1, size: 1000 });
+      setDatasets(data.content.map(mapDataset) || []);
+    } catch (error) {
+      console.error('Error fetching datasets:', error);
+      message.error('获取数据集列表失败');
     }
   };
 
-  const handleAddCustomDimension = () => {
-    if (newDimension.name.trim() && newDimension.description.trim()) {
-      const customDimension: EvaluationDimension = {
-        id: `custom_${Date.now()}`,
-        name: newDimension.name.trim(),
-        description: newDimension.description.trim(),
-        category: "custom",
-        isCustom: true,
-        isEnabled: true,
-      };
-      setAllDimensions([...allDimensions, customDimension]);
-      setNewDimension({ name: "", description: "" });
+  const fetchModels = async () => {
+    try {
+      const { data } = await queryModelListUsingGet({ page: 0, size: 1000 });
+      setModels(data.content || []);
+    } catch (error) {
+      console.error('Error fetching models:', error);
+      message.error('获取模型列表失败');
     }
   };
 
-  const handleDimensionToggle = (id: string, checked: boolean) => {
-    setAllDimensions(
-      allDimensions.map((d) => (d.id === id ? { ...d, isEnabled: checked } : d))
-    );
-  };
-
-  const handleEditDimension = (
-    id: string,
-    field: "name" | "description",
-    value: string
-  ) => {
-    setAllDimensions(
-      allDimensions.map((d) => (d.id === id ? { ...d, [field]: value } : d))
-    );
-  };
-
-  const handleDeleteCustomDimension = (id: string) => {
-    setAllDimensions(allDimensions.filter((d) => d.id !== id));
-  };
-
-  const handleDeletePresetDimension = (id: string) => {
-    setAllDimensions(
-      allDimensions.map((d) => (d.id === id ? { ...d, isEnabled: false } : d))
-    );
-  };
-
-  const handleCreateTask = () => {
-    const selectedDataset = datasets.find((d) => d.id === createForm.datasetId);
-    if (!selectedDataset) return;
-
-    const enabledDimensions = allDimensions.filter((d) => d.isEnabled);
-    const presetDimensionIds = enabledDimensions
-      .filter((d) => !d.isCustom)
-      .map((d) => d.id);
-    const customDimensions = enabledDimensions.filter((d) => d.isCustom);
-
-    let finalPrompt = createForm.modelConfig.prompt;
-    if (createForm.evaluationType === "model" && !finalPrompt.trim()) {
-      finalPrompt = generateDefaultPrompt(selectedDataset.name);
-    }
-
-    const newTask: EvaluationTask = {
-      id: Date.now().toString(),
-      name: createForm.name,
-      datasetId: createForm.datasetId,
-      datasetName: selectedDataset.name,
-      evaluationType: createForm.evaluationType,
-      status: "pending",
-      progress: 0,
-      createdAt: new Date().toLocaleString(),
-      description: `${
-        createForm.evaluationType === "model" ? "模型自动" : "人工"
-      }评估${selectedDataset.name}`,
-      dimensions: presetDimensionIds,
-      customDimensions: customDimensions,
-      modelConfig:
-        createForm.evaluationType === "model"
-          ? {
-              ...createForm.modelConfig,
-              prompt: finalPrompt,
-            }
-          : undefined,
-      metrics: {
-        accuracy: 0,
-        completeness: 0,
-        consistency: 0,
-        relevance: 0,
-      },
-      issues: [],
-    };
-
-    // 重置表单
-    setCreateForm({
-      name: "",
-      datasetId: "",
-      evaluationType: "model",
-      dimensions: [],
-      customDimensions: [],
-      modelConfig: {
-        url: "",
-        apiKey: "",
-        prompt: "",
-        temperature: 0.3,
-        maxTokens: 2000,
-      },
+  const formatDimensionsForPrompt = (dimensions: Dimension[]) => {
+    let result = "\n";
+    dimensions.forEach((dim, index) => {
+      result += `### ${index + 1}. ${dim.dimension}\n**评估标准：**\n${dim.description}\n\n`;
     });
-    navigate("/data/evaluation");
+    return result;
   };
 
+  const formatResultExample = (dimensions: Dimension[]) => {
+    return dimensions.map(dim => `\n    "${dim.dimension}": "Y",`).join('');
+  };
+
+  const fetchPromptTemplates = async () => {
+    try {
+      const response = await queryPromptTemplatesUsingGet();
+      const templates: PromptTemplate[] = response.data?.templates
+      setPromptTemplates(templates)
+      if (taskType) {
+        const template = templates.find(t => t.evalType === taskType);
+        if (template) {
+          setDimensions(template.defaultDimensions.map((dim: any, index: number) => ({
+            key: `dim-${index}`,
+            dimension: dim.dimension,
+            description: dim.description
+          })));
+        }
+      }
+    } catch (error) {
+      console.error('Error fetching prompt templates:', error);
+      message.error('获取评估维度失败');
+    }
+  };
+
+  const generateEvaluationPrompt = () => {
+    if (dimensions.length === 0) {
+      message.warning('请先添加评估维度');
+      return;
+    }
+    const template = promptTemplates.find(t => t.evalType === taskType);
+    setEvaluationPrompt(template?.prompt.replace("{dimensions}", formatDimensionsForPrompt(dimensions))
+      .replace('{result_example}', formatResultExample(dimensions)));
+    setPreviewVisible(true);
+  };
+
+  const chatModelOptions = models
+    .filter((model) => model.type === "CHAT")
+    .map((model) => ({
+      label: `${model.modelName} (${model.provider})`,
+      value: model.id,
+    }));
+
+  const handleSubmit = async (values: any) => {
+    if (dimensions.length === 0) {
+      message.warning('请至少添加一个评估维度');
+      return;
+    }
+
+    try {
+      setLoading(true);
+      const { datasetId, modelId, ...rest } = values;
+      const selectedDataset = datasets.find(d => d.id === datasetId);
+      const selectedModel = models.find(d => d.id === modelId);
+
+      const payload = {
+        ...rest,
+        sourceType: 'DATASET',
+        sourceId: datasetId,
+        sourceName: selectedDataset?.name,
+        evalConfig: {
+          modelId: selectedModel?.id,
+          dimensions: dimensions.map(d => ({
+            dimension: d.dimension,
+            description: d.description
+          }))
+        }
+      };
+
+      await createEvaluationTaskUsingPost(payload);
+      message.success('评估任务创建成功');
+      onSuccess();
+      form.resetFields();
+      onCancel();
+    } catch (error: any) {
+      console.error('Error creating task:', error);
+      message.error(error.response?.data?.message || '创建评估任务失败');
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  const columns = [
+    {
+      title: '维度',
+      dataIndex: 'dimension',
+      key: 'dimension',
+      width: '30%',
+    },
+    {
+      title: '描述',
+      dataIndex: 'description',
+      key: 'description',
+      width: '60%',
+    },
+    {
+      title: '操作',
+      key: 'action',
+      width: '10%',
+      render: (_: any, record: any) => (
+        <Space size="middle">
+          <a
+            onClick={() => handleDeleteDimension(record.key)}
+            style={{ color: dimensions.length <= 1 ? '#ccc' : '#ff4d4f' }}
+            className={dimensions.length <= 1 ? 'disabled-link' : ''}
+          >
+            删除
+          </a>
+        </Space>
+      ),
+    },
+  ];
+
   return (
-    <div className="h-full">
-      {/* 页面头部 */}
-      <div className="flex items-center mb-2">
-        <Button
-          type="text"
-          icon={<ArrowLeftOutlined />}
-          onClick={() => navigate("/data/evaluation")}
-        ></Button>
-        <div className="text-xl font-bold">创建评估任务</div>
-      </div>
+    <Modal
+      title="创建评估任务"
+      open={visible}
+      onCancel={onCancel}
+      footer={null}
+      width={800}
+      destroyOnClose
+    >
+      <Form
+        form={form}
+        layout="vertical"
+        onFinish={handleSubmit}
+        initialValues={{
+          evalMethod: DEFAULT_EVAL_METHOD,
+          taskType: DEFAULT_TASK_TYPE,
+        }}
+      >
+        <Row gutter={16}>
+          <Col span={12}>
+            <Form.Item
+              label="任务名称"
+              name="name"
+              rules={[{ required: true, message: '请输入任务名称' }]}
+            >
+              <Input placeholder="输入任务名称" />
+            </Form.Item>
+          </Col>
+          <Col span={12}>
+            <Form.Item
+              label="任务类型"
+              name="taskType"
+              rules={[{ required: true, message: '请选择任务类型' }]}
+            >
+              <Select options={TASK_TYPES} />
+            </Form.Item>
+          </Col>
+        </Row>
 
-      <Form layout="vertical">
-        {/* 基本信息 */}
-        <Card title="基本信息" style={{ marginBottom: 24 }}>
-          <Form.Item label="任务名称" required>
-            <Input
-              value={createForm.name}
-              onChange={(e) =>
-                setCreateForm({ ...createForm, name: e.target.value })
-              }
-              placeholder="输入任务名称"
-            />
-          </Form.Item>
-          <Form.Item label="选择数据集" required>
-            <Select
-              value={createForm.datasetId || undefined}
-              onChange={(value) =>
-                setCreateForm({ ...createForm, datasetId: value })
-              }
-              placeholder="选择要评估的数据集"
-            >
-              {datasets.map((dataset) => (
-                <Option key={dataset.id} value={dataset.id}>
-                  {dataset.name}（{dataset.fileCount} 文件 • {dataset.size}）
-                </Option>
-              ))}
-            </Select>
-          </Form.Item>
-          <Form.Item label="评估方式" required>
-            <Select
-              value={createForm.evaluationType}
-              onChange={(value: "model" | "manual") =>
-                setCreateForm({ ...createForm, evaluationType: value })
-              }
-            >
-              <Option value="model">模型自动评估</Option>
-              <Option value="manual">人工评估</Option>
-            </Select>
-          </Form.Item>
-        </Card>
-
-        {/* 算子配置 */}
-        <Card title="切片算子配置" style={{ marginBottom: 24 }}>
-          <Form.Item label="切片算子">
-            <Select
-              value={createForm.sliceConfig.method}
-              onChange={(value) =>
-                setCreateForm({
-                  ...createForm,
-                  sliceConfig: { ...createForm.sliceConfig, method: value },
-                })
-              }
-              placeholder="选择切片算子"
-            >
-              {sliceOperators.map((operator) => (
-                <Option key={operator.id} value={operator.name}>
-                  {operator.name}{" "}
-                  <Badge style={{ marginLeft: 8 }} count={operator.type} />
-                </Option>
-              ))}
-            </Select>
-          </Form.Item>
-          <Form.Item label="分隔符">
-            <Input
-              placeholder="输入分隔符，如 \\n\\n"
-              value={createForm.sliceConfig.delimiter}
-              onChange={(e) =>
-                setCreateForm({
-                  ...createForm,
-                  sliceConfig: {
-                    ...createForm.sliceConfig,
-                    delimiter: e.target.value,
-                  },
-                })
-              }
-            />
-          </Form.Item>
-          <Form.Item label="分块大小">
-            <Input
-              type="number"
-              value={createForm.sliceConfig.chunkSize}
-              onChange={(e) =>
-                setCreateForm({
-                  ...createForm,
-                  sliceConfig: {
-                    ...createForm.sliceConfig,
-                    chunkSize: Number(e.target.value),
-                  },
-                })
-              }
-            />
-          </Form.Item>
-          <Form.Item label="重叠长度">
-            <Input
-              type="number"
-              value={createForm.sliceConfig.overlapLength}
-              onChange={(e) =>
-                setCreateForm({
-                  ...createForm,
-                  sliceConfig: {
-                    ...createForm.sliceConfig,
-                    overlapLength: Number(e.target.value),
-                  },
-                })
-              }
-            />
-          </Form.Item>
-          <Form.Item label="抽样比例">
-            <Input
-              type="number"
-              value={createForm.sliceConfig.threshold}
-              onChange={(e) =>
-                setCreateForm({
-                  ...createForm,
-                  sliceConfig: {
-                    ...createForm.sliceConfig,
-                    threshold: Number(e.target.value),
-                  },
-                })
-              }
-            />
-          </Form.Item>
-        </Card>
-
-        {/* 评估维度配置 */}
-        <Card
-          title={
-            <div
-              style={{
-                display: "flex",
-                justifyContent: "space-between",
-                alignItems: "center",
-              }}
-            >
-              <span>评估维度配置</span>
-              <div style={{ display: "flex", alignItems: "center", gap: 8 }}>
-                <Select
-                  value={selectedTemplate}
-                  onChange={handleTemplateChange}
-                  style={{ width: 160 }}
-                >
-                  {Object.entries(evaluationTemplates).map(
-                    ([key, template]) => (
-                      <Option key={key} value={key}>
-                        {template.name}
-                      </Option>
-                    )
-                  )}
-                </Select>
-                <Badge
-                  count={allDimensions.filter((d) => d.isEnabled).length}
-                  style={{ background: "#f0f0f0", color: "#333" }}
-                />
-              </div>
-            </div>
-          }
-          style={{ marginBottom: 24 }}
+        <Form.Item
+          label="任务描述"
+          name="description"
         >
-          {/* 维度表格 */}
-          <div
-            style={{
-              border: "1px solid #f0f0f0",
-              borderRadius: 6,
-              marginBottom: 16,
-            }}
-          >
-            <div
-              style={{
-                background: "#fafafa",
-                padding: "8px 12px",
-                borderBottom: "1px solid #f0f0f0",
-                fontWeight: 500,
-                fontSize: 13,
-              }}
+          <Input.TextArea placeholder="输入任务描述（可选）" rows={3} />
+        </Form.Item>
+
+        <Row gutter={16}>
+          <Col span={12}>
+            <Form.Item
+              label="选择数据集"
+              name="datasetId"
+              rules={[{ required: true, message: '请选择数据集' }]}
             >
-              <div style={{ display: "flex", alignItems: "center" }}>
-                <div style={{ width: 60 }}>启用</div>
-                <div style={{ width: 160 }}>维度名称</div>
-                <div style={{ flex: 1 }}>描述</div>
-                <div style={{ width: 120 }}>操作</div>
-              </div>
-            </div>
-            <div style={{ maxHeight: 320, overflowY: "auto" }}>
-              {allDimensions.map((dimension) => (
-                <div
-                  key={dimension.id}
-                  style={{
-                    display: "flex",
-                    alignItems: "center",
-                    padding: "8px 12px",
-                    borderBottom: "1px solid #f5f5f5",
-                  }}
-                >
-                  <div style={{ width: 60 }}>
-                    <Checkbox
-                      checked={dimension.isEnabled}
-                      onChange={(e) =>
-                        handleDimensionToggle(dimension.id, e.target.checked!)
-                      }
-                    />
-                  </div>
-                  <div style={{ width: 160 }}>
-                    {editingDimension === dimension.id && dimension.isCustom ? (
-                      <Input
-                        value={dimension.name}
-                        onChange={(e) =>
-                          handleEditDimension(
-                            dimension.id,
-                            "name",
-                            e.target.value
-                          )
-                        }
-                        size="small"
-                      />
-                    ) : (
-                      <span style={{ fontWeight: 500 }}>
-                        {dimension.name}
-                        {dimension.isCustom && (
-                          <Badge
-                            style={{
-                              marginLeft: 4,
-                              background: "#f9f0ff",
-                              color: "#722ed1",
-                            }}
-                            count="自定义"
-                          />
-                        )}
-                      </span>
-                    )}
-                  </div>
-                  <div style={{ flex: 1 }}>
-                    {editingDimension === dimension.id && dimension.isCustom ? (
-                      <Input
-                        value={dimension.description}
-                        onChange={(e) =>
-                          handleEditDimension(
-                            dimension.id,
-                            "description",
-                            e.target.value
-                          )
-                        }
-                        size="small"
-                      />
-                    ) : (
-                      <span style={{ color: "#888" }}>
-                        {dimension.description}
-                      </span>
-                    )}
-                  </div>
-                  <div style={{ width: 120 }}>
-                    {editingDimension === dimension.id && dimension.isCustom ? (
-                      <Button
-                        type="text"
-                        icon={<SaveOutlined />}
-                        size="small"
-                        onClick={() => setEditingDimension(null)}
-                      />
-                    ) : (
-                      dimension.isCustom && (
-                        <Button
-                          type="text"
-                          icon={<EditOutlined />}
-                          size="small"
-                          onClick={() => setEditingDimension(dimension.id)}
-                        />
-                      )
-                    )}
-                    <Button
-                      type="text"
-                      icon={<DeleteOutlined />}
-                      size="small"
-                      danger
-                      onClick={() =>
-                        dimension.isCustom
-                          ? handleDeleteCustomDimension(dimension.id)
-                          : handleDeletePresetDimension(dimension.id)
-                      }
-                      disabled={
-                        allDimensions.filter((d) => d.isEnabled).length <= 1 &&
-                        dimension.isEnabled
-                      }
-                    />
-                  </div>
+              <Select
+                placeholder="请选择要评估的数据集"
+                showSearch
+                optionFilterProp="label"
+              >
+                {datasets.map((dataset) => (
+                  <Select.Option key={dataset.id} value={dataset.id} label={dataset.name}>
+                    <div className="flex justify-between w-full">
+                      <span>{dataset.name}</span>
+                      <span className="text-gray-500">{dataset.size}</span>
+                    </div>
+                  </Select.Option>
+                ))}
+              </Select>
+            </Form.Item>
+          </Col>
+          <Col span={12}>
+            <Form.Item
+              label="评估方式"
+              name="evalMethod"
+              initialValue={DEFAULT_EVAL_METHOD}
+            >
+              <Select options={EVAL_METHODS} />
+            </Form.Item>
+          </Col>
+        </Row>
+
+        <Form.Item
+          noStyle
+          shouldUpdate={(prevValues, currentValues) =>
+            prevValues.evalMethod !== currentValues.evalMethod
+          }
+        >
+          {({ getFieldValue }) => getFieldValue('evalMethod') === 'AUTO' && (
+            <>
+              <Form.Item
+                label="评估模型"
+                name="modelId"
+                rules={[{ required: true, message: '请选择评估模型' }]}
+              >
+                <Select
+                  placeholder="请选择模型"
+                  options={chatModelOptions}
+                  showSearch
+                  optionFilterProp="label"
+                />
+              </Form.Item>
+
+              <Form.Item label="评估维度">
+                <Table
+                  columns={columns}
+                  dataSource={dimensions}
+                  pagination={false}
+                  size="small"
+                  rowKey="key"
+                />
+                <div style={{ display: 'flex', gap: 8, marginBottom: 16 }}>
+                  <Input
+                    placeholder="输入维度名称"
+                    value={newDimension.dimension}
+                    onChange={(e) => setNewDimension({...newDimension, dimension: e.target.value})}
+                    style={{ flex: 1 }}
+                  />
+                  <Input
+                    placeholder="输入维度描述"
+                    value={newDimension.description}
+                    onChange={(e) => setNewDimension({...newDimension, description: e.target.value})}
+                    style={{ flex: 2 }}
+                  />
+                  <Button
+                    type="primary"
+                    onClick={handleAddDimension}
+                    disabled={!newDimension.dimension.trim()}
+                  >
+                    添加维度
+                  </Button>
                 </div>
-              ))}
-            </div>
-          </div>
-          {/* 添加自定义维度 */}
-          <div style={{ background: "#fafafa", borderRadius: 6, padding: 16 }}>
-            <div style={{ fontWeight: 500, marginBottom: 8 }}>
-              添加自定义维度
-            </div>
-            <Input
-              value={newDimension.name}
-              onChange={(e) =>
-                setNewDimension({ ...newDimension, name: e.target.value })
-              }
-              placeholder="维度名称"
-              style={{ width: 180, marginRight: 8 }}
-              size="small"
-            />
-            <Input
-              value={newDimension.description}
-              onChange={(e) =>
-                setNewDimension({
-                  ...newDimension,
-                  description: e.target.value,
-                })
-              }
-              placeholder="维度描述"
-              style={{ width: 260, marginRight: 8 }}
-              size="small"
-            />
+              </Form.Item>
+            </>
+          )}
+        </Form.Item>
+        <Form.Item>
+          <div style={{ display: 'flex', justifyContent: 'space-between', marginTop: '16px' }}>
             <Button
-              icon={<PlusOutlined />}
-              onClick={handleAddCustomDimension}
-              disabled={
-                !newDimension.name.trim() || !newDimension.description.trim()
-              }
-              size="small"
+              type="link"
+              icon={<EyeOutlined />}
+              onClick={generateEvaluationPrompt}
             >
-              添加维度
+              查看评估提示词
             </Button>
           </div>
-        </Card>
+        </Form.Item>
 
-        {/* 模型配置（仅在选择模型评估时显示） */}
-        {createForm.evaluationType === "model" && (
-          <Card title="模型配置" style={{ marginBottom: 24 }}>
-            <Form.Item label="模型 URL" required>
-              <Input
-                value={createForm.modelConfig.url}
-                onChange={(e) =>
-                  setCreateForm({
-                    ...createForm,
-                    modelConfig: {
-                      ...createForm.modelConfig,
-                      url: e.target.value,
-                    },
-                  })
-                }
-                placeholder="https://api.openai.com/v1/chat/completions"
-              />
-            </Form.Item>
-            <Form.Item label="API Key" required>
-              <Input.Password
-                value={createForm.modelConfig.apiKey}
-                onChange={(e) =>
-                  setCreateForm({
-                    ...createForm,
-                    modelConfig: {
-                      ...createForm.modelConfig,
-                      apiKey: e.target.value,
-                    },
-                  })
-                }
-                placeholder="sk-***"
-              />
-            </Form.Item>
-          </Card>
-        )}
-
-        {/* 操作按钮 */}
         <Form.Item>
-          <div style={{ display: "flex", justifyContent: "flex-end", gap: 12 }}>
-            <Button onClick={() => navigate("/data/evaluation")}>取消</Button>
-            <Button
-              type="primary"
-              onClick={handleCreateTask}
-              disabled={
-                !createForm.name ||
-                !createForm.datasetId ||
-                allDimensions.filter((d) => d.isEnabled).length === 0 ||
-                (createForm.evaluationType === "model" &&
-                  (!createForm.modelConfig.url ||
-                    !createForm.modelConfig.apiKey))
-              }
-            >
-              创建评估任务
+          <div style={{ textAlign: 'right' }}>
+            <Button onClick={onCancel} style={{ marginRight: 8 }}>
+              取消
+            </Button>
+            <Button type="primary" htmlType="submit" loading={loading}>
+              创建任务
             </Button>
           </div>
         </Form.Item>
       </Form>
-    </div>
+      <PreviewPromptModal
+        previewVisible={previewVisible}
+        onCancel={() => setPreviewVisible(false)}
+        evaluationPrompt={evaluationPrompt}
+      />
+    </Modal>
   );
 };
 
-export default EvaluationTaskCreate;
+export default CreateTaskModal;
diff --git a/frontend/src/pages/DataEvaluation/Create/PreviewPrompt.tsx b/frontend/src/pages/DataEvaluation/Create/PreviewPrompt.tsx
new file mode 100644
index 0000000..72c9cba
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/Create/PreviewPrompt.tsx
@@ -0,0 +1,42 @@
+import React from 'react';
+import { Button, message, Modal } from 'antd';
+
+interface PreviewPromptModalProps {
+  previewVisible: boolean;
+  onCancel: () => void;
+  evaluationPrompt: string;
+}
+
+const PreviewPromptModal: React.FC<PreviewPromptModalProps> = ({ previewVisible, onCancel, evaluationPrompt }) => {
+  return (
+    <Modal
+      title="评估提示词预览"
+      open={previewVisible}
+      onCancel={onCancel}
+      footer={[
+        <Button key="copy" onClick={() => {
+          navigator.clipboard.writeText(evaluationPrompt).then();
+          message.success('已复制到剪贴板');
+        }}>
+          复制
+        </Button>,
+        <Button key="close" type="primary" onClick={onCancel}>
+          关闭
+        </Button>
+      ]}
+      width={800}
+    >
+      <div style={{
+        background: '#f5f5f5',
+        padding: '16px',
+        borderRadius: '4px',
+        whiteSpace: 'pre-wrap',
+        fontFamily: 'monospace'
+      }}>
+        {evaluationPrompt}
+      </div>
+    </Modal>
+  )
+}
+
+export default PreviewPromptModal;
diff --git a/frontend/src/pages/DataEvaluation/Detail/TaskDetail.tsx b/frontend/src/pages/DataEvaluation/Detail/TaskDetail.tsx
new file mode 100644
index 0000000..940b472
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/Detail/TaskDetail.tsx
@@ -0,0 +1,152 @@
+import { Link, useParams } from "react-router";
+import { Tabs, Spin, message, Breadcrumb } from 'antd';
+import { LayoutList, Clock } from "lucide-react";
+import { useEffect, useState } from 'react';
+import { getEvaluationTaskByIdUsingGet, queryEvaluationItemsUsingGet } from '../evaluation.api';
+import { EvaluationTask, EvaluationStatus } from '../evaluation.model';
+import DetailHeader from "@/components/DetailHeader.tsx";
+import {TaskStatusMap} from "@/pages/DataCleansing/cleansing.const.tsx";
+import EvaluationItems from "@/pages/DataEvaluation/Detail/components/EvaluationItems.tsx";
+import Overview from "@/pages/DataEvaluation/Detail/components/Overview.tsx";
+
+const tabList = [
+  {
+    key: "overview",
+    label: "概览",
+  },
+  {
+    key: "evaluationItems",
+    label: "评估详情",
+  }
+];
+
+interface EvaluationItem {
+  id: string;
+  content: string;
+  status: EvaluationStatus;
+  score?: number;
+  dimensions: {
+    id: string;
+    name: string;
+    score: number;
+  }[];
+  createdAt: string;
+}
+
+const EvaluationDetailPage: React.FC = () => {
+  const { id } = useParams<{ id: string }>();
+  const [loading, setLoading] = useState<boolean>(true);
+  const [task, setTask] = useState<EvaluationTask | null>(null);
+  const [items, setItems] = useState<EvaluationItem[]>([]);
+  const [activeTab, setActiveTab] = useState("overview");
+  const [pagination, setPagination] = useState({
+    current: 1,
+    pageSize: 10,
+    total: 0,
+  });
+
+  const fetchTaskDetail = async () => {
+    try {
+      const response = await getEvaluationTaskByIdUsingGet(id);
+      setTask(response.data);
+    } catch (error) {
+      message.error('Failed to fetch task details');
+      console.error('Error fetching task detail:', error);
+    }
+  };
+
+  const fetchEvaluationItems = async (page = 1, pageSize = 10) => {
+    try {
+      const response = await queryEvaluationItemsUsingGet({
+        taskId: id,
+        page: page,
+        size: pageSize,
+      });
+      setItems(response.data.content || []);
+      setPagination({
+        ...pagination,
+        current: page,
+        total: response.data.totalElements || 0,
+      });
+    } catch (error) {
+      message.error('Failed to fetch evaluation items');
+      console.error('Error fetching evaluation items:', error);
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  useEffect(() => {
+    if (id) {
+      setLoading(true);
+      Promise.all([
+        fetchTaskDetail(),
+        fetchEvaluationItems(1, pagination.pageSize),
+      ]).finally(() => setLoading(false));
+    }
+  }, [id]);
+
+  if (loading && !task) {
+    return (
+      <div style={{ textAlign: 'center', padding: '100px 0' }}>
+        <Spin size="large" />
+      </div>
+    );
+  }
+
+  if (!task) {
+    return <div>Task not found</div>;
+  }
+
+  const breadItems = [
+    {
+      title: <Link to="/data/evaluation">数据评估</Link>,
+    },
+    {
+      title: "数据评估详情",
+    },
+  ];
+
+  const headerData = {
+    ...task,
+    icon: <LayoutList className="w-8 h-8" />,
+    status: TaskStatusMap[task?.status],
+    createdAt: task?.createdAt,
+    lastUpdated: task?.updatedAt,
+  };
+
+  // 基本信息描述项
+  const statistics = [
+    {
+      icon: <Clock className="text-blue-400 w-4 h-4" />,
+      key: "time",
+      value: task?.updatedAt,
+    },
+  ];
+
+  const operations = []
+
+  return (
+    <>
+      <Breadcrumb items={breadItems} />
+      <div className="mb-4 mt-4">
+        <div className="mb-4 mt-4">
+          <DetailHeader
+            data={headerData}
+            statistics={statistics}
+            operations={operations}
+          />
+        </div>
+      </div>
+      <div className="flex-overflow-auto p-6 pt-2 bg-white rounded-md shadow">
+        <Tabs activeKey={activeTab} items={tabList} onChange={setActiveTab} />
+        <div className="h-full overflow-auto">
+          {activeTab === "overview" && <Overview task={task} />}
+          {activeTab === "evaluationItems" && <EvaluationItems task={task} />}
+        </div>
+      </div>
+    </>
+  );
+};
+
+export default EvaluationDetailPage;
diff --git a/frontend/src/pages/DataEvaluation/Detail/components/EvaluationItems.tsx b/frontend/src/pages/DataEvaluation/Detail/components/EvaluationItems.tsx
new file mode 100644
index 0000000..dd32a91
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/Detail/components/EvaluationItems.tsx
@@ -0,0 +1,271 @@
+import { useEffect, useState } from 'react';
+import { Table, Typography, Button, Space, Spin, Empty, message, Tooltip } from 'antd';
+import { FolderOpen, FileText, ArrowLeft } from 'lucide-react';
+import { queryEvaluationFilesUsingGet, queryEvaluationItemsUsingGet } from '../../evaluation.api';
+
+const { Text } = Typography;
+
+const COLUMN_WIDTH = 520;
+ const MONO_FONT = 'ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono", "Courier New", monospace';
+ const codeBlockStyle = {
+   fontFamily: MONO_FONT,
+   fontSize: 12,
+   lineHeight: '20px',
+   color: '#334155',
+   backgroundColor: '#f8fafc',
+   border: '1px solid #f0f0f0',
+   borderRadius: 6,
+   padding: 8,
+ } as const;
+
+type EvalFile = {
+  taskId: string;
+  fileId: string;
+  fileName: string;
+  totalCount: number;
+  evaluatedCount: number;
+  pendingCount: number;
+};
+
+type EvalItem = {
+  id: string;
+  taskId: string;
+  itemId: string;
+  fileId: string;
+  evalContent: any;
+  evalScore?: number | null;
+  evalResult: any;
+  status?: string;
+};
+
+export default function EvaluationItems({ task }: { task: any }) {
+  const [loadingFiles, setLoadingFiles] = useState<boolean>(false);
+  const [files, setFiles] = useState<EvalFile[]>([]);
+  const [filePagination, setFilePagination] = useState({ current: 1, pageSize: 10, total: 0 });
+
+  const [selectedFile, setSelectedFile] = useState<{ fileId: string; fileName: string } | null>(null);
+  const [loadingItems, setLoadingItems] = useState<boolean>(false);
+  const [items, setItems] = useState<EvalItem[]>([]);
+  const [itemPagination, setItemPagination] = useState({ current: 1, pageSize: 10, total: 0 });
+
+  // Fetch files list
+  useEffect(() => {
+    if (!task?.id || selectedFile) return;
+    const fetchFiles = async () => {
+      setLoadingFiles(true);
+      try {
+        const res = await queryEvaluationFilesUsingGet({ taskId: task.id, page: filePagination.current, size: filePagination.pageSize });
+        const data = res?.data;
+        const list: EvalFile[] = data?.content || [];
+        setFiles(list);
+        setFilePagination((p) => ({ ...p, total: data?.totalElements || 0 }));
+      } catch (e) {
+        message.error('加载评估文件失败');
+        console.error(e);
+      } finally {
+        setLoadingFiles(false);
+      }
+    };
+    fetchFiles();
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [task?.id, filePagination.current, filePagination.pageSize, selectedFile]);
+
+  // Fetch items of selected file
+  useEffect(() => {
+    if (!task?.id || !selectedFile) return;
+    const fetchItems = async () => {
+      setLoadingItems(true);
+      try {
+        const res = await queryEvaluationItemsUsingGet({
+          taskId: task.id,
+          page: itemPagination.current,
+          size: itemPagination.pageSize,
+          file_id: selectedFile.fileId,
+        });
+        const data = res?.data;
+        const list: EvalItem[] = data?.content || [];
+        setItems(list);
+        setItemPagination((p) => ({ ...p, total: data?.totalElements || 0 }));
+      } catch (e) {
+        message.error('加载评估条目失败');
+        console.error(e);
+      } finally {
+        setLoadingItems(false);
+      }
+    };
+    fetchItems();
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [task?.id, selectedFile?.fileId, itemPagination.current, itemPagination.pageSize]);
+
+  const fileColumns = [
+    {
+      title: '文件名',
+      dataIndex: 'fileName',
+      key: 'fileName',
+      render: (_: any, record: EvalFile) => (
+        <Space onClick={(e) => { e.stopPropagation(); setSelectedFile({ fileId: record.fileId, fileName: record.fileName }); }} style={{ cursor: 'pointer' }}>
+          <FolderOpen size={16} />
+          <Button type="link" style={{ padding: 0 }}>{record.fileName}</Button>
+        </Space>
+      ),
+    },
+    {
+      title: '总条目',
+      dataIndex: 'totalCount',
+      key: 'totalCount',
+      width: 120,
+    },
+    {
+      title: '已评估',
+      dataIndex: 'evaluatedCount',
+      key: 'evaluatedCount',
+      width: 120,
+    },
+    {
+      title: '待评估',
+      dataIndex: 'pendingCount',
+      key: 'pendingCount',
+      width: 120,
+    },
+  ];
+
+  const renderEvalObject = (rec: EvalItem) => {
+    const c = rec.evalContent;
+    let jsonString = '';
+    try {
+      if (typeof c === 'string') {
+        // 尝试将字符串解析为 JSON，失败则按原字符串显示
+        try {
+          jsonString = JSON.stringify(JSON.parse(c), null, 2);
+        } catch {
+          jsonString = JSON.stringify({ value: c }, null, 2);
+        }
+      } else {
+        jsonString = JSON.stringify(c, null, 2);
+      }
+    } catch {
+      jsonString = 'null';
+    }
+    return (
+      <Tooltip
+        color="#fff"
+        title={<pre style={{ ...codeBlockStyle, margin: 0, maxWidth: COLUMN_WIDTH, whiteSpace: 'pre-wrap' }}>{jsonString}</pre>}
+        overlayInnerStyle={{ maxHeight: 600, overflow: 'auto', width: COLUMN_WIDTH }}
+      >
+        <Typography.Paragraph
+          style={{ margin: 0, whiteSpace: 'pre-wrap', fontFamily: MONO_FONT, fontSize: 12, lineHeight: '20px', color: '#334155' }}
+          ellipsis={{ rows: 6 }}
+        >
+          <pre style={{ ...codeBlockStyle, whiteSpace: 'pre-wrap', margin: 0 }}>{jsonString}</pre>
+        </Typography.Paragraph>
+      </Tooltip>
+    );
+  };
+
+  const renderEvalResult = (rec: EvalItem) => {
+    const r = rec.evalResult;
+    let jsonString = '';
+    try {
+      if (typeof r === 'string') {
+        try {
+          jsonString = JSON.stringify(JSON.parse(r), null, 2);
+        } catch {
+          jsonString = JSON.stringify({ value: r, score: rec.evalScore ?? undefined }, null, 2);
+        }
+      } else {
+        const withScore = rec.evalScore !== undefined && rec.evalScore !== null ? { ...r, evalScore: rec.evalScore } : r;
+        jsonString = JSON.stringify(withScore, null, 2);
+      }
+    } catch {
+      jsonString = 'null';
+    }
+    // 判空展示未评估
+    const isEmpty = !r || (typeof r === 'string' && r.trim() === '') || (typeof r === 'object' && r !== null && Object.keys(r).length === 0);
+    if (isEmpty) {
+      return <Text type="secondary">未评估</Text>;
+    }
+    return (
+      <Tooltip
+        color="#fff"
+        title={<pre style={{ ...codeBlockStyle, margin: 0, maxWidth: 800, whiteSpace: 'pre-wrap' }}>{jsonString}</pre>}
+        overlayInnerStyle={{ maxHeight: 600, overflow: 'auto' }}
+      >
+        <Typography.Paragraph
+          style={{ margin: 0, whiteSpace: 'pre-wrap', fontFamily: MONO_FONT, fontSize: 12, lineHeight: '20px', color: '#334155' }}
+          ellipsis={{ rows: 6 }}
+        >
+          <pre style={{ ...codeBlockStyle, whiteSpace: 'pre-wrap', margin: 0 }}>{jsonString}</pre>
+        </Typography.Paragraph>
+      </Tooltip>
+    );
+  };
+
+  const itemColumns = [
+    {
+      title: '评估对象',
+      dataIndex: 'evalContent',
+      key: 'evalContent',
+      render: (_: any, record: EvalItem) => renderEvalObject(record),
+      width: COLUMN_WIDTH,
+    },
+    {
+      title: '评估结果',
+      dataIndex: 'evalResult',
+      key: 'evalResult',
+      render: (_: any, record: EvalItem) => renderEvalResult(record),
+      width: COLUMN_WIDTH,
+    },
+  ];
+
+  if (!task?.id) return <Empty description="任务不存在" />;
+
+  return (
+    <div className="flex flex-col gap-4">
+      {!selectedFile ? (
+        <Table
+          rowKey={(r: EvalFile) => r.fileId}
+          columns={fileColumns}
+          dataSource={files}
+          loading={loadingFiles}
+          size="middle"
+          onRow={(record) => ({ onClick: () => setSelectedFile({ fileId: record.fileId, fileName: record.fileName }) })}
+          pagination={{
+            current: filePagination.current,
+            pageSize: filePagination.pageSize,
+            total: filePagination.total,
+            onChange: (current, pageSize) => setFilePagination({ current, pageSize, total: filePagination.total }),
+          }}
+        />
+      ) : (
+        <div className="flex flex-col gap-3">
+          <div className="sticky top-0 z-10 bg-white py-2" style={{ borderBottom: '1px solid #f0f0f0' }}>
+            <Space wrap>
+              <Button icon={<ArrowLeft size={16} />} onClick={() => { setSelectedFile(null); setItems([]); }}>
+                返回文件列表
+              </Button>
+              <Space>
+                <FileText size={16} />
+                <Text strong>{selectedFile.fileName}</Text>
+                <Text type="secondary">文件ID：{selectedFile.fileId}</Text>
+                <Text type="secondary">共 {itemPagination.total} 条</Text>
+              </Space>
+            </Space>
+          </div>
+          <Table
+            rowKey={(r: EvalItem) => r.id}
+            columns={itemColumns}
+            dataSource={items}
+            loading={loadingItems}
+            size="middle"
+            pagination={{
+              current: itemPagination.current,
+              pageSize: itemPagination.pageSize,
+              total: itemPagination.total,
+              onChange: (current, pageSize) => setItemPagination({ current, pageSize, total: itemPagination.total }),
+            }}
+          />
+        </div>
+      )}
+    </div>
+  );
+}
diff --git a/frontend/src/pages/DataEvaluation/Detail/components/Overview.tsx b/frontend/src/pages/DataEvaluation/Detail/components/Overview.tsx
new file mode 100644
index 0000000..29cb639
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/Detail/components/Overview.tsx
@@ -0,0 +1,122 @@
+import { useState } from 'react';
+import { Descriptions, Empty, DescriptionsProps, Table, Button, message } from 'antd';
+import { CheckCircle, XCircle, Clock as ClockIcon } from 'lucide-react';
+import { EyeOutlined } from '@ant-design/icons';
+import { EvaluationStatus } from '../../evaluation.model';
+import PreviewPromptModal from "@/pages/DataEvaluation/Create/PreviewPrompt.tsx";
+
+const statusMap = {
+  [EvaluationStatus.PENDING]: { color: 'blue', text: '待处理', icon: <ClockIcon className="mr-1" size={14} /> },
+  [EvaluationStatus.RUNNING]: { color: 'processing', text: '进行中', icon: <ClockIcon className="mr-1" size={14} /> },
+  [EvaluationStatus.COMPLETED]: { color: 'success', text: '已完成', icon: <CheckCircle className="mr-1" size={14} /> },
+  [EvaluationStatus.FAILED]: { color: 'error', text: '失败', icon: <XCircle className="mr-1" size={14} /> },
+  [EvaluationStatus.PAUSED]: { color: 'warning', text: '已暂停', icon: <ClockIcon className="mr-1" size={14} /> },
+};
+
+const Overview = ({ task }) => {
+  const [previewVisible, setPreviewVisible] = useState(false);
+  if (!task) {
+    return <Empty description="未找到评估任务信息" />;
+  }
+
+  const generateEvaluationPrompt = () => {
+    setPreviewVisible(true);
+  };
+
+  const statusInfo = statusMap[task.status] || { color: 'default', text: '未知状态' };
+
+  // 基本信息
+  const items: DescriptionsProps["items"] = [
+    {
+      key: "id",
+      label: "ID",
+      children: task.id,
+    },
+    {
+      key: "name",
+      label: "名称",
+      children: task.name,
+    },
+    {
+      key: "status",
+      label: "状态",
+      children: statusInfo.text || "未知",
+    },
+    {
+      key: "createdBy",
+      label: "创建者",
+      children: task.createdBy || "未知",
+    },
+    {
+      key: "createdAt",
+      label: "创建时间",
+      children: task.createdAt,
+    },
+    {
+      key: "updatedAt",
+      label: "更新时间",
+      children: task.updatedAt,
+    },
+    {
+      key: "description",
+      label: "描述",
+      children: task.description || "无",
+    },
+  ];
+
+  const columns = [
+    {
+      title: '维度',
+      dataIndex: 'dimension',
+      key: 'dimension',
+      width: '30%',
+    },
+    {
+      title: '描述',
+      dataIndex: 'description',
+      key: 'description',
+      width: '60%',
+    },
+  ];
+
+  return (
+    <>
+      <div className=" flex flex-col gap-4">
+        {/* 基本信息 */}
+        <Descriptions
+          title="基本信息"
+          layout="vertical"
+          size="small"
+          items={items}
+          column={5}
+        />
+        <h2 className="text-base font-semibold mt-8">评估维度</h2>
+        <div className="overflow-x-auto">
+          <Table
+            size="middle"
+            rowKey="id"
+            columns={columns}
+            dataSource={task?.evalConfig?.dimensions}
+            scroll={{ x: "max-content", y: 600 }}
+          />
+        </div>
+        <div className="flex items-center gap-2 p-3 bg-blue-50 rounded-lg border border-blue-200">
+          <Button
+            type="link"
+            icon={<EyeOutlined />}
+            onClick={generateEvaluationPrompt}
+          >
+            查看评估提示词
+          </Button>
+        </div>
+        <PreviewPromptModal
+          previewVisible={previewVisible}
+          onCancel={() => setPreviewVisible(false)}
+          evaluationPrompt={task?.evalPrompt}
+        />
+      </div>
+    </>
+  );
+};
+
+export default Overview;
diff --git a/frontend/src/pages/DataEvaluation/Home/DataEvaluation.tsx b/frontend/src/pages/DataEvaluation/Home/DataEvaluation.tsx
index d589b37..6e99f4f 100644
--- a/frontend/src/pages/DataEvaluation/Home/DataEvaluation.tsx
+++ b/frontend/src/pages/DataEvaluation/Home/DataEvaluation.tsx
@@ -1,482 +1,267 @@
-import { useState, useEffect } from "react";
-import { Button, Card, Badge, Progress, Table } from "antd";
+import { useState } from "react";
+import {
+  Button,
+  Card,
+  Table,
+  Tag,
+  Space,
+  Typography,
+  Progress,
+  Popconfirm,
+  App,
+} from "antd";
 import {
   PlusOutlined,
-  DeleteOutlined,
-  EyeOutlined,
-  ClockCircleOutlined,
-  DatabaseOutlined,
-  CheckCircleOutlined,
-  CloseCircleOutlined,
-  ExclamationCircleOutlined,
-  ReloadOutlined,
-  UserOutlined,
-  RobotOutlined,
-  EditOutlined,
+  DeleteOutlined
 } from "@ant-design/icons";
 import { SearchControls } from "@/components/SearchControls";
-import { mockTasks } from "@/mock/evaluation";
-import CardView from "@/components/CardView";
 import { useNavigate } from "react-router";
-import type { Dataset } from "@/pages/DataManagement/dataset.model";
-import DevelopmentInProgress from "@/components/DevelopmentInProgress";
+import { deleteEvaluationTaskUsingGet, getPagedEvaluationTaskUsingGet } from "@/pages/DataEvaluation/evaluation.api";
+import CardView from "@/components/CardView";
+import CreateTaskModal from "@/pages/DataEvaluation/Create/CreateTask.tsx";
+import useFetchData from "@/hooks/useFetchData.ts";
+import { EvaluationTask } from "@/pages/DataEvaluation/evaluation.model.ts";
+import { mapEvaluationTask } from "@/pages/DataEvaluation/evaluation.const.tsx";
+
+const { Text, Title } = Typography;
+
+const statusMap = {
+  PENDING: { text: '等待中', color: 'warning'},
+  RUNNING: { text: '运行中', color: 'processing'},
+  COMPLETED: { text: '已完成', color: 'success'},
+  STOPPED: { text: '已停止', color: 'default'},
+  FAILED: { text: '失败', color: 'error'},
+};
 
 export default function DataEvaluationPage() {
+  const { message } = App.useApp();
   const navigate = useNavigate();
-  const [tasks, setTasks] = useState<EvaluationTask[]>(mockTasks);
-  const [datasets, setDatasets] = useState<Dataset[]>([]);
+  const [viewMode, setViewMode] = useState<"card" | "list">("list");
+  const {
+    loading,
+    tableData,
+    pagination,
+    searchParams,
+    setSearchParams,
+    handleFiltersChange,
+    fetchData,
+  } = useFetchData<EvaluationTask>(
+    getPagedEvaluationTaskUsingGet,
+    mapEvaluationTask,
+    30000,
+    true,
+    [],
+    0
+  );
 
-  // 搜索和过滤状态
-  const [searchTerm, setSearchTerm] = useState("");
-  const [selectedFilters, setSelectedFilters] = useState<
-    Record<string, string[]>
-  >({});
-  const [sortBy, setSortBy] = useState("createdAt");
-  const [sortOrder, setSortOrder] = useState<"asc" | "desc">("desc");
-  const [viewMode, setViewMode] = useState<"card" | "table">("card");
+  const [isModalVisible, setIsModalVisible] = useState(false);
 
-  useEffect(() => {
-    if (typeof window !== "undefined") {
-      localStorage.setItem("evaluation-tasks", JSON.stringify(tasks));
+  const handleDeleteTask = async (task: EvaluationTask) => {
+    try {
+      // 调用删除接口
+      await deleteEvaluationTaskUsingGet(task.id);
+      message.success("任务删除成功");
+      // 重新加载数据
+      fetchData().then();
+    } catch (error) {
+      message.error("任务删除失败，请稍后重试");
     }
-  }, [tasks]);
+  };
 
-  // 搜索和过滤配置
   const filterOptions = [
     {
-      key: "evaluationType",
-      label: "评估方式",
+      key: 'status',
+      label: '状态',
+      options: Object.entries(statusMap).map(([value, { text }]) => ({
+        value,
+        label: text,
+      })),
+    },
+    {
+      key: 'taskType',
+      label: '任务类型',
       options: [
-        { label: "模型评估", value: "model" },
-        { label: "人工评估", value: "manual" },
+        { value: 'QA', label: 'QA评估' },
       ],
     },
     {
-      key: "status",
-      label: "状态",
+      key: 'evalMethod',
+      label: '评估方式',
       options: [
-        { label: "待处理", value: "pending" },
-        { label: "运行中", value: "running" },
-        { label: "已完成", value: "completed" },
-        { label: "失败", value: "failed" },
+        { value: 'AUTO', label: '自动评估' },
+        { value: 'MANUAL', label: '人工评估' },
       ],
     },
+  ];
+
+  const columns = [
     {
-      key: "dataset",
-      label: "数据集",
-      options: datasets.map((d) => ({ label: d.name, value: d.id })),
+      title: '任务名称',
+      dataIndex: 'name',
+      key: 'name',
+      render: (name, record) => (
+        <Button
+          type="link"
+          onClick={() => navigate(`/data/evaluation/detail/${record.id}`)}
+        >
+          {name}
+        </Button>
+      ),
+    },
+    {
+      title: '任务类型',
+      dataIndex: 'taskType',
+      key: 'taskType',
+      render: (text: string) => (
+        <Tag color={text === 'QA' ? 'blue' : 'default'}>
+          {text === 'QA' ? 'QA评估' : text}
+        </Tag>
+      ),
+    },
+    {
+      title: '评估方式',
+      dataIndex: 'evalMethod',
+      key: 'evalMethod',
+      render: (text: string) => (
+        <Tag color={text === 'AUTO' ? 'geekblue' : 'orange'}>
+          {text === 'AUTO' ? '自动评估' : '人工评估'}
+        </Tag>
+      ),
+    },
+    {
+      title: '状态',
+      dataIndex: 'status',
+      key: 'status',
+      render: (status: any) => {
+        return (<Tag color={status.color}> {status.label} </Tag>);
+      },
+    },
+    {
+      title: '进度',
+      dataIndex: 'evalProcess',
+      key: 'evalProcess',
+      render: (progress: number, record: EvaluationTask) => (
+        <Progress
+          percent={Math.round(progress * 100)}
+          size="small"
+          status={record.status === 'FAILED' ? 'exception' : 'active'}
+        />
+      ),
+    },
+    {
+      title: "创建时间",
+      dataIndex: "updatedAt",
+      key: "updatedAt",
+      width: 180,
+    },
+    {
+      title: '操作',
+      key: 'action',
+      render: (_: any, task: EvaluationTask) => (
+        <div className="flex items-center gap-2">
+          {operations.map((op) => {
+            if (op.confirm) {
+              <Popconfirm
+                title={op.confirm.title}
+                description={op.confirm.description}
+                onConfirm={() => op.onClick(task)}
+              >
+                <Button type="text" icon={op.icon} />
+              </Popconfirm>;
+            }
+            return (
+              <Button
+                key={op.key}
+                type="text"
+                icon={op.icon}
+                danger={op.danger}
+                onClick={() => op.onClick(task)}
+              />
+            );
+          })}
+        </div>
+      ),
     },
   ];
 
-  const sortOptions = [
-    { label: "创建时间", value: "createdAt" },
-    { label: "任务名称", value: "name" },
-    { label: "完成时间", value: "completedAt" },
-    { label: "评分", value: "score" },
+  const operations = [
+    {
+      key: "delete",
+      label: "删除",
+      danger: true,
+      confirm: {
+        title: "确认删除该任务？",
+        description: "删除后该任务将无法恢复，请谨慎操作。",
+        okText: "删除",
+        cancelText: "取消",
+        okType: "danger",
+      },
+      icon: <DeleteOutlined />,
+      onClick: handleDeleteTask,
+    }
   ];
 
-  // 过滤和排序逻辑
-  const filteredTasks = tasks.filter((task) => {
-    // 搜索过滤
-    if (
-      searchTerm &&
-      !task.name.toLowerCase().includes(searchTerm.toLowerCase()) &&
-      !task.datasetName.toLowerCase().includes(searchTerm.toLowerCase())
-    ) {
-      return false;
-    }
-
-    // 评估方式过滤
-    if (
-      selectedFilters.evaluationType?.length &&
-      !selectedFilters.evaluationType.includes(task.evaluationType)
-    ) {
-      return false;
-    }
-
-    // 状态过滤
-    if (
-      selectedFilters.status?.length &&
-      !selectedFilters.status.includes(task.status)
-    ) {
-      return false;
-    }
-
-    // 数据集过滤
-    if (
-      selectedFilters.dataset?.length &&
-      !selectedFilters.dataset.includes(task.datasetId)
-    ) {
-      return false;
-    }
-
-    return true;
-  });
-
-  // 排序
-  const sortedTasks = [...filteredTasks].sort((a, b) => {
-    let aValue: any = a[sortBy as keyof EvaluationTask];
-    let bValue: any = b[sortBy as keyof EvaluationTask];
-
-    if (sortBy === "score") {
-      aValue = a.score || 0;
-      bValue = b.score || 0;
-    }
-
-    if (typeof aValue === "string") {
-      aValue = aValue.toLowerCase();
-      bValue = bValue.toLowerCase();
-    }
-
-    if (sortOrder === "asc") {
-      return aValue > bValue ? 1 : -1;
-    } else {
-      return aValue < bValue ? 1 : -1;
-    }
-  });
-
-  const getStatusColor = (status: string) => {
-    switch (status) {
-      case "completed":
-        return "green";
-      case "running":
-        return "blue";
-      case "failed":
-        return "red";
-      case "pending":
-        return "gold";
-      default:
-        return "gray";
-    }
-  };
-
-  const getStatusIcon = (status: string) => {
-    switch (status) {
-      case "completed":
-        return <CheckCircleOutlined />;
-      case "running":
-        return <ReloadOutlined spin />;
-      case "failed":
-        return <CloseCircleOutlined />;
-      case "pending":
-        return <ClockCircleOutlined />;
-      default:
-        return <ExclamationCircleOutlined />;
-    }
-  };
-
-  // 开始人工评估
-  const handleStartManualEvaluation = (task: EvaluationTask) => {
-    navigate(`/data/evaluation/manual-evaluate/${task.id}`);
-  };
-
-  // 查看评估报告
-  const handleViewReport = (task: EvaluationTask) => {
-    navigate(`/data/evaluation/task-report/${task.id}`);
-  };
-
-  // 删除任务
-  const handleDeleteTask = (taskId: string) => {
-    setTasks(tasks.filter((task) => task.id !== taskId));
-  };
-
-  return <DevelopmentInProgress showTime="2025.11.30" />;
-  // 主列表界面
   return (
-    <div>
-      {/* 页面头部 */}
-      <div className="flex items-center justify-between">
-        <h1 className="text-xl font-bold">数据评估</h1>
+    <div className="p-4">
+      <div className="flex items-center justify-between mb-4">
+        <Title level={4} style={{ margin: 0 }}>数据评估</Title>
         <Button
           type="primary"
           icon={<PlusOutlined />}
-          onClick={() => navigate("/data/evaluation/create-task")}
+          onClick={() => setIsModalVisible(true)}
         >
           创建评估任务
         </Button>
       </div>
-
-      {/* 搜索和过滤控件 */}
-      <SearchControls
-        searchTerm={searchTerm}
-        onSearchChange={setSearchTerm}
-        searchPlaceholder="搜索任务名称或数据集..."
-        filters={filterOptions}
-        selectedFilters={selectedFilters}
-        onFiltersChange={setSelectedFilters}
-        viewMode={viewMode}
-        onViewModeChange={setViewMode}
-      />
-
-      {/* 任务列表 */}
-      {viewMode === "card" ? (
-        <CardView
-          data={sortedTasks.map((task) => ({
-            id: task.id,
-            name: task.name,
-            type: task.evaluationType,
-            icon:
-              task.evaluationType === "model" ? (
-                <RobotOutlined style={{ fontSize: 24, color: "#722ed1" }} />
-              ) : (
-                <UserOutlined style={{ fontSize: 24, color: "#52c41a" }} />
-              ),
-            iconColor: "",
-            status: {
-              label:
-                task.status === "completed"
-                  ? "已完成"
-                  : task.status === "running"
-                  ? "运行中"
-                  : task.status === "failed"
-                  ? "失败"
-                  : "待处理",
-              icon: getStatusIcon(task.status),
-              color: getStatusColor(task.status),
-            },
-            description: task.description,
-            tags: [task.datasetName],
-            statistics: [
-              {
-                label: "进度",
-                value: task.progress !== undefined ? `${task.progress}%` : "-",
-              },
-              { label: "评分", value: task.score ? `${task.score}分` : "-" },
-            ],
-            lastModified: task.createdAt,
-          }))}
-          operations={[
-            {
-              key: "view",
-              label: "查看报告",
-              icon: <EyeOutlined />,
-              onClick: (item) => {
-                const task = tasks.find((t) => t.id === item.id);
-                if (task) handleViewReport(task);
-              },
-            },
-            {
-              key: "evaluate",
-              label: "开始评估",
-              icon: <EditOutlined />,
-              onClick: (item) => {
-                const task = tasks.find((t) => t.id === item.id);
-                if (task) handleStartManualEvaluation(task);
-              },
-            },
-            {
-              key: "delete",
-              label: "删除",
-              icon: <DeleteOutlined />,
-              onClick: (item) => handleDeleteTask(item.id as string),
-            },
-          ]}
-          onView={(item) => {
-            const task = tasks.find((t) => t.id === item.id);
-            if (task) handleViewReport(task);
-          }}
+      <>
+        {/* 搜索、筛选和视图控制 */}
+        <SearchControls
+          searchTerm={searchParams.keyword}
+          onSearchChange={(keyword) =>
+            setSearchParams({ ...searchParams, keyword })
+          }
+          searchPlaceholder="搜索任务名称..."
+          filters={filterOptions}
+          onFiltersChange={handleFiltersChange}
+          onClearFilters={() =>
+            setSearchParams({ ...searchParams, filter: {} })
+          }
+          viewMode={viewMode}
+          onViewModeChange={setViewMode}
+          showViewToggle
+          onReload={fetchData}
         />
-      ) : (
-        <Card>
-          <Table
-            rowKey="id"
-            dataSource={sortedTasks}
-            pagination={false}
-            scroll={{ x: "max-content" }}
-            columns={[
-              {
-                title: "任务名称",
-                dataIndex: "name",
-                key: "name",
-                render: (text, record) => (
-                  <div>
-                    <div style={{ fontWeight: 500 }}>{text}</div>
-                    <div style={{ fontSize: 13, color: "#888" }}>
-                      {record.description}
-                    </div>
-                  </div>
-                ),
-              },
-              {
-                title: "数据集",
-                dataIndex: "datasetName",
-                key: "datasetName",
-                render: (text) => (
-                  <div
-                    style={{ display: "flex", alignItems: "center", gap: 4 }}
-                  >
-                    <DatabaseOutlined />
-                    <span style={{ fontSize: 13 }}>{text}</span>
-                  </div>
-                ),
-              },
-              {
-                title: "评估方式",
-                dataIndex: "evaluationType",
-                key: "evaluationType",
-                render: (type) => (
-                  <div
-                    style={{ display: "flex", alignItems: "center", gap: 4 }}
-                  >
-                    {type === "model" ? (
-                      <RobotOutlined style={{ color: "#722ed1" }} />
-                    ) : (
-                      <UserOutlined style={{ color: "#52c41a" }} />
-                    )}
-                    <span style={{ fontSize: 13 }}>
-                      {type === "model" ? "模型评估" : "人工评估"}
-                    </span>
-                  </div>
-                ),
-              },
-              {
-                title: "状态",
-                dataIndex: "status",
-                key: "status",
-                render: (status) => (
-                  <Badge
-                    color={getStatusColor(status)}
-                    style={{ background: "none", padding: 0 }}
-                    count={
-                      <span
-                        style={{
-                          display: "flex",
-                          alignItems: "center",
-                          gap: 4,
-                        }}
-                      >
-                        {getStatusIcon(status)}
-                        <span>
-                          {status === "completed" && "已完成"}
-                          {status === "running" && "运行中"}
-                          {status === "failed" && "失败"}
-                          {status === "pending" && "待处理"}
-                        </span>
-                      </span>
-                    }
-                    showZero={false}
-                  />
-                ),
-              },
-              {
-                title: "进度",
-                dataIndex: "progress",
-                key: "progress",
-                render: (progress) =>
-                  progress !== undefined ? (
-                    <div style={{ width: 100 }}>
-                      <Progress
-                        percent={progress}
-                        size="small"
-                        showInfo={false}
-                      />
-                      <div
-                        style={{
-                          fontSize: 12,
-                          color: "#888",
-                          textAlign: "right",
-                        }}
-                      >
-                        {progress}%
-                      </div>
-                    </div>
-                  ) : (
-                    <span style={{ fontSize: 13, color: "#bbb" }}>-</span>
-                  ),
-              },
-              {
-                title: "评分",
-                dataIndex: "score",
-                key: "score",
-                render: (score) =>
-                  score ? (
-                    <span style={{ fontWeight: 500, color: "#389e0d" }}>
-                      {score}分
-                    </span>
-                  ) : (
-                    <span style={{ fontSize: 13, color: "#bbb" }}>-</span>
-                  ),
-              },
-              {
-                title: "创建时间",
-                dataIndex: "createdAt",
-                key: "createdAt",
-                render: (text) => <span style={{ fontSize: 13 }}>{text}</span>,
-              },
-              {
-                title: "操作",
-                key: "action",
-                render: (_, task) => (
-                  <div style={{ display: "flex", gap: 8 }}>
-                    {task.status === "completed" && (
-                      <Button
-                        type="default"
-                        size="small"
-                        icon={<EyeOutlined />}
-                        onClick={() => handleViewReport(task)}
-                      >
-                        报告
-                      </Button>
-                    )}
-                    {task.evaluationType === "manual" &&
-                      task.status === "pending" && (
-                        <Button
-                          type="default"
-                          size="small"
-                          icon={<EditOutlined />}
-                          onClick={() => handleStartManualEvaluation(task)}
-                        >
-                          评估
-                        </Button>
-                      )}
-                    <Button
-                      type="text"
-                      size="small"
-                      danger
-                      icon={<DeleteOutlined />}
-                      onClick={() => handleDeleteTask(task.id)}
-                    />
-                  </div>
-                ),
-              },
-            ]}
-            locale={{
-              emptyText: (
-                <div
-                  style={{ textAlign: "center", padding: 48, color: "#bbb" }}
-                >
-                  <DatabaseOutlined style={{ fontSize: 48, marginBottom: 8 }} />
-                  <div style={{ marginTop: 8 }}>暂无评估任务</div>
-                  <div style={{ fontSize: 13, color: "#ccc" }}>
-                    点击"创建评估任务"开始评估数据集质量
-                  </div>
-                </div>
-              ),
+        {/* 任务列表 */}
+        {viewMode === "list" ? (
+          <Card>
+            <Table
+              columns={columns}
+              dataSource={tableData}
+              pagination={pagination}
+              rowKey="id"
+              scroll={{ x: "max-content", y: "calc(100vh - 30rem)" }}
+            />
+          </Card>
+        ) : (
+          <CardView
+            loading={loading}
+            data={tableData}
+            operations={operations}
+            pagination={pagination}
+            onView={(task) => {
+              navigate(`/data/evaluation/detail/${task.id}`);
             }}
           />
-        </Card>
-      )}
-
-      {sortedTasks.length === 0 && (
-        <div style={{ textAlign: "center", padding: "48px 0" }}>
-          <DatabaseOutlined
-            style={{ fontSize: 64, color: "#bbb", marginBottom: 16 }}
-          />
-          <div style={{ fontSize: 18, fontWeight: 500, marginBottom: 8 }}>
-            暂无评估任务
-          </div>
-          <div style={{ color: "#888", marginBottom: 24 }}>
-            创建您的第一个数据评估任务
-          </div>
-          <Button
-            type="primary"
-            icon={<PlusOutlined />}
-            onClick={() => navigate("/data/evaluation/create-task")}
-          >
-            创建评估任务
-          </Button>
-        </div>
-      )}
+        )}
+      </>
+      <CreateTaskModal
+        visible={isModalVisible}
+        onCancel={() => setIsModalVisible(false)}
+        onSuccess={() => {
+          setIsModalVisible(false);
+          fetchData();
+        }}
+      />
     </div>
   );
 }
diff --git a/frontend/src/pages/DataEvaluation/data-evaluation.d.ts b/frontend/src/pages/DataEvaluation/data-evaluation.d.ts
deleted file mode 100644
index b67878f..0000000
--- a/frontend/src/pages/DataEvaluation/data-evaluation.d.ts
+++ /dev/null
@@ -1,73 +0,0 @@
-interface EvaluationDimension {
-    id: string
-    name: string
-    description: string
-    category: "quality" | "accuracy" | "completeness" | "consistency" | "bias" | "custom"
-    isCustom?: boolean
-    isEnabled?: boolean
-}
-
-interface EvaluationTask {
-    id: string
-    name: string
-    datasetId: string
-    datasetName: string
-    evaluationType: "model" | "manual"
-    status: "running" | "completed" | "failed" | "pending"
-    score?: number
-    progress?: number
-    createdAt: string
-    completedAt?: string
-    description: string
-    dimensions: string[]
-    customDimensions: EvaluationDimension[]
-    sliceConfig?: {
-        threshold: number
-        sampleCount: number
-        method: string
-    }
-    modelConfig?: {
-        url: string
-        apiKey: string
-        prompt: string
-        temperature: number
-        maxTokens: number
-    }
-    metrics: {
-        accuracy: number
-        completeness: number
-        consistency: number
-        relevance: number
-    }
-    issues: {
-        type: string
-        count: number
-        severity: "high" | "medium" | "low"
-    }[]
-}
-
-interface EvaluationSlice {
-    id: string
-    content: string
-    sourceFile: string
-    sliceIndex: number
-    sliceType: string
-    metadata: {
-        startPosition?: number
-        endPosition?: number
-        pageNumber?: number
-        section?: string
-        processingMethod: string
-    }
-    scores?: { [dimensionId: string]: number }
-    comment?: string
-}
-
-interface QAPair {
-    id: string
-    question: string
-    answer: string
-    sliceId: string
-    score: number
-    feedback?: string
-}
\ No newline at end of file
diff --git a/frontend/src/pages/DataEvaluation/data-evaluation.api.ts b/frontend/src/pages/DataEvaluation/evaluation.api.ts
similarity index 89%
rename from frontend/src/pages/DataEvaluation/data-evaluation.api.ts
rename to frontend/src/pages/DataEvaluation/evaluation.api.ts
index 3a44921..5794f94 100644
--- a/frontend/src/pages/DataEvaluation/data-evaluation.api.ts
+++ b/frontend/src/pages/DataEvaluation/evaluation.api.ts
@@ -1,5 +1,46 @@
 import { get, post, put, del, download } from "@/utils/request";
 
+export function createEvaluationTaskUsingPost(data: any) {
+  return post("/api/evaluation/tasks", data);
+}
+
+export function getPagedEvaluationTaskUsingGet(params?: any) {
+  return get("/api/evaluation/tasks", params);
+}
+
+export function deleteEvaluationTaskUsingGet(id: string) {
+  const url = `/api/evaluation/tasks?ids=${id}`;
+  return del(url);
+}
+
+export function queryPromptTemplatesUsingGet() {
+  return get("/api/evaluation/prompt-templates");
+}
+
+export function getEvaluationTaskByIdUsingGet(taskId: string | number) {
+  return get(`/api/evaluation/tasks/${taskId}`);
+}
+
+export function queryEvaluationFilesUsingGet(params: {
+  taskId: string;
+  page?: number;
+  size?: number;
+}) {
+  const { taskId, ...rest } = params;
+  return get(`/api/evaluation/tasks/${taskId}/files`, rest);
+}
+
+export function queryEvaluationItemsUsingGet(params: {
+  taskId: string;
+  page?: number;
+  size?: number;
+  status?: string;
+  file_id?: string;
+}) {
+  const { taskId, ...rest } = params;
+  return get(`/api/evaluation/tasks/${taskId}/items`, rest);
+}
+
 // 数据质量评估相关接口
 export function evaluateDataQualityUsingPost(data: any) {
   return post("/api/v1/evaluation/quality", data);
@@ -113,14 +154,6 @@ export function queryEvaluationTasksUsingGet(params?: any) {
   return get("/api/v1/evaluation/tasks", params);
 }
 
-export function createEvaluationTaskUsingPost(data: any) {
-  return post("/api/v1/evaluation/tasks", data);
-}
-
-export function getEvaluationTaskByIdUsingGet(taskId: string | number) {
-  return get(`/api/v1/evaluation/tasks/${taskId}`);
-}
-
 export function cancelEvaluationTaskUsingPost(taskId: string | number) {
   return post(`/api/v1/evaluation/tasks/${taskId}/cancel`);
 }
@@ -240,4 +273,4 @@ export function enableEvaluationScheduleUsingPost(scheduleId: string | number) {
 
 export function disableEvaluationScheduleUsingPost(scheduleId: string | number) {
   return post(`/api/v1/evaluation/schedules/${scheduleId}/disable`);
-}
\ No newline at end of file
+}
diff --git a/frontend/src/pages/DataEvaluation/evaluation.const.tsx b/frontend/src/pages/DataEvaluation/evaluation.const.tsx
new file mode 100644
index 0000000..369ed6f
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/evaluation.const.tsx
@@ -0,0 +1,67 @@
+import { formatDate } from "@/utils/unit";
+import { BarChart3 } from "lucide-react";
+import { EvaluationStatus, EvaluationTask } from "@/pages/DataEvaluation/evaluation.model.ts";
+
+export const evalTaskStatusMap: Record<
+  string,
+  {
+    value: EvaluationStatus;
+    label: string;
+    color: string;
+  }
+> = {
+  [EvaluationStatus.PENDING]: {
+    value: EvaluationStatus.PENDING,
+    label: "等待中",
+    color: "gray",
+  },
+  [EvaluationStatus.RUNNING]: {
+    value: EvaluationStatus.RUNNING,
+    label: "运行中",
+    color: "blue",
+  },
+  [EvaluationStatus.COMPLETED]: {
+    value: EvaluationStatus.COMPLETED,
+    label: "已完成",
+    color: "green",
+  },
+  [EvaluationStatus.FAILED]: {
+    value: EvaluationStatus.FAILED,
+    label: "失败",
+    color: "red",
+  },
+  [EvaluationStatus.PAUSED]: {
+    value: EvaluationStatus.PAUSED,
+    label: "已暂停",
+    color: "orange",
+  },
+};
+
+export function mapEvaluationTask(task: Partial<EvaluationTask>): EvaluationTask {
+  return {
+    ...task,
+    status: evalTaskStatusMap[task.status || EvaluationStatus.PENDING],
+    createdAt: formatDate(task.createdAt),
+    updatedAt: formatDate(task.updatedAt),
+    description: task.description,
+    icon: <BarChart3 />,
+    iconColor: task.ratio_method === "DATASET" ? "bg-blue-100" : "bg-green-100",
+    statistics: [
+      {
+        label: "任务类型",
+        icon: <BarChart3 className="w-4 h-4 text-gray-500" />,
+        value: (task.taskType ?? 0).toLocaleString(),
+      },
+      {
+        label: "评估方式",
+        icon: <BarChart3 className="w-4 h-4 text-gray-500" />,
+        value: (task.evalMethod ?? 0).toLocaleString(),
+      },
+      {
+        label: "数据源",
+        icon: <BarChart3 className="w-4 h-4 text-gray-500" />,
+        value: (task.sourceName ?? 0).toLocaleString(),
+      },
+    ],
+  };
+}
diff --git a/frontend/src/pages/DataEvaluation/evaluation.model.ts b/frontend/src/pages/DataEvaluation/evaluation.model.ts
new file mode 100644
index 0000000..d867ffc
--- /dev/null
+++ b/frontend/src/pages/DataEvaluation/evaluation.model.ts
@@ -0,0 +1,57 @@
+export enum EvaluationStatus {
+  PENDING = "PENDING",
+  RUNNING = "RUNNING",
+  COMPLETED = "COMPLETED",
+  FAILED = "FAILED",
+  PAUSED = "PAUSED",
+}
+
+export interface EvaluationTask {
+  id: string;
+  name: string;
+  description?: string;
+  taskType: string;
+  sourceType: string;
+  sourceId: string;
+  sourceName: string;
+  status: 'PENDING' | 'RUNNING' | 'COMPLETED' | 'STOPPED' | 'FAILED';
+  evalProcess: number;
+  evalMethod: 'AUTO' | 'MANUAL';
+  createdAt: string;
+  updatedAt: string;
+}
+
+interface EvaluationDimension {
+    id: string
+    name: string
+    description: string
+    category: "quality" | "accuracy" | "completeness" | "consistency" | "bias" | "custom"
+    isCustom?: boolean
+    isEnabled?: boolean
+}
+
+interface EvaluationSlice {
+    id: string
+    content: string
+    sourceFile: string
+    sliceIndex: number
+    sliceType: string
+    metadata: {
+        startPosition?: number
+        endPosition?: number
+        pageNumber?: number
+        section?: string
+        processingMethod: string
+    }
+    scores?: { [dimensionId: string]: number }
+    comment?: string
+}
+
+interface QAPair {
+    id: string
+    question: string
+    answer: string
+    sliceId: string
+    score: number
+    feedback?: string
+}
diff --git a/frontend/src/pages/DataManagement/dataset.model.ts b/frontend/src/pages/DataManagement/dataset.model.ts
index a685c14..688de58 100644
--- a/frontend/src/pages/DataManagement/dataset.model.ts
+++ b/frontend/src/pages/DataManagement/dataset.model.ts
@@ -55,7 +55,7 @@ export interface Dataset {
   updatedAt: string;
   tags: string[];
   targetLocation?: string;
-  distribution?: Record<string, number>;
+  distribution?: Record<string, Record<string, number>>;
 }
 
 export interface TagItem {
diff --git a/frontend/src/pages/RatioTask/Create/CreateRatioTask.tsx b/frontend/src/pages/RatioTask/Create/CreateRatioTask.tsx
index de1a64b..5598f01 100644
--- a/frontend/src/pages/RatioTask/Create/CreateRatioTask.tsx
+++ b/frontend/src/pages/RatioTask/Create/CreateRatioTask.tsx
@@ -1,4 +1,4 @@
-import { useMemo, useState } from "react";
+import { useState } from "react";
 import { Button, Form, message } from "antd";
 import { ArrowLeft, ChevronRight } from "lucide-react";
 import { createRatioTaskUsingPost } from "@/pages/RatioTask/ratio.api.ts";
diff --git a/frontend/src/pages/RatioTask/Create/components/RatioConfig.tsx b/frontend/src/pages/RatioTask/Create/components/RatioConfig.tsx
index 7fefcbd..fcb58d2 100644
--- a/frontend/src/pages/RatioTask/Create/components/RatioConfig.tsx
+++ b/frontend/src/pages/RatioTask/Create/components/RatioConfig.tsx
@@ -2,14 +2,12 @@ import React, { useMemo, useState, useEffect, FC } from "react";
 import {
   Badge,
   Card,
-  Progress,
   Button,
   Select,
   Table,
   InputNumber,
-  Space,
 } from "antd";
-import { BarChart3, Filter } from "lucide-react";
+import { BarChart3 } from "lucide-react";
 import type { Dataset } from "@/pages/DataManagement/dataset.model.ts";
 
 const TIME_RANGE_OPTIONS = [
@@ -20,6 +18,11 @@ const TIME_RANGE_OPTIONS = [
   { label: '最近30天', value: 30 },
 ];
 
+interface LabelFilter {
+  label: string;
+  value: string;
+}
+
 interface RatioConfigItem {
   id: string;
   name: string;
@@ -27,7 +30,7 @@ interface RatioConfigItem {
   quantity: number;
   percentage: number;
   source: string; // dataset id
-  labelFilter?: string;
+  labelFilter?: LabelFilter;
   dateRange?: number;
 }
 
@@ -36,7 +39,8 @@ interface RatioConfigProps {
   selectedDatasets: string[];
   datasets: Dataset[];
   totalTargetCount: number;
-  distributions: Record<string, Record<string, number>>;
+  // distributions now: { datasetId: { labelName: { labelValue: count } } }
+  distributions: Record<string, Record<string, Record<string, number>>>;
   onChange?: (configs: RatioConfigItem[]) => void;
 }
 
@@ -63,6 +67,10 @@ const RatioConfig: FC<RatioConfigProps> = ({
     return Object.keys(dist);
   };
 
+  const getLabelValues = (datasetId: string, label: string): string[] => {
+    return Object.keys(distributions[String(datasetId)]?.[label] || {});
+  };
+
   const addConfig = (datasetId: string) => {
     const dataset = datasets.find((d) => String(d.id) === datasetId);
     const newConfig: RatioConfigItem = {
@@ -208,46 +216,85 @@ const RatioConfig: FC<RatioConfigProps> = ({
               );
 
               const labels = getDatasetLabels(datasetId);
-              const usedLabels = datasetConfigs
-                .map((c) => c.labelFilter)
-                .filter(Boolean) as string[];
+
+              // helper: used values per label for this dataset (exclude a given row when needed)
+              const getUsedValuesForLabel = (label: string, excludeId?: string) => {
+                return new Set(
+                  datasetConfigs
+                    .filter((c) => c.id !== excludeId && c.labelFilter?.label === label)
+                    .map((c) => c.labelFilter?.value)
+                    .filter(Boolean) as string[]
+                );
+              };
 
               const columns = [
                 {
-                  title: "配比项",
-                  dataIndex: "id",
-                  key: "id",
-                  render: (_: any, record: RatioConfigItem) => (
-                    <Space>
-                      <Filter size={14} className="text-gray-400" />
-                      <span className="text-sm">{record.name}</span>
-                    </Space>
-                  ),
-                },
-                {
-                  title: "标签筛选",
+                  title: "标签",
                   dataIndex: "labelFilter",
                   key: "labelFilter",
                   render: (_: any, record: RatioConfigItem) => {
                     const availableLabels = labels
-                      .map((l) => ({ label: l, value: l }))
-                      .filter(
-                        (opt) =>
-                          opt.value === record.labelFilter ||
-                          !usedLabels.includes(opt.value)
-                      );
+                      .map((l) => ({
+                        label: l,
+                        value: l,
+                        disabled: getLabelValues(datasetId, l).every((v) => getUsedValuesForLabel(l, record.id).has(v)),
+                      }))
                     return (
                       <Select
                         style={{ width: "160px" }}
                         placeholder="选择标签"
-                        value={record.labelFilter}
+                        value={record.labelFilter?.label}
                         options={availableLabels}
                         allowClear
-                        onChange={(value) =>
+                        onChange={(value) => {
+                          if (!value) {
+                            updateConfig(record.id, { labelFilter: undefined });
+                          } else {
+                            // reset value when label changes
+                            updateConfig(record.id, {
+                              labelFilter: { label: value, value: "" },
+                            });
+                          }
+                        }}
+                      />
+                    );
+                  },
+                },
+                {
+                  title: "标签值",
+                  dataIndex: "labelValue",
+                  key: "labelValue",
+                  render: (_: any, record: RatioConfigItem) => {
+                    const selectedLabel = record.labelFilter?.label;
+                    const options = selectedLabel
+                      ? getLabelValues(datasetId, selectedLabel).map((v) => ({
+                        label: v,
+                        value: v,
+                        disabled: datasetConfigs.some(
+                          (c) =>
+                            c.id !== record.id &&
+                            c.labelFilter?.label === selectedLabel &&
+                            c.labelFilter?.value === v
+                        ),
+                      }))
+                      : [];
+                    return (
+                      <Select
+                        style={{ width: "180px" }}
+                        placeholder="选择标签值"
+                        value={record.labelFilter?.value || undefined}
+                        options={options}
+                        allowClear
+                        disabled={!selectedLabel}
+                        onChange={(value) => {
+                          if (!selectedLabel) return;
                           updateConfig(record.id, {
-                            labelFilter: value || undefined,
-                          })
-                        }
+                            labelFilter: {
+                              label: selectedLabel,
+                              value: value || "",
+                            },
+                          });
+                        }}
                       />
                     );
                   },
@@ -286,23 +333,6 @@ const RatioConfig: FC<RatioConfigProps> = ({
                     />
                   ),
                 },
-                {
-                  title: "占比",
-                  dataIndex: "percentage",
-                  key: "percentage",
-                  render: (_: any, record: RatioConfigItem) => (
-                    <div style={{ minWidth: 140 }}>
-                      <div className="text-xs mb-1">
-                        {record.percentage ?? 0}%
-                      </div>
-                      <Progress
-                        percent={record.percentage ?? 0}
-                        size="small"
-                        showInfo={false}
-                      />
-                    </div>
-                  ),
-                },
                 {
                   title: "操作",
                   dataIndex: "actions",
diff --git a/frontend/src/pages/RatioTask/Create/components/SelectDataset.tsx b/frontend/src/pages/RatioTask/Create/components/SelectDataset.tsx
index ef04988..377171c 100644
--- a/frontend/src/pages/RatioTask/Create/components/SelectDataset.tsx
+++ b/frontend/src/pages/RatioTask/Create/components/SelectDataset.tsx
@@ -1,3 +1,4 @@
+// typescript
 import React, { useEffect, useState } from "react";
 import { Badge, Button, Card, Checkbox, Input, Pagination } from "antd";
 import { Search as SearchIcon } from "lucide-react";
@@ -5,32 +6,48 @@ import type { Dataset } from "@/pages/DataManagement/dataset.model.ts";
 import {
   queryDatasetsUsingGet,
   queryDatasetByIdUsingGet,
-  queryDatasetStatisticsByIdUsingGet,
 } from "@/pages/DataManagement/dataset.api.ts";
 
 interface SelectDatasetProps {
   selectedDatasets: string[];
   onSelectedDatasetsChange: (next: string[]) => void;
+  // distributions now: { datasetId: { labelName: { labelValue: count } } }
   onDistributionsChange?: (
-    next: Record<string, Record<string, number>>
+    next: Record<string, Record<string, Record<string, number>>>
   ) => void;
   onDatasetsChange?: (list: Dataset[]) => void;
 }
 
 const SelectDataset: React.FC<SelectDatasetProps> = ({
-  selectedDatasets,
-  onSelectedDatasetsChange,
-  onDistributionsChange,
-  onDatasetsChange,
-}) => {
+                                                       selectedDatasets,
+                                                       onSelectedDatasetsChange,
+                                                       onDistributionsChange,
+                                                       onDatasetsChange,
+                                                     }) => {
   const [datasets, setDatasets] = useState<Dataset[]>([]);
   const [loading, setLoading] = useState(false);
   const [searchQuery, setSearchQuery] = useState("");
   const [pagination, setPagination] = useState({ page: 1, size: 10, total: 0 });
   const [distributions, setDistributions] = useState<
-    Record<string, Record<string, number>>
+    Record<string, Record<string, Record<string, number>>>
   >({});
 
+  // Helper: flatten nested distribution for preview and filter logic
+  const flattenDistribution = (
+    dist?: Record<string, Record<string, number>>
+  ): Array<{ label: string; value: string; count: number }> => {
+    if (!dist) return [];
+    const items: Array<{ label: string; value: string; count: number }> = [];
+    Object.entries(dist).forEach(([label, values]) => {
+      if (values && typeof values === "object") {
+        Object.entries(values).forEach(([val, cnt]) => {
+          items.push({ label, value: val, count: cnt });
+        });
+      }
+    });
+    return items;
+  };
+
   // Fetch dataset list
   useEffect(() => {
     const fetchDatasets = async () => {
@@ -52,10 +69,10 @@ const SelectDataset: React.FC<SelectDatasetProps> = ({
         setLoading(false);
       }
     };
-    fetchDatasets();
+    fetchDatasets().then(() => {});
   }, [pagination.page, pagination.size, searchQuery]);
 
-  // Fetch label distributions when in label mode
+  // Fetch label distributions when datasets change
   useEffect(() => {
     const fetchDistributions = async () => {
       if (!datasets?.length) return;
@@ -64,74 +81,25 @@ const SelectDataset: React.FC<SelectDatasetProps> = ({
         .filter((id) => !distributions[id]);
       if (!idsToFetch.length) return;
       try {
-        const results = await Promise.all(
-          idsToFetch.map(async (id) => {
-            try {
-              const statRes = await queryDatasetStatisticsByIdUsingGet(id);
-              return { id, stats: statRes?.data };
-            } catch {
-              return { id, stats: null };
-            }
-          })
-        );
-
-        const next: Record<string, Record<string, number>> = {
-          ...distributions,
-        };
-        for (const { id, stats } of results) {
-          let dist: Record<string, number> | undefined = undefined;
-          if (stats) {
-            const candidates: any[] = [
-              (stats as any).labelDistribution,
-              (stats as any).tagDistribution,
-              (stats as any).label_stats,
-              (stats as any).labels,
-              (stats as any).distribution,
-            ];
-            let picked = candidates.find(
-              (c) => c && (typeof c === "object" || Array.isArray(c))
-            );
-            if (Array.isArray(picked)) {
-              const obj: Record<string, number> = {};
-              picked.forEach((it: any) => {
-                const key = it?.label ?? it?.name ?? it?.tag ?? it?.key;
-                const val = it?.count ?? it?.value ?? it?.num ?? it?.total;
-                if (key != null && typeof val === "number")
-                  obj[String(key)] = val;
-              });
-              dist = obj;
-            } else if (picked && typeof picked === "object") {
-              dist = picked as Record<string, number>;
-            }
-          }
-          if (!dist) {
-            try {
-              const detRes = await queryDatasetByIdUsingGet(id);
-              const det = detRes?.data;
-              if (det) {
-                let picked =
-                  (det as any).distribution ||
-                  (det as any).labelDistribution ||
-                  (det as any).tagDistribution ||
-                  (det as any).label_stats ||
-                  (det as any).labels ||
-                  undefined;
-                if (Array.isArray(picked)) {
-                  const obj: Record<string, number> = {};
-                  picked.forEach((it: any) => {
-                    const key = it?.label ?? it?.name ?? it?.tag ?? it?.key;
-                    const val = it?.count ?? it?.value ?? it?.num ?? it?.total;
-                    if (key != null && typeof val === "number")
-                      obj[String(key)] = val;
-                  });
-                  dist = obj;
-                } else if (picked && typeof picked === "object") {
-                  dist = picked as Record<string, number>;
-                }
+        const next: Record<
+          string,
+          Record<string, Record<string, number>>
+        > = { ...distributions };
+        for (const id of idsToFetch) {
+          let dist: Record<string, Record<string, number>> | undefined =
+            undefined;
+          try {
+            const detRes = await queryDatasetByIdUsingGet(id);
+            const det = detRes?.data;
+            if (det) {
+              const picked = det?.distribution;
+              if (picked && typeof picked === "object") {
+                // Assume picked is now { labelName: { labelValue: count } }
+                dist = picked as Record<string, Record<string, number>>;
               }
-            } catch {
-              dist = undefined;
             }
+          } catch {
+            dist = undefined;
           }
           next[String(id)] = dist || {};
         }
@@ -141,7 +109,7 @@ const SelectDataset: React.FC<SelectDatasetProps> = ({
         // ignore
       }
     };
-    fetchDistributions();
+    fetchDistributions().then(() => {});
     // eslint-disable-next-line react-hooks/exhaustive-deps
   }, [datasets]);
 
@@ -195,6 +163,8 @@ const SelectDataset: React.FC<SelectDatasetProps> = ({
             datasets.map((dataset) => {
               const idStr = String(dataset.id);
               const checked = selectedDatasets.includes(idStr);
+              const distFor = distributions[idStr];
+              const flat = flattenDistribution(distFor);
               return (
                 <Card
                   key={dataset.id}
@@ -224,17 +194,15 @@ const SelectDataset: React.FC<SelectDatasetProps> = ({
                         <span>{dataset.size}</span>
                       </div>
                       <div className="mt-2">
-                        {distributions[idStr] ? (
-                          Object.entries(distributions[idStr]).length > 0 ? (
+                        {distFor ? (
+                          flat.length > 0 ? (
                             <div className="flex flex-wrap gap-2 text-xs">
-                              {Object.entries(distributions[idStr])
-                                .slice(0, 8)
-                                .map(([tag, count]) => (
-                                  <Badge
-                                    key={tag}
-                                    color="gray"
-                                  >{`${tag}: ${count}`}</Badge>
-                                ))}
+                              {flat.slice(0, 8).map((it) => (
+                                <Badge
+                                  key={`${it.label}_${it.value}`}
+                                  color="gray"
+                                >{`${it.label}/${it.value}: ${it.count}`}</Badge>
+                              ))}
                             </div>
                           ) : (
                             <div className="text-xs text-gray-400">
diff --git a/frontend/src/pages/RatioTask/ratio.const.tsx b/frontend/src/pages/RatioTask/ratio.const.tsx
index 55d8fb4..d18f244 100644
--- a/frontend/src/pages/RatioTask/ratio.const.tsx
+++ b/frontend/src/pages/RatioTask/ratio.const.tsx
@@ -45,9 +45,7 @@ export function mapRatioTask(task: Partial<RatioTaskItem>): RatioTaskItem {
     status: ratioTaskStatusMap[task.status || RatioStatus.PENDING],
     createdAt: formatDate(task.created_at),
     updatedAt: formatDate(task.updated_at),
-    description:
-      task.description ||
-      (task.ratio_method === "DATASET" ? "按数据集配比" : "按标签配比"),
+    description: task.description,
     icon: <BarChart3 />,
     iconColor: task.ratio_method === "DATASET" ? "bg-blue-100" : "bg-green-100",
     statistics: [
@@ -73,16 +71,5 @@ export function mapRatioTask(task: Partial<RatioTaskItem>): RatioTaskItem {
         value: task.created_at || "-",
       },
     ],
-    type: task.ratio_method === "DATASET" ? "数据集配比" : "标签配比",
-    // progress: 100,
-    // sourceDatasets: ["sentiment_dataset", "news_classification"],
-    // targetRatio: { 正面: 33, 负面: 33, 中性: 34 },
-    // currentRatio: { 正面: 33, 负面: 33, 中性: 34 },
-    // totalRecords: 15000,
-    // processedRecords: 15000,
-    // estimatedTime: "已完成",
-    // quality: 95,
-    // strategy: "随机下采样",
-    // outputPath: "/data/balanced/sentiment_balanced_20250120",
   };
 }
diff --git a/frontend/src/pages/SettingsPage/ModelAccess.tsx b/frontend/src/pages/SettingsPage/ModelAccess.tsx
index da60908..055d652 100644
--- a/frontend/src/pages/SettingsPage/ModelAccess.tsx
+++ b/frontend/src/pages/SettingsPage/ModelAccess.tsx
@@ -32,6 +32,7 @@ export interface ModelI {
   modelName: string;
   provider: string;
   model: string;
+  baseUrl: string;
   apiKey: string;
   type: string;
   isEnabled: boolean;
diff --git a/frontend/src/routes/routes.ts b/frontend/src/routes/routes.ts
index 9a7d02d..c6f46d5 100644
--- a/frontend/src/routes/routes.ts
+++ b/frontend/src/routes/routes.ts
@@ -22,7 +22,6 @@ import InstructionTemplateCreate from "@/pages/SynthesisTask/CreateTemplate";
 import SynthesisTaskCreate from "@/pages/SynthesisTask/CreateTask";
 
 import DataEvaluationPage from "@/pages/DataEvaluation/Home/DataEvaluation";
-import EvaluationTaskCreate from "@/pages/DataEvaluation/Create/CreateTask";
 import EvaluationTaskReport from "@/pages/DataEvaluation/Report/EvaluationReport";
 import ManualEvaluatePage from "@/pages/DataEvaluation/Evaluate/ManualEvaluate";
 
@@ -41,6 +40,7 @@ import { withErrorBoundary } from "@/components/ErrorBoundary";
 import AgentPage from "@/pages/Agent/Agent.tsx";
 import RatioTaskDetail from "@/pages/RatioTask/Detail/RatioTaskDetail";
 import CleansingTemplateDetail from "@/pages/DataCleansing/Detail/TemplateDetail";
+import EvaluationDetailPage from "@/pages/DataEvaluation/Detail/TaskDetail.tsx";
 
 const router = createBrowserRouter([
   {
@@ -189,8 +189,8 @@ const router = createBrowserRouter([
             Component: DataEvaluationPage,
           },
           {
-            path: "create-task",
-            Component: EvaluationTaskCreate,
+            path: "detail/:id",
+            Component: EvaluationDetailPage,
           },
           {
             path: "task-report/:id",
diff --git a/runtime/datamate-python/app/core/exception.py b/runtime/datamate-python/app/core/exception.py
new file mode 100644
index 0000000..0c4b096
--- /dev/null
+++ b/runtime/datamate-python/app/core/exception.py
@@ -0,0 +1,17 @@
+from enum import Enum
+
+class BusinessErrorCode:
+    def __init__(self, message: str, error_code: str):
+        self.message = message
+        self.error_code = error_code
+
+
+class BusinessException(RuntimeError):
+    def __init__(self, business_error_code: BusinessErrorCode):
+        self.message = business_error_code.message
+        self.error_code = business_error_code.error_code
+        super().__init__(self.message)
+
+
+class BusinessErrorCodeEnum(Enum):
+    TASK_TYPE_ERROR = BusinessErrorCode("任务类型错误", "evaluation.0001")
diff --git a/runtime/datamate-python/app/db/models/__init__.py b/runtime/datamate-python/app/db/models/__init__.py
index 5a237f3..2b83de2 100644
--- a/runtime/datamate-python/app/db/models/__init__.py
+++ b/runtime/datamate-python/app/db/models/__init__.py
@@ -16,6 +16,11 @@ from .annotation_management import (
     LabelingProject
 )
 
+from .data_evaluation import (
+    EvaluationTask,
+    EvaluationItem
+)
+
 __all__ = [
     "Dataset",
     "DatasetTag",
@@ -25,4 +30,6 @@ __all__ = [
     "User",
     "AnnotationTemplate",
     "LabelingProject",
-]
\ No newline at end of file
+    "EvaluationTask",
+    "EvaluationItem",
+]
diff --git a/runtime/datamate-python/app/db/models/data_evaluation.py b/runtime/datamate-python/app/db/models/data_evaluation.py
new file mode 100644
index 0000000..5234064
--- /dev/null
+++ b/runtime/datamate-python/app/db/models/data_evaluation.py
@@ -0,0 +1,81 @@
+"""
+Tables for Data Evaluation module
+
+Derived from scripts/db/data-evaluation-init.sql
+ - t_de_eval_task
+ - t_de_eval_item
+"""
+
+import uuid
+from sqlalchemy import Column, String, Text, Float, TIMESTAMP, ForeignKey, Integer
+from sqlalchemy.sql import func
+
+from app.db.session import Base
+
+
+class EvaluationTask(Base):
+    """评估任务表（UUID 主键） -> t_de_eval_task
+
+    Columns per data-evaluation-init.sql:
+      id, name, description, task_type, source_type, source_id, source_name,
+      status, eval_process, eval_promt, eval_config, created_at, updated_at,
+      created_by, updated_by
+    """
+
+    __tablename__ = "t_de_eval_task"
+
+    id = Column(String(36), primary_key=True, default=lambda: str(uuid.uuid4()), comment="UUID")
+    name = Column(String(255), nullable=False, comment="评估任务名称")
+    description = Column(Text, nullable=True, comment="评估任务描述")
+    task_type = Column(String(50), nullable=False, comment="评估任务类型：QA")
+    source_type = Column(String(36), nullable=True, comment="待评估对象类型：DATASET/SYNTHESIS")
+    source_id = Column(String(36), nullable=True, comment="待评估对象ID")
+    source_name = Column(String(255), nullable=True, comment="待评估对象名称")
+    status = Column(String(50), server_default="PENDING", nullable=False, comment="状态：PENDING/RUNNING/COMPLETED/STOPPED/FAILED")
+    eval_process = Column(Float, nullable=False, server_default="0", comment="评估进度")
+    eval_prompt = Column(Text, nullable=True, comment="评估提示词")
+    eval_config = Column(Text, nullable=True, comment="评估配置")
+    created_at = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="创建时间")
+    updated_at = Column(TIMESTAMP, server_default=func.current_timestamp(), onupdate=func.current_timestamp(), comment="更新时间")
+    created_by = Column(String(255), nullable=True, comment="创建者")
+    updated_by = Column(String(255), nullable=True, comment="更新者")
+
+
+class EvaluationFile(Base):
+    """评估条目表（UUID 主键） -> t_de_eval_file"""
+
+    __tablename__ = "t_de_eval_file"
+
+    id = Column(String(36), primary_key=True, default=lambda: str(uuid.uuid4()), comment="UUID")
+    task_id = Column(String(36), ForeignKey('t_de_eval_task.id'), nullable=False, comment="评估任务ID")
+    file_id = Column(String(36), ForeignKey('t_dm_dataset_files.id'), nullable=True, comment="文件ID")
+    file_name = Column(String(255), nullable=False, comment="文件名")
+    total_count = Column(Integer, nullable=False, default=0, comment="总数")
+    evaluated_count = Column(Integer, nullable=False, default=0, comment="已评估数")
+    created_at = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="创建时间")
+    updated_at = Column(TIMESTAMP, server_default=func.current_timestamp(), onupdate=func.current_timestamp(), comment="更新时间")
+    created_by = Column(String(255), nullable=True, comment="创建者")
+    updated_by = Column(String(255), nullable=True, comment="更新者")
+
+
+class EvaluationItem(Base):
+    """评估条目表（UUID 主键） -> t_de_eval_item
+
+    Columns per data-evaluation-init.sql:
+      id, task_id, item_id, eval_score, eval_result, status
+    """
+
+    __tablename__ = "t_de_eval_item"
+
+    id = Column(String(36), primary_key=True, default=lambda: str(uuid.uuid4()), comment="UUID")
+    task_id = Column(String(36), ForeignKey('t_de_eval_task.id'), nullable=False, comment="评估任务ID")
+    file_id = Column(String(36), ForeignKey('t_dm_dataset_files.id'), nullable=True, comment="文件ID")
+    item_id = Column(String(36), nullable=False, comment="评估条目ID")
+    eval_content = Column(Text, nullable=True, comment="评估内容")
+    eval_score = Column(Float, nullable=False, server_default="0", comment="评估分数")
+    eval_result = Column(Text, nullable=True, comment="评估结果")
+    status = Column(String(50), server_default="PENDING", nullable=False, comment="状态：PENDING/EVALUATED")
+    created_at = Column(TIMESTAMP, server_default=func.current_timestamp(), comment="创建时间")
+    updated_at = Column(TIMESTAMP, server_default=func.current_timestamp(), onupdate=func.current_timestamp(), comment="更新时间")
+    created_by = Column(String(255), nullable=True, comment="创建者")
+    updated_by = Column(String(255), nullable=True, comment="更新者")
diff --git a/runtime/datamate-python/app/module/__init__.py b/runtime/datamate-python/app/module/__init__.py
index 0295370..2eea538 100644
--- a/runtime/datamate-python/app/module/__init__.py
+++ b/runtime/datamate-python/app/module/__init__.py
@@ -2,8 +2,9 @@ from fastapi import APIRouter
 
 from .system.interface import router as system_router
 from .annotation.interface import router as annotation_router
-from .synthesis.interface import router as ratio_router
+from .ratio.interface import router as ratio_router
 from .generation.interface import router as generation_router
+from .evaluation.interface import router as evaluation_router
 
 router = APIRouter(
     prefix="/api"
@@ -13,5 +14,6 @@ router.include_router(system_router)
 router.include_router(annotation_router)
 router.include_router(ratio_router)
 router.include_router(generation_router)
+router.include_router(evaluation_router)
 
 __all__ = ["router"]
diff --git a/runtime/datamate-python/app/module/dataset/schema/dataset_file.py b/runtime/datamate-python/app/module/dataset/schema/dataset_file.py
index bf9269c..f31bba8 100644
--- a/runtime/datamate-python/app/module/dataset/schema/dataset_file.py
+++ b/runtime/datamate-python/app/module/dataset/schema/dataset_file.py
@@ -46,8 +46,7 @@ class DatasetFileTag(BaseModel):
             tags.append(tag_values)
         # 如果 from_name 不为空，添加前缀
         if self.from_name:
-            tags = [f"{self.from_name} {tag}" for tag in tags]
-
+            tags = [f"{self.from_name}@{tag}" for tag in tags]
         return tags
 
 
diff --git a/runtime/datamate-python/app/module/synthesis/__init__.py b/runtime/datamate-python/app/module/evaluation/__init__.py
similarity index 100%
rename from runtime/datamate-python/app/module/synthesis/__init__.py
rename to runtime/datamate-python/app/module/evaluation/__init__.py
diff --git a/runtime/datamate-python/app/module/evaluation/interface/__init__.py b/runtime/datamate-python/app/module/evaluation/interface/__init__.py
new file mode 100644
index 0000000..d770ef7
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/interface/__init__.py
@@ -0,0 +1,11 @@
+from fastapi import APIRouter
+
+router = APIRouter(
+    prefix="/evaluation",
+    tags = ["evaluation"]
+)
+
+# Include sub-routers
+from .evaluation import router as evaluation_router
+
+router.include_router(evaluation_router)
diff --git a/runtime/datamate-python/app/module/evaluation/interface/evaluation.py b/runtime/datamate-python/app/module/evaluation/interface/evaluation.py
new file mode 100644
index 0000000..9b62bca
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/interface/evaluation.py
@@ -0,0 +1,429 @@
+import asyncio
+import uuid
+import json
+from typing import Optional
+from fastapi import APIRouter, Depends, HTTPException, Query, BackgroundTasks
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy import select, func, or_, text, and_
+from pydantic import ValidationError
+
+from app.core.logging import get_logger
+from app.db.models.data_evaluation import EvaluationFile
+from app.db.session import get_db
+from app.db.models import EvaluationTask, EvaluationItem, DatasetFiles
+from app.module.evaluation.schema.evaluation import (
+    CreateEvaluationTaskRequest,
+    PagedEvaluationTaskResponse,
+    EvaluationTaskDetailResponse,
+    PagedEvaluationItemsResponse,
+    EvaluationItemResponse, PagedEvaluationFilesResponse, EvaluationFileResponse
+)
+from app.module.evaluation.schema.prompt import get_prompt
+from app.module.evaluation.schema.prompt_template import PromptTemplateResponse
+from app.module.evaluation.service.prompt_template_service import PromptTemplateService
+from app.module.evaluation.service.evaluation import EvaluationTaskService
+from app.module.shared.schema.common import StandardResponse, TaskStatus
+
+router = APIRouter(
+    prefix="",
+    tags=["evaluation"],
+)
+
+logger = get_logger(__name__)
+
+
+@router.get("/prompt-templates", response_model=StandardResponse[PromptTemplateResponse])
+async def get_prompt_templates():
+    """
+    Get all available evaluation prompt templates
+
+    Returns:
+        StandardResponse with list of prompt templates
+    """
+    try:
+        templates = PromptTemplateService.get_prompt_templates()
+        return StandardResponse(
+            code=200,
+            message="Success",
+            data=templates
+        )
+    except Exception as e:
+        logger.error(f"Failed to get prompt templates: {str(e)}")
+        raise HTTPException(
+            status_code=500,
+            detail="Failed to retrieve prompt templates"
+        )
+
+
+@router.post("/tasks", response_model=StandardResponse[EvaluationTaskDetailResponse], status_code=201)
+async def create_evaluation_task(
+    request: CreateEvaluationTaskRequest,
+    db: AsyncSession = Depends(get_db)
+):
+    """
+    创建评估任务
+
+    Args:
+        request: 创建评估任务请求
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[EvaluationTaskDetailResponse]: 创建的任务详情
+    """
+    try:
+        # 检查任务名称是否已存在
+        existing_task = await db.execute(
+            select(EvaluationTask).where(EvaluationTask.name == request.name)
+        )
+        if existing_task.scalar_one_or_none():
+            raise HTTPException(status_code=400, detail=f"Evaluation task with name '{request.name}' already exists")
+
+        # 创建评估任务
+        task = EvaluationTask(
+            id=str(uuid.uuid4()),
+            name=request.name,
+            description=request.description,
+            task_type=request.task_type,
+            source_type=request.source_type,
+            source_id=request.source_id,
+            source_name=request.source_name,
+            eval_prompt=request.eval_prompt,
+            eval_config=json.dumps({
+                "model_id": request.eval_config.model_id,
+                "dimensions": request.eval_config.dimensions,
+            }),
+            status=TaskStatus.PENDING.value,
+            eval_process=0.0,
+        )
+
+        db.add(task)
+        # Commit first to persist the task before scheduling background work
+        await db.commit()
+        # Schedule background execution without blocking the current request
+        asyncio.create_task(EvaluationTaskService.run_evaluation_task(task.id))
+
+        # Refresh the task to return latest state
+        await db.refresh(task)
+
+        # 转换响应模型
+        response = _map_to_task_detail_response(task)
+        return StandardResponse(
+            code=200,
+            message="Evaluation task created successfully",
+            data=response
+        )
+
+    except ValidationError as e:
+        await db.rollback()
+        logger.error(f"Validation error: {str(e)}")
+        raise HTTPException(status_code=400, detail=str(e))
+    except Exception as e:
+        await db.rollback()
+        logger.error(f"Failed to create evaluation task: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+
+@router.get("/tasks", response_model=StandardResponse[PagedEvaluationTaskResponse])
+async def list_evaluation_tasks(
+    page: int = Query(1, ge=1, description="页码，从1开始"),
+    size: int = Query(10, ge=1, le=100, description="每页数量"),
+    name: Optional[str] = Query(None, description="任务名称模糊查询"),
+    status: Optional[str] = Query(None, description="任务状态过滤"),
+    task_type: Optional[str] = Query(None, description="任务类型过滤"),
+    db: AsyncSession = Depends(get_db),
+):
+    """
+    分页查询评估任务
+
+    Args:
+        page: 页码，从1开始
+        size: 每页数量
+        name: 任务名称模糊查询
+        status: 任务状态过滤
+        task_type: 任务类型过滤
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[PagedEvaluationTaskResponse]: 分页的评估任务列表
+    """
+    try:
+        # 构建查询条件
+        query = select(EvaluationTask)
+
+        if name:
+            query = query.where(EvaluationTask.name.ilike(f"%{name}%"))
+        if status:
+            query = query.where(EvaluationTask.status == status)
+        if task_type:
+            query = query.where(EvaluationTask.task_type == task_type)
+
+        # 获取总数
+        count_query = select(func.count()).select_from(query.subquery())
+        total = (await db.execute(count_query)).scalar_one()
+
+        # 分页查询
+        offset = (page - 1) * size
+        tasks = (await db.execute(
+            query.order_by(EvaluationTask.created_at.desc())
+                 .offset(offset)
+                 .limit(size)
+        )).scalars().all()
+
+        # 转换为响应模型
+        items = [_map_to_task_detail_response(task) for task in tasks]
+        total_pages = (total + size - 1) // size if size > 0 else 0
+
+        return StandardResponse(
+            code=200,
+            message="Success",
+            data=PagedEvaluationTaskResponse(
+                content=items,
+                totalElements=total,
+                totalPages=total_pages,
+                page=page,
+                size=size,
+            )
+        )
+
+    except Exception as e:
+        logger.error(f"Failed to list evaluation tasks: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+@router.get("/tasks/{task_id}/files", response_model=StandardResponse[PagedEvaluationFilesResponse])
+async def list_evaluation_items(
+    task_id: str,
+    page: int = Query(1, ge=1, description="页码，从1开始"),
+    size: int = Query(10, ge=1, le=100, description="每页数量"),
+    db: AsyncSession = Depends(get_db),
+):
+    """
+    分页查询评估文件
+
+    Args:
+        task_id: 评估任务ID
+        page: 页码，从1开始
+        size: 每页数量
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[PagedEvaluationFilesResponse]: 分页的评估文件列表
+    """
+    try:
+        task = await db.get(EvaluationTask, task_id)
+        if not task:
+            raise HTTPException(status_code=404, detail="Evaluation task not found")
+        offset = (page - 1) * size
+        query = select(EvaluationFile).where(EvaluationFile.task_id == task_id)
+        count_query = select(func.count()).select_from(query.subquery())
+        total = (await db.execute(count_query)).scalar_one()
+        files = (await db.execute(query.offset(offset).limit(size))).scalars().all()
+        total_pages = (total + size - 1) // size if size > 0 else 0
+        file_responses = [
+            EvaluationFileResponse(
+                taskId=file.task_id,
+                fileId=file.file_id,
+                fileName=file.file_name,
+                totalCount=file.total_count,
+                evaluatedCount=file.evaluated_count,
+                pendingCount=file.total_count - file.evaluated_count
+            )
+            for file in files
+        ]
+        return StandardResponse(
+            code=200,
+            message="Success",
+            data=PagedEvaluationFilesResponse(
+                content=file_responses,
+                totalElements=total,
+                totalPages=total_pages,
+                page=page,
+                size=size,
+            )
+        )
+    except Exception as e:
+        logger.error(f"Failed to list evaluation items: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+
+@router.get("/tasks/{task_id}/items", response_model=StandardResponse[PagedEvaluationItemsResponse])
+async def list_evaluation_items(
+    task_id: str,
+    page: int = Query(1, ge=1, description="页码，从1开始"),
+    size: int = Query(10, ge=1, le=100, description="每页数量"),
+    status: Optional[str] = Query(None, description="状态过滤"),
+    file_id: Optional[str] = Query(None, description="文件过滤"),
+    db: AsyncSession = Depends(get_db),
+):
+    """
+    分页查询评估条目
+
+    Args:
+        task_id: 评估任务ID
+        page: 页码，从1开始
+        size: 每页数量
+        status: 状态过滤
+        file_id: 文件过滤
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[PagedEvaluationItemsResponse]: 分页的评估条目列表
+    """
+    try:
+        # 检查任务是否存在
+        task = await db.get(EvaluationTask, task_id)
+        if not task:
+            raise HTTPException(status_code=404, detail="Evaluation task not found")
+
+        # 构建查询条件
+        query = select(EvaluationItem).where(EvaluationItem.task_id == task_id)
+
+        if status:
+            query = query.where(EvaluationItem.status == status)
+
+        if file_id:
+            query = query.where(EvaluationItem.file_id == file_id)
+
+        # 获取总数
+        count_query = select(func.count()).select_from(query.subquery())
+        total = (await db.execute(count_query)).scalar_one()
+
+        # 分页查询
+        offset = (page - 1) * size
+        items = (await db.execute(query.offset(offset).limit(size))).scalars().all()
+
+        # 转换为响应模型
+        item_responses = [
+            EvaluationItemResponse(
+                id=item.id,
+                taskId=item.task_id,
+                itemId=item.item_id,
+                fileId=item.file_id,
+                evalContent=json.loads(item.eval_content),
+                evalScore=float(item.eval_score) if item.eval_score else None,
+                evalResult=json.loads(item.eval_result),
+                status=item.status
+            )
+            for item in items
+        ]
+
+        total_pages = (total + size - 1) // size if size > 0 else 0
+
+        return StandardResponse(
+            code=200,
+            message="Success",
+            data=PagedEvaluationItemsResponse(
+                content=item_responses,
+                totalElements=total,
+                totalPages=total_pages,
+                page=page,
+                size=size,
+            )
+        )
+    except Exception as e:
+        logger.error(f"Failed to list evaluation items: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+
+@router.get("/tasks/{task_id}", response_model=StandardResponse[EvaluationTaskDetailResponse])
+async def get_evaluation_task(
+    task_id: str,
+    db: AsyncSession = Depends(get_db),
+):
+    """
+    获取评估任务详情
+
+    Args:
+        task_id: 任务ID
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[EvaluationTaskDetailResponse]: 评估任务详情
+    """
+    try:
+        task = await db.get(EvaluationTask, task_id)
+        if not task:
+            raise HTTPException(status_code=404, detail="Evaluation task not found")
+
+        # 转换为响应模型
+        response = _map_to_task_detail_response(task)
+        return StandardResponse(
+            code=200,
+            message="Success",
+            data=response
+        )
+
+    except HTTPException:
+        raise
+    except Exception as e:
+        logger.error(f"Failed to get evaluation task: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+
+@router.delete("/tasks", response_model=StandardResponse[str], status_code=200)
+async def delete_eval_tasks(
+    ids: list[str] = Query(..., description="要删除的评估任务ID列表"),
+    db: AsyncSession = Depends(get_db),
+):
+    """
+    删除评估任务
+
+    Args:
+        ids: 任务ID
+        db: 数据库会话
+
+    Returns:
+        StandardResponse[str]: 删除结果
+    """
+    try:
+        # 检查任务是否存在
+        task_id = ids[0]
+        task = await db.get(EvaluationTask, task_id)
+        if not task:
+            raise HTTPException(status_code=404, detail="Evaluation task not found")
+
+        # 删除评估项
+        await db.execute(
+            EvaluationItem.__table__.delete()
+            .where(EvaluationItem.task_id == task_id)
+        )
+
+        # 删除任务
+        await db.delete(task)
+        await db.commit()
+
+        return StandardResponse(
+            code=200,
+            message="Evaluation task deleted successfully",
+            data="success"
+        )
+
+    except HTTPException:
+        await db.rollback()
+        raise
+    except Exception as e:
+        await db.rollback()
+        logger.error(f"Failed to delete evaluation task: {str(e)}")
+        raise HTTPException(status_code=500, detail="Internal server error")
+
+
+def _map_to_task_detail_response(
+    task: EvaluationTask
+) -> EvaluationTaskDetailResponse:
+    """将数据库模型转换为任务详情响应模型"""
+    task_response = EvaluationTaskDetailResponse(
+        id=task.id,
+        name=task.name,
+        description=task.description,
+        taskType=task.task_type,
+        sourceType=task.source_type,
+        sourceId=task.source_id,
+        sourceName=task.source_name,
+        status=task.status,
+        evalProcess=task.eval_process,
+        evalPrompt=task.eval_prompt,
+        evalConfig=json.loads(task.eval_config),
+        createdAt=task.created_at.isoformat() if task.created_at else None,
+        updatedAt=task.updated_at.isoformat() if task.updated_at else None,
+    )
+    task_response.eval_prompt = get_prompt(task_response.task_type, task_response.eval_config.get("dimensions"))
+    return task_response
diff --git a/runtime/datamate-python/app/module/synthesis/schema/__init__.py b/runtime/datamate-python/app/module/evaluation/schema/__init__.py
similarity index 100%
rename from runtime/datamate-python/app/module/synthesis/schema/__init__.py
rename to runtime/datamate-python/app/module/evaluation/schema/__init__.py
diff --git a/runtime/datamate-python/app/module/evaluation/schema/evaluation.py b/runtime/datamate-python/app/module/evaluation/schema/evaluation.py
new file mode 100644
index 0000000..ffdd41b
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/schema/evaluation.py
@@ -0,0 +1,101 @@
+from typing import List, Optional, Dict, Any
+from pydantic import BaseModel, Field, field_validator
+from enum import Enum
+
+from app.core.logging import get_logger
+from app.module.shared.schema.common import TaskStatus
+
+logger = get_logger(__name__)
+
+class EvaluationConfig(BaseModel):
+    """评估配置项"""
+    model_id: str = Field(..., alias="modelId", description="模型id")
+    dimensions: list[dict] = Field(..., alias="dimensions", description="评估维度")
+
+
+class CreateEvaluationTaskRequest(BaseModel):
+    """创建评估任务请求"""
+    name: str = Field(..., description="评估任务名称")
+    description: Optional[str] = Field(None, description="评估任务描述")
+    task_type: str = Field(..., alias="taskType", description="评估任务类型：QA/QUALITY/COMPATIBILITY/VALUE")
+    source_type: str = Field(..., alias="sourceType", description="待评估对象类型：DATASET/SYNTHESIS")
+    source_id: str = Field(..., alias="sourceId", description="待评估对象ID")
+    source_name: str = Field(..., alias="sourceName", description="待评估对象名称")
+    eval_method: str = Field("AUTO", alias="evalMethod", description="评估提示词")
+    eval_prompt: Optional[str] = Field(None, alias="evalPrompt", description="评估提示词")
+    eval_config: EvaluationConfig = Field(..., alias="evalConfig", description="评估配置项列表")
+
+
+class EvaluationTaskItem(BaseModel):
+    """评估任务列表项"""
+    id: str = Field(..., description="任务ID")
+    name: str = Field(..., description="任务名称")
+    description: Optional[str] = Field(None, description="任务描述")
+    task_type: Optional[str] = Field(..., alias="taskType", description="任务类型")
+    source_type: Optional[str] = Field(..., alias="sourceType", description="数据源类型")
+    source_id: Optional[str] = Field(..., alias="sourceId", description="数据源ID")
+    source_name: Optional[str] = Field(None, alias="sourceName", description="数据源名称")
+    status: TaskStatus = Field(..., description="任务状态")
+    eval_process: Optional[float] = Field(0, alias="evalProcess", description="评估进度")
+    created_at: Optional[str] = Field(None, alias="createdAt", description="创建时间")
+    updated_at: Optional[str] = Field(None, alias="updatedAt", description="更新时间")
+
+
+class PagedEvaluationTaskResponse(BaseModel):
+    """分页评估任务响应"""
+    content: List[EvaluationTaskItem]
+    total_elements: int = Field(..., alias="totalElements")
+    total_pages: int = Field(..., alias="totalPages")
+    page: int
+    size: int
+
+
+class EvaluationTaskDetailResponse(EvaluationTaskItem):
+    """评估任务详情响应"""
+    eval_prompt: Optional[str] = Field(None, alias="evalPrompt", description="评估提示词")
+    eval_config: Optional[Dict[str, Any]] = Field(None, alias="evalConfig", description="评估配置")
+    eval_result: Optional[Dict[str, Any]] = Field(None, alias="evalResult", description="评估结果")
+
+
+class EvaluationItemResponse(BaseModel):
+    """评估条目响应"""
+    id: str = Field(..., description="条目ID")
+    task_id: str = Field(..., alias="taskId", description="任务ID")
+    file_id: str = Field(..., alias="fileId", description="文件ID")
+    item_id: str = Field(..., alias="itemId", description="评估项ID")
+    eval_content: Optional[Dict[str, Any]] = Field(None, alias="evalContent", description="评估内容")
+    eval_score: Optional[float] = Field(None, alias="evalScore", description="评估分数")
+    eval_result: Optional[Dict[str, Any]] = Field(None, alias="evalResult", description="评估结果详情")
+    status: str = Field(..., description="评估状态")
+
+class EvaluationFileResponse(BaseModel):
+    """评估文件响应"""
+    task_id: str = Field(..., alias="taskId", description="任务ID")
+    file_id: str = Field(..., alias="fileId", description="文件ID")
+    file_name: str = Field(..., alias="fileName", description="文件名")
+    total_count: int = Field(..., alias="totalCount", description="总数")
+    evaluated_count: int = Field(..., alias="evaluatedCount", description="已评估数")
+    pending_count: int = Field(..., alias="pendingCount", description="待评估数")
+
+
+
+class PagedEvaluationItemsResponse(BaseModel):
+    """分页评估任务响应"""
+    content: List[EvaluationItemResponse]
+    total_elements: int = Field(..., alias="totalElements")
+    total_pages: int = Field(..., alias="totalPages")
+    page: int
+    size: int
+
+class PagedEvaluationFilesResponse(BaseModel):
+    """分页评估任务响应"""
+    content: List[EvaluationFileResponse]
+    total_elements: int = Field(..., alias="totalElements")
+    total_pages: int = Field(..., alias="totalPages")
+    page: int
+    size: int
+
+
+class SourceType(Enum):
+    DATASET = "DATASET"
+    SYNTHESIS = "SYNTHESIS"
diff --git a/runtime/datamate-python/app/module/evaluation/schema/prompt.py b/runtime/datamate-python/app/module/evaluation/schema/prompt.py
new file mode 100644
index 0000000..b151d71
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/schema/prompt.py
@@ -0,0 +1,87 @@
+EVALUATION_PROMPT_TEMPLATE = [
+    {
+        "evalType": "QA",
+        "defaultDimensions": [
+            {
+                "dimension": "问题是否独立",
+                "description": "仅分析问题，问题的主体和客体都比较明确，即使有省略，也符合语言习惯。在不需要补充其他信息的情况下不会引起疑惑。"
+            },
+            {
+                "dimension": "语法是否错误",
+                "description": "问题为疑问句，答案为陈述句; 不存在词语搭配不当的情况;连接词和标点符号不存在错用情况；逻辑混乱的情况不存在；语法结构都正确且完整。"
+            },
+            {
+                "dimension": "回答是否有针对性",
+                "description": "回答应对问题中的所有疑问点提供正面、直接的回答，不应引起疑惑。同时，答案不应有任何内容的遗漏，需构成一个完整的陈述。"
+            }
+        ],
+        "prompt": """
+# Role: 问答对质量评估专家
+## Profile:
+- Description: 你是一名专业的对话文本质量评估专家，擅长从多个维度对问答对进行质量评估，为机器学习模型训练提供高质量的数据筛选建议。具备深度学习、自然语言处理和数据科学的专业背景。
+
+## Skills:
+1. 能够从多个维度对问答对进行综合评估
+2. 擅长识别问答对中的潜在问题，如答案不准确、问题模糊、文本不匹配、逻辑错误等
+3. 能够给出具体的改进建议和质量评分，并提供可操作的优化方案
+4. 熟悉机器学习训练数据的质量标准和最佳实践
+5. 能够区分不同类型的问题（事实性、推理性、创造性）并采用相应的评估标准
+
+## 评估维度:
+{dimensions}
+
+## 原始文本块内容:
+{content}
+
+## 问题:
+{question}
+
+## 答案:
+{answer}
+
+## 评估说明:
+1. **数据集类型识别**：如果原始文本块内容为空或显示"Distilled Content"，说明这是一个蒸馏数据集，没有原始文本参考。请重点评估问题的质量、答案的合理性和逻辑性，以及问答的一致性。
+2. **评估原则**：采用严格的评估标准，确保筛选出的数据集能够有效提升模型性能。
+
+## 注意事项:
+- 评估结论要具体指出优点和不足，提供可操作的改进建议
+- 评估结论控制在150字以内，简洁明了但要涵盖关键信息
+
+## 输出要求:
+请按照以下JSON格式输出评估结果，评估结果为Y/N，符合标注输出Y，不符合标准输出N：
+
+{
+  "result": {{result_example}
+  },
+  "evaluation": "这是一个高质量的问答数据集。问题表述清晰具体，答案准确完整且逻辑性强，与原始文本高度相关。建议：可以进一步丰富答案的细节描述。"
+}
+"""
+    }
+]
+
+def get_dimensions_for_qa(dimensions: list[dict]) -> str:
+    dimensions_str = "\n"
+    index = 1
+    for dimension in dimensions:
+        dimensions_str += f"### {index}. {dimension.get("dimension")}\n**评估标准：**\n{dimension.get("description")}\n\n"
+        index += 1
+    return dimensions_str
+
+def get_result_example_for_qa(dimensions: list[dict]) -> str:
+    result_example = ""
+    for dimension in dimensions:
+        result_example += f'\n    "{dimension.get("dimension")}": "Y",'
+    return result_example
+
+def get_prompt(task_type: str, dimensions: list[dict]) -> str:
+    template = None
+    for t in EVALUATION_PROMPT_TEMPLATE:
+        if t.get("evalType") == task_type:
+            template = t.get("prompt")
+            break
+    if not template:
+        template = EVALUATION_PROMPT_TEMPLATE[0].get("prompt", "")
+    if not dimensions or len(dimensions) == 0:
+        return template
+    return (template.replace("{dimensions}", get_dimensions_for_qa(dimensions))
+                       .replace("{result_example}", get_result_example_for_qa(dimensions)))
diff --git a/runtime/datamate-python/app/module/evaluation/schema/prompt_template.py b/runtime/datamate-python/app/module/evaluation/schema/prompt_template.py
new file mode 100644
index 0000000..e25166a
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/schema/prompt_template.py
@@ -0,0 +1,29 @@
+"""
+Schema for evaluation prompt templates.
+"""
+from typing import List, Dict, Any
+from pydantic import BaseModel, Field
+
+
+class PromptTemplateDimension(BaseModel):
+    """A single dimension in the prompt template"""
+    dimension: str = Field(..., description="Dimension name")
+    description: str = Field(..., description="Description of the dimension")
+
+
+class PromptTemplateItem(BaseModel):
+    """A single prompt template item"""
+    evalType: str = Field(..., description="Evaluation type")
+    defaultDimensions: List[PromptTemplateDimension] = Field(
+        default_factory=list,
+        description="List of default dimensions for this evaluation type"
+    )
+    prompt: str = Field(..., description="The prompt template string")
+
+
+class PromptTemplateResponse(BaseModel):
+    """Response model for getting prompt templates"""
+    templates: List[PromptTemplateItem] = Field(
+        ...,
+        description="List of available prompt templates"
+    )
diff --git a/runtime/datamate-python/app/module/synthesis/service/__init__.py b/runtime/datamate-python/app/module/evaluation/service/__init__.py
similarity index 100%
rename from runtime/datamate-python/app/module/synthesis/service/__init__.py
rename to runtime/datamate-python/app/module/evaluation/service/__init__.py
diff --git a/runtime/datamate-python/app/module/evaluation/service/evaluation.py b/runtime/datamate-python/app/module/evaluation/service/evaluation.py
new file mode 100644
index 0000000..d187ae7
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/service/evaluation.py
@@ -0,0 +1,207 @@
+import json
+import uuid
+import asyncio
+
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.core.exception import BusinessErrorCodeEnum, BusinessException
+from app.core.logging import get_logger
+from app.db.models import EvaluationItem, EvaluationTask, DatasetFiles
+from app.db.models.data_evaluation import EvaluationFile
+from app.db.models.data_synthesis import DataSynthesisFileInstance, SynthesisData
+from app.db.session import AsyncSessionLocal
+from app.module.evaluation.schema.evaluation import SourceType
+from app.module.shared.schema import TaskStatus
+from app.module.shared.util.model_chat import call_openai_style_model
+from app.module.evaluation.schema.prompt import get_prompt
+from app.module.shared.util.structured_file import StructuredFileHandlerFactory
+from app.module.system.service.common_service import get_model_by_id
+
+logger = get_logger(__name__)
+
+class EvaluationExecutor:
+    def __init__(self, db: AsyncSession, task: EvaluationTask):
+        self.db = db
+        self.task = task
+
+    async def save_eval_items(self):
+        pass
+
+    def get_eval_prompt(self, item: EvaluationItem) -> str:
+        prompt_text = get_prompt(self.task.task_type, json.loads(self.task.eval_config).get("dimensions"))
+        eval_content = json.loads(item.eval_content)
+        if self.task.task_type == "QA":
+            prompt_text = ((prompt_text.replace("{content}", eval_content.get("input"))
+                            .replace("{question}", eval_content.get("instruction")))
+                           .replace("{answer}", eval_content.get("output")))
+        return prompt_text
+
+    async def execute(self):
+        eval_config = json.loads(self.task.eval_config)
+        model_config = await get_model_by_id(self.db, eval_config.get("model_id"))
+        semaphore = asyncio.Semaphore(10)
+        files = (await self.db.execute(
+            select(EvaluationFile).where(EvaluationFile.task_id == self.task.id)
+        )).scalars().all()
+        for file in files:
+            items = (await self.db.execute(
+                select(EvaluationItem).where(EvaluationItem.task_id == self.task.id)
+                .where(EvaluationItem.file_id == file.file_id)
+            )).scalars().all()
+            tasks = [
+                self.evaluate_item(model_config, item, semaphore)
+                for item in items
+            ]
+            await asyncio.gather(*tasks, return_exceptions=True)
+            file.evaluated_count = len(items)
+            await self.db.commit()
+
+    async def evaluate_item(self, model_config, item: EvaluationItem, semaphore: asyncio.Semaphore):
+        async with semaphore:
+            prompt_text = self.get_eval_prompt(item)
+            resp_text = await asyncio.to_thread(
+                call_openai_style_model, model_config.base_url, model_config.api_key, model_config.model_name,
+                prompt_text,
+            )
+            item.eval_result = resp_text
+            item.status = TaskStatus.COMPLETED.value
+            await self.db.commit()
+
+
+    def get_source_type(self) -> SourceType:
+        pass
+
+
+class DatasetEvaluationExecutor(EvaluationExecutor):
+    def __init__(self, db: AsyncSession, task: EvaluationTask):
+        super().__init__(db, task)
+
+    async def save_eval_items(self):
+        dataset_files = ((await self.db.execute(select(DatasetFiles)
+                                               .where(DatasetFiles.dataset_id == self.task.source_id)))
+                         .scalars().all())
+        handler = StructuredFileHandlerFactory().get_handler(self.task.task_type)
+        for dataset_file in dataset_files:
+            if dataset_file.file_type.upper() != "JSON" and dataset_file.file_type.upper() != "JSONL":
+                continue
+            items = handler.get_items_from_file(dataset_file.file_path)
+            logger.info(f"parse {len(items)} items from file {dataset_file.file_name}")
+            for item in items:
+                self.db.add(EvaluationItem(
+                    id=str(uuid.uuid4()),
+                    task_id=self.task.id,
+                    file_id=dataset_file.id,
+                    item_id=item.get("id") if item.get("id") else str(uuid.uuid4()),
+                    eval_content=json.dumps(item, ensure_ascii=False),
+                    status=TaskStatus.PENDING.value,
+                    created_by=self.task.created_by,
+                    updated_by=self.task.updated_by,
+                ))
+            self.db.add(EvaluationFile(
+                id=str(uuid.uuid4()),
+                task_id=self.task.id,
+                file_id=dataset_file.id,
+                file_name=dataset_file.file_name,
+                total_count=len(items),
+                evaluated_count=0,
+                created_by=self.task.created_by,
+                updated_by=self.task.updated_by,
+            ))
+
+    def get_source_type(self) -> SourceType:
+        return SourceType.DATASET
+
+
+class SynthesisEvaluationExecutor(EvaluationExecutor):
+    def __init__(self, db: AsyncSession, task: EvaluationTask):
+        super().__init__(db, task)
+
+    async def save_eval_items(self):
+        synthesis_files = ((await self.db.execute(select(DataSynthesisFileInstance)
+                               .where(DataSynthesisFileInstance.task_id == self.task.source_id)))
+                           .scalars().all())
+        for synthesis_file in synthesis_files:
+            synthesis_datas = ((await self.db.execute(select(SynthesisData)
+                                                     .where(SynthesisData.synthesis_file_instance_id == synthesis_file.id)))
+                               .scalars().all())
+            logger.info(f"get {len(synthesis_datas)} items from file {synthesis_file.file_name}")
+            for synthesis_data in synthesis_datas:
+                self.db.add(EvaluationItem(
+                    id=str(uuid.uuid4()),
+                    task_id=self.task.id,
+                    file_id=synthesis_file.id,
+                    item_id=synthesis_data.id,
+                    eval_content=synthesis_data.data,
+                    status=TaskStatus.PENDING.value,
+                    created_by=self.task.created_by,
+                    updated_by=self.task.updated_by,
+                ))
+            self.db.add(EvaluationFile(
+                id=str(uuid.uuid4()),
+                task_id=self.task.id,
+                file_id=synthesis_file.id,
+                file_name=synthesis_file.file_name,
+                total_count=len(synthesis_datas),
+                evaluated_count=0,
+                created_by=self.task.created_by,
+                updated_by=self.task.updated_by,
+            ))
+        pass
+
+    def get_source_type(self) -> SourceType:
+        return SourceType.SYNTHESIS
+
+
+class EvaluationExecutorFactory:
+    def __init__(self, db: AsyncSession, task: EvaluationTask):
+        self.db = db
+        self.executors: list[EvaluationExecutor] = []
+        self.executors.append(DatasetEvaluationExecutor(db, task))
+        self.executors.append(SynthesisEvaluationExecutor(db, task))
+
+    def get_executor(self, source_type: str) -> EvaluationExecutor:
+        for executor in self.executors:
+            if executor.get_source_type().value == source_type:
+                return executor
+        raise BusinessException(BusinessErrorCodeEnum.TASK_TYPE_ERROR.value)
+
+
+class EvaluationTaskService:
+
+    @staticmethod
+    async def run_evaluation_task(task_id: str):
+        """
+        Background worker to run evaluations.
+        - task_id: id of EvaluationTaskModel
+        """
+        logger.info(f"Background evaluation worker started add items for task {task_id}")
+        async with AsyncSessionLocal() as session:
+            try:
+                task = await session.execute(select(EvaluationTask).where(EvaluationTask.id == task_id))
+                task = task.scalar_one_or_none()
+                factory = EvaluationExecutorFactory(session, task)
+                executor = factory.get_executor(task.source_type)
+                await executor.save_eval_items()
+                task.status = TaskStatus.RUNNING.value
+            except Exception as e:
+                logger.error(f"Background worker encountered error for task {task_id}: {e}")
+                task.status = TaskStatus.FAILED.value
+            finally:
+                await session.commit()
+
+        logger.info(f"Background evaluation worker started for task {task_id}")
+        async with AsyncSessionLocal() as session:
+            try:
+                task = await session.execute(select(EvaluationTask).where(EvaluationTask.id == task_id))
+                task = task.scalar_one_or_none()
+                factory = EvaluationExecutorFactory(session, task)
+                executor = factory.get_executor(task.source_type)
+                await executor.execute()
+                logger.info(f"Background evaluation worker finished for task {task_id}")
+                task.status = TaskStatus.COMPLETED.value
+            except Exception as e:
+                logger.error(f"Background worker encountered error for task {task_id}: {e}")
+                task.status = TaskStatus.FAILED.value
+            finally:
+                await session.commit()
diff --git a/runtime/datamate-python/app/module/evaluation/service/prompt_template_service.py b/runtime/datamate-python/app/module/evaluation/service/prompt_template_service.py
new file mode 100644
index 0000000..aeea011
--- /dev/null
+++ b/runtime/datamate-python/app/module/evaluation/service/prompt_template_service.py
@@ -0,0 +1,45 @@
+"""
+Service for managing evaluation prompt templates.
+"""
+from typing import List, Dict, Any
+
+from app.module.evaluation.schema.prompt import EVALUATION_PROMPT_TEMPLATE
+from app.module.evaluation.schema.prompt_template import (
+    PromptTemplateItem,
+    PromptTemplateDimension,
+    PromptTemplateResponse
+)
+
+
+class PromptTemplateService:
+    """Service for managing evaluation prompt templates"""
+
+    @staticmethod
+    def get_prompt_templates() -> PromptTemplateResponse:
+        """
+        Get all available prompt templates
+        
+        Returns:
+            PromptTemplateResponse containing all prompt templates
+        """
+        templates = []
+        
+        for template in EVALUATION_PROMPT_TEMPLATE:
+            # Convert dimensions to the proper schema
+            dimensions = [
+                PromptTemplateDimension(
+                    dimension=dim.get("dimension"),
+                    description=dim.get("description", "")
+                )
+                for dim in template.get("defaultDimensions", [])
+            ]
+            
+            # Create template item
+            template_item = PromptTemplateItem(
+                evalType=template.get("evalType", ""),
+                defaultDimensions=dimensions,
+                prompt=template.get("prompt", "")
+            )
+            templates.append(template_item)
+        
+        return PromptTemplateResponse(templates=templates)
diff --git a/runtime/datamate-python/app/module/ratio/__init__.py b/runtime/datamate-python/app/module/ratio/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/runtime/datamate-python/app/module/synthesis/interface/__init__.py b/runtime/datamate-python/app/module/ratio/interface/__init__.py
similarity index 100%
rename from runtime/datamate-python/app/module/synthesis/interface/__init__.py
rename to runtime/datamate-python/app/module/ratio/interface/__init__.py
diff --git a/runtime/datamate-python/app/module/synthesis/interface/ratio_task.py b/runtime/datamate-python/app/module/ratio/interface/ratio_task.py
similarity index 99%
rename from runtime/datamate-python/app/module/synthesis/interface/ratio_task.py
rename to runtime/datamate-python/app/module/ratio/interface/ratio_task.py
index 09d3a16..8a3e13f 100644
--- a/runtime/datamate-python/app/module/synthesis/interface/ratio_task.py
+++ b/runtime/datamate-python/app/module/ratio/interface/ratio_task.py
@@ -13,7 +13,7 @@ from app.db.models import Dataset
 from app.db.session import get_db
 from app.module.dataset import DatasetManagementService
 from app.module.shared.schema import StandardResponse, TaskStatus
-from app.module.synthesis.schema.ratio_task import (
+from app.module.ratio.schema.ratio_task import (
     CreateRatioTaskResponse,
     CreateRatioTaskRequest,
     PagedRatioTaskResponse,
@@ -21,7 +21,7 @@ from app.module.synthesis.schema.ratio_task import (
     TargetDatasetInfo,
     RatioTaskDetailResponse,
 )
-from app.module.synthesis.service.ratio_task import RatioTaskService
+from app.module.ratio.service.ratio_task import RatioTaskService
 from app.db.models.ratio_task import RatioInstance, RatioRelation, RatioRelation as RatioRelationModel
 
 router = APIRouter(
diff --git a/runtime/datamate-python/app/module/ratio/schema/__init__.py b/runtime/datamate-python/app/module/ratio/schema/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/runtime/datamate-python/app/module/synthesis/schema/ratio_task.py b/runtime/datamate-python/app/module/ratio/schema/ratio_task.py
similarity index 94%
rename from runtime/datamate-python/app/module/synthesis/schema/ratio_task.py
rename to runtime/datamate-python/app/module/ratio/schema/ratio_task.py
index 1b01023..5e4791e 100644
--- a/runtime/datamate-python/app/module/synthesis/schema/ratio_task.py
+++ b/runtime/datamate-python/app/module/ratio/schema/ratio_task.py
@@ -7,9 +7,13 @@ from app.module.shared.schema.common import TaskStatus
 
 logger = get_logger(__name__)
 
+class LabelFilter(BaseModel):
+    label: Optional[str] = Field(..., description="标签")
+    value: Optional[str] = Field(None, description="标签值")
+
 class FilterCondition(BaseModel):
     date_range: Optional[str] = Field(None, description="数据范围", alias="dateRange")
-    label: Optional[str] = Field(None, description="标签")
+    label: Optional[LabelFilter] = Field(None, description="标签")
 
     @field_validator("date_range")
     @classmethod
diff --git a/runtime/datamate-python/app/module/ratio/service/__init__.py b/runtime/datamate-python/app/module/ratio/service/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/runtime/datamate-python/app/module/synthesis/service/ratio_task.py b/runtime/datamate-python/app/module/ratio/service/ratio_task.py
similarity index 97%
rename from runtime/datamate-python/app/module/synthesis/service/ratio_task.py
rename to runtime/datamate-python/app/module/ratio/service/ratio_task.py
index ea0a1bb..943d23b 100644
--- a/runtime/datamate-python/app/module/synthesis/service/ratio_task.py
+++ b/runtime/datamate-python/app/module/ratio/service/ratio_task.py
@@ -7,7 +7,6 @@ import shutil
 import asyncio
 
 from sqlalchemy import select
-
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.core.logging import get_logger
@@ -16,7 +15,7 @@ from app.db.models import Dataset, DatasetFiles
 from app.db.session import AsyncSessionLocal
 from app.module.dataset.schema.dataset_file import DatasetFileTag
 from app.module.shared.schema import TaskStatus
-from app.module.synthesis.schema.ratio_task import FilterCondition
+from app.module.ratio.schema.ratio_task import FilterCondition
 
 logger = get_logger(__name__)
 
@@ -59,7 +58,10 @@ class RatioTaskService:
                 counts=int(item.get("counts", 0)),
                 filter_conditions=json.dumps({
                     'date_range': item.get("filter_conditions").date_range,
-                    'label': item.get("filter_conditions").label,
+                    'label': {
+                        "label":item.get("filter_conditions").label.label,
+                        "value":item.get("filter_conditions").label.value,
+                    },
                 })
             )
             logger.info(f"Relation created: {relation.id}, {relation}, {item}, {config}")
@@ -285,7 +287,7 @@ class RatioTaskService:
             try:
                 # tags could be a list of strings or list of objects with 'name'
                 tag_names = RatioTaskService.get_all_tags(tags)
-                return conditions.label in tag_names
+                return f"{conditions.label.label}@{conditions.label.value}" in tag_names
             except Exception as e:
                 logger.exception(f"Failed to get tags for {file}", e)
                 return False
diff --git a/runtime/datamate-python/app/module/shared/util/__init__.py b/runtime/datamate-python/app/module/shared/util/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/runtime/datamate-python/app/module/shared/util/model_chat.py b/runtime/datamate-python/app/module/shared/util/model_chat.py
new file mode 100644
index 0000000..7b8fcbc
--- /dev/null
+++ b/runtime/datamate-python/app/module/shared/util/model_chat.py
@@ -0,0 +1,15 @@
+from openai import OpenAI
+
+
+def call_openai_style_model(base_url, api_key, model_name, prompt, **kwargs):
+    client = OpenAI(
+        base_url=base_url,
+        api_key=api_key
+    )
+
+    response = client.chat.completions.create(
+        model=model_name,
+        messages=[{"role": "user", "content": prompt}],
+        **kwargs
+    )
+    return response.choices[0].message.content
diff --git a/runtime/datamate-python/app/module/shared/util/structured_file.py b/runtime/datamate-python/app/module/shared/util/structured_file.py
new file mode 100644
index 0000000..dc990b5
--- /dev/null
+++ b/runtime/datamate-python/app/module/shared/util/structured_file.py
@@ -0,0 +1,85 @@
+import json
+
+from enum import Enum
+from jsonschema import validate
+
+class ItemTypes(Enum):
+    QA = "QA"
+
+
+class StructuredFileItemHandler:
+    def __init__(self):
+        pass
+
+    def get_item_type(self) -> ItemTypes:
+        pass
+
+    def get_items_from_file(self, file_path: str) -> list[dict]:
+        pass
+
+    def check_file(self) -> bool:
+        pass
+
+class QAItemHandler(StructuredFileItemHandler):
+    def __init__(self):
+        self.schema_alpaca = {
+            "type": "object",
+            "properties": {
+                "instruction": {"type": "string"},
+                "input": {"type": "string"},
+                "output": {"type": "string"}
+            },
+            "required": ["instruction", "output"],
+        }
+        self.schema_alpaca_list = {
+            "type": "array",
+            "items": self.schema_alpaca,
+        }
+        super().__init__()
+
+    def get_item_type(self):
+        return ItemTypes.QA
+
+    def validate_json(self, data):
+        try:
+            validate(instance=data, schema=self.schema_alpaca)
+            return True
+        except Exception as e:
+            try:
+                validate(instance=data, schema=self.schema_alpaca_list)
+                return True
+            except Exception as e:
+                return False
+
+    def get_items_from_file(self, file_path: str) -> list[dict]:
+        file_type = file_path.split(".")[-1].upper()
+        items = []
+        if file_type == "JSON":
+            with open(file_path, "r", encoding="utf-8") as f:
+                data = json.load(f)
+                if not self.validate_json(data):
+                    return items
+                items = data
+        elif file_type == "JSONL":
+            with open(file_path, "r", encoding="utf-8") as f:
+                for line in f:
+                    data = json.loads(line)
+                    if not self.validate_json(data):
+                        continue
+                    items.append(data)
+        return items
+
+    def check_file(self) -> bool:
+        pass
+
+
+class StructuredFileHandlerFactory:
+    def __init__(self):
+        self.handlers: list[StructuredFileItemHandler] = []
+        self.handlers.append(QAItemHandler())
+
+    def get_handler(self, item_type: str) -> StructuredFileItemHandler:
+        for handler in self.handlers:
+            if handler.get_item_type().value == item_type:
+                return handler
+        raise ValueError(f"Unsupported item type: {item_type}")
diff --git a/runtime/datamate-python/poetry.lock b/runtime/datamate-python/poetry.lock
index a339c9f..c14c2fd 100644
--- a/runtime/datamate-python/poetry.lock
+++ b/runtime/datamate-python/poetry.lock
@@ -1260,6 +1260,43 @@ files = [
     {file = "jsonpointer-3.0.0.tar.gz", hash = "sha256:2b2d729f2091522d61c3b31f82e11870f60b68f43fbc705cb76bf4b832af59ef"},
 ]
 
+[[package]]
+name = "jsonschema"
+version = "4.25.1"
+description = "An implementation of JSON Schema validation for Python"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "jsonschema-4.25.1-py3-none-any.whl", hash = "sha256:3fba0169e345c7175110351d456342c364814cfcf3b964ba4587f22915230a63"},
+    {file = "jsonschema-4.25.1.tar.gz", hash = "sha256:e4a9655ce0da0c0b67a085847e00a3a51449e1157f4f75e9fb5aa545e122eb85"},
+]
+
+[package.dependencies]
+attrs = ">=22.2.0"
+jsonschema-specifications = ">=2023.03.6"
+referencing = ">=0.28.4"
+rpds-py = ">=0.7.1"
+
+[package.extras]
+format = ["fqdn", "idna", "isoduration", "jsonpointer (>1.13)", "rfc3339-validator", "rfc3987", "uri-template", "webcolors (>=1.11)"]
+format-nongpl = ["fqdn", "idna", "isoduration", "jsonpointer (>1.13)", "rfc3339-validator", "rfc3986-validator (>0.1.0)", "rfc3987-syntax (>=1.1.0)", "uri-template", "webcolors (>=24.6.0)"]
+
+[[package]]
+name = "jsonschema-specifications"
+version = "2025.9.1"
+description = "The JSON Schema meta-schemas and vocabularies, exposed as a Registry"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "jsonschema_specifications-2025.9.1-py3-none-any.whl", hash = "sha256:98802fee3a11ee76ecaca44429fda8a41bff98b00a0f2838151b113f210cc6fe"},
+    {file = "jsonschema_specifications-2025.9.1.tar.gz", hash = "sha256:b540987f239e745613c7a9176f3edb72b832a4ac465cf02712288397832b5e8d"},
+]
+
+[package.dependencies]
+referencing = ">=0.31.0"
+
 [[package]]
 name = "langchain"
 version = "1.1.0"
@@ -2864,6 +2901,23 @@ files = [
 [package.extras]
 all = ["numpy"]
 
+[[package]]
+name = "referencing"
+version = "0.37.0"
+description = "JSON Referencing + Python"
+optional = false
+python-versions = ">=3.10"
+groups = ["main"]
+files = [
+    {file = "referencing-0.37.0-py3-none-any.whl", hash = "sha256:381329a9f99628c9069361716891d34ad94af76e461dcb0335825aecc7692231"},
+    {file = "referencing-0.37.0.tar.gz", hash = "sha256:44aefc3142c5b842538163acb373e24cce6632bd54bdb01b21ad5863489f50d8"},
+]
+
+[package.dependencies]
+attrs = ">=22.2.0"
+rpds-py = ">=0.7.0"
+typing-extensions = {version = ">=4.4.0", markers = "python_version < \"3.13\""}
+
 [[package]]
 name = "regex"
 version = "2025.11.3"
@@ -3026,6 +3080,131 @@ files = [
 [package.dependencies]
 requests = ">=2.0.1,<3.0.0"
 
+[[package]]
+name = "rpds-py"
+version = "0.30.0"
+description = "Python bindings to Rust's persistent data structures (rpds)"
+optional = false
+python-versions = ">=3.10"
+groups = ["main"]
+files = [
+    {file = "rpds_py-0.30.0-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:679ae98e00c0e8d68a7fda324e16b90fd5260945b45d3b824c892cec9eea3288"},
+    {file = "rpds_py-0.30.0-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:4cc2206b76b4f576934f0ed374b10d7ca5f457858b157ca52064bdfc26b9fc00"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:389a2d49eded1896c3d48b0136ead37c48e221b391c052fba3f4055c367f60a6"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:32c8528634e1bf7121f3de08fa85b138f4e0dc47657866630611b03967f041d7"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f207f69853edd6f6700b86efb84999651baf3789e78a466431df1331608e5324"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:67b02ec25ba7a9e8fa74c63b6ca44cf5707f2fbfadae3ee8e7494297d56aa9df"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0c0e95f6819a19965ff420f65578bacb0b00f251fefe2c8b23347c37174271f3"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_31_riscv64.whl", hash = "sha256:a452763cc5198f2f98898eb98f7569649fe5da666c2dc6b5ddb10fde5a574221"},
+    {file = "rpds_py-0.30.0-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:e0b65193a413ccc930671c55153a03ee57cecb49e6227204b04fae512eb657a7"},
+    {file = "rpds_py-0.30.0-cp310-cp310-musllinux_1_2_aarch64.whl", hash = "sha256:858738e9c32147f78b3ac24dc0edb6610000e56dc0f700fd5f651d0a0f0eb9ff"},
+    {file = "rpds_py-0.30.0-cp310-cp310-musllinux_1_2_i686.whl", hash = "sha256:da279aa314f00acbb803da1e76fa18666778e8a8f83484fba94526da5de2cba7"},
+    {file = "rpds_py-0.30.0-cp310-cp310-musllinux_1_2_x86_64.whl", hash = "sha256:7c64d38fb49b6cdeda16ab49e35fe0da2e1e9b34bc38bd78386530f218b37139"},
+    {file = "rpds_py-0.30.0-cp310-cp310-win32.whl", hash = "sha256:6de2a32a1665b93233cde140ff8b3467bdb9e2af2b91079f0333a0974d12d464"},
+    {file = "rpds_py-0.30.0-cp310-cp310-win_amd64.whl", hash = "sha256:1726859cd0de969f88dc8673bdd954185b9104e05806be64bcd87badbe313169"},
+    {file = "rpds_py-0.30.0-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:a2bffea6a4ca9f01b3f8e548302470306689684e61602aa3d141e34da06cf425"},
+    {file = "rpds_py-0.30.0-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:dc4f992dfe1e2bc3ebc7444f6c7051b4bc13cd8e33e43511e8ffd13bf407010d"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:422c3cb9856d80b09d30d2eb255d0754b23e090034e1deb4083f8004bd0761e4"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:07ae8a593e1c3c6b82ca3292efbe73c30b61332fd612e05abee07c79359f292f"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:12f90dd7557b6bd57f40abe7747e81e0c0b119bef015ea7726e69fe550e394a4"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:99b47d6ad9a6da00bec6aabe5a6279ecd3c06a329d4aa4771034a21e335c3a97"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:33f559f3104504506a44bb666b93a33f5d33133765b0c216a5bf2f1e1503af89"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_31_riscv64.whl", hash = "sha256:946fe926af6e44f3697abbc305ea168c2c31d3e3ef1058cf68f379bf0335a78d"},
+    {file = "rpds_py-0.30.0-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:495aeca4b93d465efde585977365187149e75383ad2684f81519f504f5c13038"},
+    {file = "rpds_py-0.30.0-cp311-cp311-musllinux_1_2_aarch64.whl", hash = "sha256:d9a0ca5da0386dee0655b4ccdf46119df60e0f10da268d04fe7cc87886872ba7"},
+    {file = "rpds_py-0.30.0-cp311-cp311-musllinux_1_2_i686.whl", hash = "sha256:8d6d1cc13664ec13c1b84241204ff3b12f9bb82464b8ad6e7a5d3486975c2eed"},
+    {file = "rpds_py-0.30.0-cp311-cp311-musllinux_1_2_x86_64.whl", hash = "sha256:3896fa1be39912cf0757753826bc8bdc8ca331a28a7c4ae46b7a21280b06bb85"},
+    {file = "rpds_py-0.30.0-cp311-cp311-win32.whl", hash = "sha256:55f66022632205940f1827effeff17c4fa7ae1953d2b74a8581baaefb7d16f8c"},
+    {file = "rpds_py-0.30.0-cp311-cp311-win_amd64.whl", hash = "sha256:a51033ff701fca756439d641c0ad09a41d9242fa69121c7d8769604a0a629825"},
+    {file = "rpds_py-0.30.0-cp311-cp311-win_arm64.whl", hash = "sha256:47b0ef6231c58f506ef0b74d44e330405caa8428e770fec25329ed2cb971a229"},
+    {file = "rpds_py-0.30.0-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:a161f20d9a43006833cd7068375a94d035714d73a172b681d8881820600abfad"},
+    {file = "rpds_py-0.30.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:6abc8880d9d036ecaafe709079969f56e876fcf107f7a8e9920ba6d5a3878d05"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ca28829ae5f5d569bb62a79512c842a03a12576375d5ece7d2cadf8abe96ec28"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:a1010ed9524c73b94d15919ca4d41d8780980e1765babf85f9a2f90d247153dd"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f8d1736cfb49381ba528cd5baa46f82fdc65c06e843dab24dd70b63d09121b3f"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:d948b135c4693daff7bc2dcfc4ec57237a29bd37e60c2fabf5aff2bbacf3e2f1"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:47f236970bccb2233267d89173d3ad2703cd36a0e2a6e92d0560d333871a3d23"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_31_riscv64.whl", hash = "sha256:2e6ecb5a5bcacf59c3f912155044479af1d0b6681280048b338b28e364aca1f6"},
+    {file = "rpds_py-0.30.0-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:a8fa71a2e078c527c3e9dc9fc5a98c9db40bcc8a92b4e8858e36d329f8684b51"},
+    {file = "rpds_py-0.30.0-cp312-cp312-musllinux_1_2_aarch64.whl", hash = "sha256:73c67f2db7bc334e518d097c6d1e6fed021bbc9b7d678d6cc433478365d1d5f5"},
+    {file = "rpds_py-0.30.0-cp312-cp312-musllinux_1_2_i686.whl", hash = "sha256:5ba103fb455be00f3b1c2076c9d4264bfcb037c976167a6047ed82f23153f02e"},
+    {file = "rpds_py-0.30.0-cp312-cp312-musllinux_1_2_x86_64.whl", hash = "sha256:7cee9c752c0364588353e627da8a7e808a66873672bcb5f52890c33fd965b394"},
+    {file = "rpds_py-0.30.0-cp312-cp312-win32.whl", hash = "sha256:1ab5b83dbcf55acc8b08fc62b796ef672c457b17dbd7820a11d6c52c06839bdf"},
+    {file = "rpds_py-0.30.0-cp312-cp312-win_amd64.whl", hash = "sha256:a090322ca841abd453d43456ac34db46e8b05fd9b3b4ac0c78bcde8b089f959b"},
+    {file = "rpds_py-0.30.0-cp312-cp312-win_arm64.whl", hash = "sha256:669b1805bd639dd2989b281be2cfd951c6121b65e729d9b843e9639ef1fd555e"},
+    {file = "rpds_py-0.30.0-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:f83424d738204d9770830d35290ff3273fbb02b41f919870479fab14b9d303b2"},
+    {file = "rpds_py-0.30.0-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:e7536cd91353c5273434b4e003cbda89034d67e7710eab8761fd918ec6c69cf8"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2771c6c15973347f50fece41fc447c054b7ac2ae0502388ce3b6738cd366e3d4"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:0a59119fc6e3f460315fe9d08149f8102aa322299deaa5cab5b40092345c2136"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:76fec018282b4ead0364022e3c54b60bf368b9d926877957a8624b58419169b7"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:692bef75a5525db97318e8cd061542b5a79812d711ea03dbc1f6f8dbb0c5f0d2"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9027da1ce107104c50c81383cae773ef5c24d296dd11c99e2629dbd7967a20c6"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_31_riscv64.whl", hash = "sha256:9cf69cdda1f5968a30a359aba2f7f9aa648a9ce4b580d6826437f2b291cfc86e"},
+    {file = "rpds_py-0.30.0-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:a4796a717bf12b9da9d3ad002519a86063dcac8988b030e405704ef7d74d2d9d"},
+    {file = "rpds_py-0.30.0-cp313-cp313-musllinux_1_2_aarch64.whl", hash = "sha256:5d4c2aa7c50ad4728a094ebd5eb46c452e9cb7edbfdb18f9e1221f597a73e1e7"},
+    {file = "rpds_py-0.30.0-cp313-cp313-musllinux_1_2_i686.whl", hash = "sha256:ba81a9203d07805435eb06f536d95a266c21e5b2dfbf6517748ca40c98d19e31"},
+    {file = "rpds_py-0.30.0-cp313-cp313-musllinux_1_2_x86_64.whl", hash = "sha256:945dccface01af02675628334f7cf49c2af4c1c904748efc5cf7bbdf0b579f95"},
+    {file = "rpds_py-0.30.0-cp313-cp313-win32.whl", hash = "sha256:b40fb160a2db369a194cb27943582b38f79fc4887291417685f3ad693c5a1d5d"},
+    {file = "rpds_py-0.30.0-cp313-cp313-win_amd64.whl", hash = "sha256:806f36b1b605e2d6a72716f321f20036b9489d29c51c91f4dd29a3e3afb73b15"},
+    {file = "rpds_py-0.30.0-cp313-cp313-win_arm64.whl", hash = "sha256:d96c2086587c7c30d44f31f42eae4eac89b60dabbac18c7669be3700f13c3ce1"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-macosx_10_12_x86_64.whl", hash = "sha256:eb0b93f2e5c2189ee831ee43f156ed34e2a89a78a66b98cadad955972548be5a"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-macosx_11_0_arm64.whl", hash = "sha256:922e10f31f303c7c920da8981051ff6d8c1a56207dbdf330d9047f6d30b70e5e"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:cdc62c8286ba9bf7f47befdcea13ea0e26bf294bda99758fd90535cbaf408000"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:47f9a91efc418b54fb8190a6b4aa7813a23fb79c51f4bb84e418f5476c38b8db"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1f3587eb9b17f3789ad50824084fa6f81921bbf9a795826570bda82cb3ed91f2"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:39c02563fc592411c2c61d26b6c5fe1e51eaa44a75aa2c8735ca88b0d9599daa"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:51a1234d8febafdfd33a42d97da7a43f5dcb120c1060e352a3fbc0c6d36e2083"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_31_riscv64.whl", hash = "sha256:eb2c4071ab598733724c08221091e8d80e89064cd472819285a9ab0f24bcedb9"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:6bdfdb946967d816e6adf9a3d8201bfad269c67efe6cefd7093ef959683c8de0"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-musllinux_1_2_aarch64.whl", hash = "sha256:c77afbd5f5250bf27bf516c7c4a016813eb2d3e116139aed0096940c5982da94"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-musllinux_1_2_i686.whl", hash = "sha256:61046904275472a76c8c90c9ccee9013d70a6d0f73eecefd38c1ae7c39045a08"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-musllinux_1_2_x86_64.whl", hash = "sha256:4c5f36a861bc4b7da6516dbdf302c55313afa09b81931e8280361a4f6c9a2d27"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-win32.whl", hash = "sha256:3d4a69de7a3e50ffc214ae16d79d8fbb0922972da0356dcf4d0fdca2878559c6"},
+    {file = "rpds_py-0.30.0-cp313-cp313t-win_amd64.whl", hash = "sha256:f14fc5df50a716f7ece6a80b6c78bb35ea2ca47c499e422aa4463455dd96d56d"},
+    {file = "rpds_py-0.30.0-cp314-cp314-macosx_10_12_x86_64.whl", hash = "sha256:68f19c879420aa08f61203801423f6cd5ac5f0ac4ac82a2368a9fcd6a9a075e0"},
+    {file = "rpds_py-0.30.0-cp314-cp314-macosx_11_0_arm64.whl", hash = "sha256:ec7c4490c672c1a0389d319b3a9cfcd098dcdc4783991553c332a15acf7249be"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f251c812357a3fed308d684a5079ddfb9d933860fc6de89f2b7ab00da481e65f"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:ac98b175585ecf4c0348fd7b29c3864bda53b805c773cbf7bfdaffc8070c976f"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3e62880792319dbeb7eb866547f2e35973289e7d5696c6e295476448f5b63c87"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4e7fc54e0900ab35d041b0601431b0a0eb495f0851a0639b6ef90f7741b39a18"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:47e77dc9822d3ad616c3d5759ea5631a75e5809d5a28707744ef79d7a1bcfcad"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_31_riscv64.whl", hash = "sha256:b4dc1a6ff022ff85ecafef7979a2c6eb423430e05f1165d6688234e62ba99a07"},
+    {file = "rpds_py-0.30.0-cp314-cp314-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:4559c972db3a360808309e06a74628b95eaccbf961c335c8fe0d590cf587456f"},
+    {file = "rpds_py-0.30.0-cp314-cp314-musllinux_1_2_aarch64.whl", hash = "sha256:0ed177ed9bded28f8deb6ab40c183cd1192aa0de40c12f38be4d59cd33cb5c65"},
+    {file = "rpds_py-0.30.0-cp314-cp314-musllinux_1_2_i686.whl", hash = "sha256:ad1fa8db769b76ea911cb4e10f049d80bf518c104f15b3edb2371cc65375c46f"},
+    {file = "rpds_py-0.30.0-cp314-cp314-musllinux_1_2_x86_64.whl", hash = "sha256:46e83c697b1f1c72b50e5ee5adb4353eef7406fb3f2043d64c33f20ad1c2fc53"},
+    {file = "rpds_py-0.30.0-cp314-cp314-win32.whl", hash = "sha256:ee454b2a007d57363c2dfd5b6ca4a5d7e2c518938f8ed3b706e37e5d470801ed"},
+    {file = "rpds_py-0.30.0-cp314-cp314-win_amd64.whl", hash = "sha256:95f0802447ac2d10bcc69f6dc28fe95fdf17940367b21d34e34c737870758950"},
+    {file = "rpds_py-0.30.0-cp314-cp314-win_arm64.whl", hash = "sha256:613aa4771c99f03346e54c3f038e4cc574ac09a3ddfb0e8878487335e96dead6"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-macosx_10_12_x86_64.whl", hash = "sha256:7e6ecfcb62edfd632e56983964e6884851786443739dbfe3582947e87274f7cb"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-macosx_11_0_arm64.whl", hash = "sha256:a1d0bc22a7cdc173fedebb73ef81e07faef93692b8c1ad3733b67e31e1b6e1b8"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0d08f00679177226c4cb8c5265012eea897c8ca3b93f429e546600c971bcbae7"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:5965af57d5848192c13534f90f9dd16464f3c37aaf166cc1da1cae1fd5a34898"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9a4e86e34e9ab6b667c27f3211ca48f73dba7cd3d90f8d5b11be56e5dbc3fb4e"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:e5d3e6b26f2c785d65cc25ef1e5267ccbe1b069c5c21b8cc724efee290554419"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:626a7433c34566535b6e56a1b39a7b17ba961e97ce3b80ec62e6f1312c025551"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_31_riscv64.whl", hash = "sha256:acd7eb3f4471577b9b5a41baf02a978e8bdeb08b4b355273994f8b87032000a8"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:fe5fa731a1fa8a0a56b0977413f8cacac1768dad38d16b3a296712709476fbd5"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-musllinux_1_2_aarch64.whl", hash = "sha256:74a3243a411126362712ee1524dfc90c650a503502f135d54d1b352bd01f2404"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-musllinux_1_2_i686.whl", hash = "sha256:3e8eeb0544f2eb0d2581774be4c3410356eba189529a6b3e36bbbf9696175856"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-musllinux_1_2_x86_64.whl", hash = "sha256:dbd936cde57abfee19ab3213cf9c26be06d60750e60a8e4dd85d1ab12c8b1f40"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-win32.whl", hash = "sha256:dc824125c72246d924f7f796b4f63c1e9dc810c7d9e2355864b3c3a73d59ade0"},
+    {file = "rpds_py-0.30.0-cp314-cp314t-win_amd64.whl", hash = "sha256:27f4b0e92de5bfbc6f86e43959e6edd1425c33b5e69aab0984a72047f2bcf1e3"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-macosx_10_12_x86_64.whl", hash = "sha256:c2262bdba0ad4fc6fb5545660673925c2d2a5d9e2e0fb603aad545427be0fc58"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-macosx_11_0_arm64.whl", hash = "sha256:ee6af14263f25eedc3bb918a3c04245106a42dfd4f5c2285ea6f997b1fc3f89a"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3adbb8179ce342d235c31ab8ec511e66c73faa27a47e076ccc92421add53e2bb"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:250fa00e9543ac9b97ac258bd37367ff5256666122c2d0f2bc97577c60a1818c"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9854cf4f488b3d57b9aaeb105f06d78e5529d3145b1e4a41750167e8c213c6d3"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:993914b8e560023bc0a8bf742c5f303551992dcb85e247b1e5c7f4a7d145bda5"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:58edca431fb9b29950807e301826586e5bbf24163677732429770a697ffe6738"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_31_riscv64.whl", hash = "sha256:dea5b552272a944763b34394d04577cf0f9bd013207bc32323b5a89a53cf9c2f"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:ba3af48635eb83d03f6c9735dfb21785303e73d22ad03d489e88adae6eab8877"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-musllinux_1_2_aarch64.whl", hash = "sha256:dff13836529b921e22f15cb099751209a60009731a68519630a24d61f0b1b30a"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-musllinux_1_2_i686.whl", hash = "sha256:1b151685b23929ab7beec71080a8889d4d6d9fa9a983d213f07121205d48e2c4"},
+    {file = "rpds_py-0.30.0-pp311-pypy311_pp73-musllinux_1_2_x86_64.whl", hash = "sha256:ac37f9f516c51e5753f27dfdef11a88330f04de2d564be3991384b2f3535d02e"},
+    {file = "rpds_py-0.30.0.tar.gz", hash = "sha256:dd8ff7cf90014af0c0f787eea34794ebf6415242ee1d6fa91eaba725cc441e84"},
+]
+
 [[package]]
 name = "six"
 version = "1.17.0"
@@ -4261,4 +4440,4 @@ cffi = ["cffi (>=1.17,<2.0) ; platform_python_implementation != \"PyPy\" and pyt
 [metadata]
 lock-version = "2.1"
 python-versions = ">=3.12,<4.0.0"
-content-hash = "9c64ec78daa702d8344683faba327cb686b7fa79cf9bb12006c3dcf55dae789d"
+content-hash = "26fa6096f5efdd91a8d5b17a2f5efd68cd5f26e4dcdd516a39b77708c36b45e8"
diff --git a/runtime/datamate-python/pyproject.toml b/runtime/datamate-python/pyproject.toml
index 8ce731e..73c4455 100644
--- a/runtime/datamate-python/pyproject.toml
+++ b/runtime/datamate-python/pyproject.toml
@@ -27,6 +27,8 @@ dependencies = [
     "langchain (>=1.1.0,<2.0.0)",
     "langchain-community (>=0.4.1,<0.5.0)",
     "langchain-openai (>=1.1.0,<2.0.0)",
+    "openai (>=2.8.1,<3.0.0)",
+    "jsonschema (>=4.25.1,<5.0.0)",
 ]
 
 
diff --git a/scripts/db/data-evaluation-init.sql b/scripts/db/data-evaluation-init.sql
new file mode 100644
index 0000000..a2a1e72
--- /dev/null
+++ b/scripts/db/data-evaluation-init.sql
@@ -0,0 +1,53 @@
+-- 使用现有的datamate数据库
+USE datamate;
+
+
+-- 数据集表（支持医学影像、文本、问答等多种类型）
+CREATE TABLE IF NOT EXISTS t_de_eval_task (
+    id VARCHAR(36) PRIMARY KEY COMMENT 'UUID',
+    name VARCHAR(255) NOT NULL COMMENT '评估任务名称',
+    description TEXT COMMENT '评估任务描述',
+    task_type VARCHAR(50) NOT NULL COMMENT '评估任务类型：QA',
+    source_type VARCHAR(36) COMMENT '待评估对象类型：DATASET/SYNTHESIS',
+    source_id VARCHAR(36) COMMENT '待评估对象ID',
+    source_name VARCHAR(255) COMMENT '待评估对象名称',
+    status VARCHAR(50) DEFAULT 'PENDING' COMMENT '状态：PENDING/RUNNING/COMPLETED/STOPPED/FAILED',
+    eval_method VARCHAR(50) DEFAULT 'AUTO' COMMENT '状态：AUTO/MANUAL',
+    eval_process DOUBLE PRECISION NOT NULL DEFAULT 0 COMMENT '评估进度',
+    eval_prompt TEXT COMMENT '评估提示词',
+    eval_config TEXT COMMENT '评估配置',
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
+    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
+    created_by VARCHAR(255) COMMENT '创建者',
+    updated_by VARCHAR(255) COMMENT '更新者',
+    INDEX idx_dm_status (status),
+    INDEX idx_dm_created_at (created_at)
+) COMMENT='评估任务表（UUID 主键）';
+
+CREATE TABLE IF NOT EXISTS t_de_eval_file (
+    id VARCHAR(36) PRIMARY KEY COMMENT 'UUID',
+    task_id VARCHAR(36) NOT NULL COMMENT '评估任务ID',
+    file_id VARCHAR(36) COMMENT '文件ID',
+    file_name VARCHAR(255) COMMENT '文件名',
+    total_count INT DEFAULT 0 COMMENT '总数',
+    evaluated_count INT DEFAULT 0 COMMENT '已评估数',
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
+    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
+    created_by VARCHAR(255) COMMENT '创建者',
+    updated_by VARCHAR(255) COMMENT '更新者'
+) COMMENT='评估文件表（UUID 主键）';
+
+CREATE TABLE IF NOT EXISTS t_de_eval_item (
+    id VARCHAR(36) PRIMARY KEY COMMENT 'UUID',
+    task_id VARCHAR(36) NOT NULL COMMENT '评估任务ID',
+    file_id VARCHAR(36) COMMENT '文件ID',
+    item_id VARCHAR(36) NOT NULL COMMENT '评估条目ID',
+    eval_content TEXT COMMENT '评估内容',
+    eval_score DOUBLE PRECISION NOT NULL DEFAULT 0 COMMENT '评估分数',
+    eval_result TEXT COMMENT '评估结果',
+    status VARCHAR(50) DEFAULT 'PENDING' COMMENT '状态：PENDING/EVALUATED',
+    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP COMMENT '创建时间',
+    updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP COMMENT '更新时间',
+    created_by VARCHAR(255) COMMENT '创建者',
+    updated_by VARCHAR(255) COMMENT '更新者'
+) COMMENT='评估条目表（UUID 主键）';
diff --git a/scripts/db/setting-management-init.sql b/scripts/db/setting-management-init.sql
index cf440c0..785df98 100644
--- a/scripts/db/setting-management-init.sql
+++ b/scripts/db/setting-management-init.sql
@@ -47,4 +47,5 @@ values ('1', 'sys.knowledge.base.count', '200', 'number', '10,200,500', '知识
        ('4', 'BRAVE_SEARCH_API_KEY', 'api-xxx', 'string', '', 'deer-flow使用的搜索引擎所需的apiKey', 1, 1, 1, 'system',
         'system'),
        ('5', 'JINA_API_KEY', '', 'string', '', 'deer-flow使用的JINA搜索引擎所需的apiKey', 1, 1, 1, 'system', 'system'),
-       ('6', 'test_bool', 'true', 'boolean', '', '测试布尔值', 1, 1, 1, 'system', 'system');
+       ('6', 'sys.management.dataset.pvc.name', 'dataset-pvc', 'string', '', '数据集所在pvc名称', 1, 0, 1, 'system', 'system'),
+       ('7', 'test_bool', 'true', 'boolean', '', '测试布尔值', 1, 1, 1, 'system', 'system');