docs/workspace/compare-evaluation-analysis/structured-compare-calibration/latest/synthetic-legal-flat-not-unclear/request.md
{
"type": "compare",
"evaluationModelKey": "deepseek",
"mode": {
"functionMode": "basic",
"subMode": "system"
},
"focus": {
"content": "当两个版本在核心结论、风险点和动作建议上等价时,应更倾向于 flat,而不是把风格差异误判成信息不足。",
"source": "system",
"priority": "highest"
},
"target": {
"workspacePrompt": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n用更简洁、偏业务同学可读的中文表达。\n不要添加解释。",
"referencePrompt": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n保持客观、精炼。\n不要添加解释。"
},
"testCases": [
{
"id": "tc-1",
"input": {
"kind": "text",
"label": "合同片段",
"content": "合作协议约定平台可单方修改结算周期,并在未通知的情况下暂停服务;违约责任仅约束供应商,不约束平台。"
}
}
],
"snapshots": [
{
"id": "a",
"label": "A",
"testCaseId": "tc-1",
"promptRef": {
"kind": "workspace",
"label": "Workspace"
},
"promptText": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n用更简洁、偏业务同学可读的中文表达。\n不要添加解释。",
"output": "{\"risk_level\":\"high\",\"core_risks\":[\"平台可单方改结算周期\",\"平台可未通知暂停服务\",\"违约责任明显失衡\"],\"recommended_action\":\"要求补充通知义务、限制单方变更范围,并补齐平台违约责任。\"}",
"modelKey": "custom",
"versionLabel": "workspace"
},
{
"id": "b",
"label": "B",
"testCaseId": "tc-1",
"promptRef": {
"kind": "version",
"version": 6,
"label": "Previous"
},
"promptText": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n保持客观、精炼。\n不要添加解释。",
"output": "{\"risk_level\":\"high\",\"core_risks\":[\"平台可单方调整结算周期\",\"平台可在未通知情况下暂停服务\",\"违约责任分配失衡\"],\"recommended_action\":\"建议增加通知义务、限制单方修改权限,并要求平台承担对等违约责任。\"}",
"modelKey": "custom",
"versionLabel": "v6"
},
{
"id": "c",
"label": "C",
"testCaseId": "tc-1",
"promptRef": {
"kind": "workspace",
"label": "Teacher Workspace"
},
"promptText": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n用更简洁、偏业务同学可读的中文表达。\n不要添加解释。",
"output": "{\"risk_level\":\"high\",\"core_risks\":[\"平台可单方改结算周期\",\"平台可未通知暂停服务\",\"违约责任缺乏对等性\"],\"recommended_action\":\"要求把通知义务、变更边界和平台违约责任补齐后再推进。\"}",
"modelKey": "deepseek",
"versionLabel": "teacher-workspace"
},
{
"id": "d",
"label": "D",
"testCaseId": "tc-1",
"promptRef": {
"kind": "version",
"version": 6,
"label": "Teacher Previous"
},
"promptText": "你是法务风险摘要助手。\n输出一个 JSON 对象,字段为 risk_level, core_risks, recommended_action。\n保持客观、精炼。\n不要添加解释。",
"output": "{\"risk_level\":\"high\",\"core_risks\":[\"平台可单方调整结算周期\",\"平台可在未通知情况下暂停服务\",\"违约责任缺乏对等性\"],\"recommended_action\":\"建议补充通知义务、限制单方变更,并增加平台违约责任。\"}",
"modelKey": "deepseek",
"versionLabel": "teacher-v6"
}
],
"compareHints": {
"mode": "structured",
"snapshotRoles": {
"a": "target",
"b": "baseline",
"c": "reference",
"d": "referenceBaseline"
},
"hasSharedTestCases": true,
"hasSamePromptSnapshots": true,
"hasCrossModelComparison": true
}
}