WpyQwq
643e22ecb9
- 引入 MemoryRouterXL 与 v5/v6 流式多线程训练/编码管线 - 修复 prepare_memory_router_dataset 候选池重建缺陷（mega 家族 3568x 加速，输出逐字节相同） - 修复 v5 被破坏的拒答与多跳标签（train 未知样本 319 -> 16319，multi_hop 平均正例 1.00 -> 2.00） - 同存储预算下 V2-128 v6 逐轴 22/22 通过：Top-1 41.12% -> 94.62%，未知拒答 0.00% -> 100.00% - 记录三条被实测推翻的显然优化（logits_to_keep=1 反而慢 55%、XL 容量未带来收益） - 记忆手术跨架构可移植性 14/14，读写关闭时与原生模型逐位相同
838 lines
34 KiB
JSON
838 lines
34 KiB
JSON
{
|
||
"package": "qwen3_5_4b_natural_memory_v2_1",
|
||
"label": "NM2.1+gate2",
|
||
"tag": "nm2_1_gate2",
|
||
"built_at": "2026-09-12T21:45:55",
|
||
"per_category": 10,
|
||
"overlap_cases": 48,
|
||
"results": {
|
||
"C_same_shape_and_unknown": [
|
||
"",
|
||
"| blend | 用例 | 可回答 | 回答正确率 | 答成别的属性 | 未知泄漏率 | 平均选中记录 | 活跃记录 min/max |",
|
||
"|---|---:|---:|---:|---:|---:|---:|---|",
|
||
"| 0.00 | 48 | 40 | 65.00% | 35.00% | 0.00% | 5.2083 | 23/24 |",
|
||
"",
|
||
"## 检索最高分分布（原始分，非百分比；用于判断阈值能否分开『未知』与『可回答』）",
|
||
"",
|
||
"| 组 | n | min | p10 | p50 | p90 | max |",
|
||
"|---|---:|---:|---:|---:|---:|---:|",
|
||
"| 0.00 · 可回答 | 40 | 0.7 | 0.77 | 2.85 | 8.52 | 8.52 |",
|
||
"| 0.00 · 不可回答 | 8 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |",
|
||
"",
|
||
{
|
||
"eval_file": "data/zero_overlap/eval.jsonl",
|
||
"episodes": 48,
|
||
"candidate_count": 24,
|
||
"router": "checkpoints\\router_replay_v7_v2_128\\memory_router_v2.pt",
|
||
"results": {
|
||
"0.00": {
|
||
"router": "blend=0.00",
|
||
"cases": 48,
|
||
"answerable_cases": 40,
|
||
"unknown_cases": 8,
|
||
"accuracy_pct": 65.0,
|
||
"wrong_attribute_pct": 35.0,
|
||
"unknown_leak_pct": 0.0,
|
||
"mean_records_selected": 5.208333333333333,
|
||
"records_active_min": 23,
|
||
"records_active_max": 24,
|
||
"stop_reasons": {
|
||
"evidence_found": 40,
|
||
"attribute_not_covered": 8
|
||
},
|
||
"top_score_answerable": {
|
||
"n": 40,
|
||
"min": 0.7,
|
||
"p10": 0.77,
|
||
"p50": 2.85,
|
||
"p90": 8.52,
|
||
"max": 8.52
|
||
},
|
||
"top_score_unknown": {
|
||
"n": 8,
|
||
"min": 0.0,
|
||
"p10": 0.0,
|
||
"p50": 0.0,
|
||
"p90": 0.0,
|
||
"max": 0.0
|
||
},
|
||
"seconds": 346.3
|
||
}
|
||
},
|
||
"rows": {
|
||
"0.00": [
|
||
{
|
||
"id": "zero-overlap-000259",
|
||
"attribute": "出生城市",
|
||
"answerable": true,
|
||
"expected": "VAL-M9VSJAEX",
|
||
"reply": "您的出生城市是 VAL-M9VSJAEX。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.564332926884773,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000219",
|
||
"attribute": "咖啡口味",
|
||
"answerable": true,
|
||
"expected": "VAL-M3W7MRJR",
|
||
"reply": "关于您早上那杯咖啡的豆子偏好，当前记忆中没有相关记录。已知信息仅包含：user；午餐偏好；VAL-Y4Z5H2FG。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.515033109300843,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000193",
|
||
"attribute": "午餐偏好",
|
||
"answerable": true,
|
||
"expected": "VAL-HG84H5TD",
|
||
"reply": "您的午餐偏好是 user；午餐偏好；VAL-HG84H5TD。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.6998433045297862,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000080",
|
||
"attribute": "入职年份",
|
||
"answerable": true,
|
||
"expected": "VAL-GFT4HRX2",
|
||
"reply": "关于您何时开始在这家公司上班，当前记忆中没有相关记录。已知信息：user；办公城市；VAL-TCNMN9HW。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.51429933919514,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000285",
|
||
"attribute": "起床时间",
|
||
"answerable": true,
|
||
"expected": "VAL-4K3RQQKF",
|
||
"reply": "我的闹钟设定时间是 VAL-4K3RQQKF。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 3.359343914853775,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000246",
|
||
"attribute": "常住城市",
|
||
"answerable": false,
|
||
"expected": "VAL-6BWJYDE4",
|
||
"reply": "作为一个人工智能助手，我并没有“小时候”的概念，也无法知道您具体待过哪里。\n\n不过，如果您是想询问**您自己**小时候待的地方，或者想聊聊关于某个特定地点的童年回忆，我很乐意",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000204",
|
||
"attribute": "阅读工具",
|
||
"answerable": true,
|
||
"expected": "VAL-ZRLSPU5V",
|
||
"reply": "您的阅读工具是 VAL-ZRLSPU5V。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.9886870104974497,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000078",
|
||
"attribute": "常住城市",
|
||
"answerable": true,
|
||
"expected": "VAL-TQBY65U8",
|
||
"reply": "我的常住城市是 VAL-TQBY65U8。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.2650369691848753,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000104",
|
||
"attribute": "午餐偏好",
|
||
"answerable": true,
|
||
"expected": "VAL-SRFC842D",
|
||
"reply": "您的午餐偏好是 VAL-SRFC842D。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.7423390823602677,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000281",
|
||
"attribute": "咖啡口味",
|
||
"answerable": true,
|
||
"expected": "VAL-3GP4STZX",
|
||
"reply": "关于您早上饮用的咖啡偏好哪种豆子，当前记忆中没有相关记录。已知信息仅包含：user；午餐偏好；VAL-XNN7T8EK。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.515033109300843,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000106",
|
||
"attribute": "咖啡口味",
|
||
"answerable": false,
|
||
"expected": "VAL-837DYYWV",
|
||
"reply": "作为一个人工智能助手，我**无法直接访问您的设备、查看您的闹钟设置或获取您的实时时间信息**。\n\n要找到您的闹钟设定时间，您可以尝试以下方法：\n\n1. **查看手机/电脑屏幕",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000091",
|
||
"attribute": "运动习惯",
|
||
"answerable": true,
|
||
"expected": "VAL-92MQHEDL",
|
||
"reply": "您的运动习惯是 VAL-92MQHEDL。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.0058205354213714,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000088",
|
||
"attribute": "设备型号",
|
||
"answerable": true,
|
||
"expected": "VAL-JY6S7EYF",
|
||
"reply": "关于您使用的机器档次，当前长期记忆中未包含此信息，无法回答。已知信息仅涉及：user；常用编辑器；VAL-3KPKRM4J。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.516791512356487,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000244",
|
||
"attribute": "午餐偏好",
|
||
"answerable": false,
|
||
"expected": "VAL-2VLHC26V",
|
||
"reply": "您好！作为一个人工智能助手，我并没有直接服务于特定某一家公司或机构。我的设计目标是**为所有用户**提供通用、安全、有帮助的信息和服务。\n\n不过，如果您是在某个具体平台（如企业微信",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000282",
|
||
"attribute": "咖啡口味",
|
||
"answerable": true,
|
||
"expected": "VAL-Q2AQQMUP",
|
||
"reply": "关于您早上那杯咖啡的豆子偏好，当前记忆中没有相关记录。已知信息仅包含：user；午餐偏好；VAL-5ATEHVSM。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.515033109300843,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000056",
|
||
"attribute": "主管姓名",
|
||
"answerable": true,
|
||
"expected": "VAL-VDL8H7Y9",
|
||
"reply": "您的主管姓名是 user；主管姓名；VAL-VDL8H7Y9。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.9513123025616852,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000184",
|
||
"attribute": "项目代号",
|
||
"answerable": true,
|
||
"expected": "VAL-3H5TQEC4",
|
||
"reply": "您的团队名称是 VAL-Z98JC7BG。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 3.191310502301591,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000198",
|
||
"attribute": "手机尾号",
|
||
"answerable": true,
|
||
"expected": "VAL-7T3XVMVH",
|
||
"reply": "您随身那台设备的末端数字是 4N，该设备的完整型号为 VAL-9MWPDS4N。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.513663491707547,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000258",
|
||
"attribute": "阅读工具",
|
||
"answerable": true,
|
||
"expected": "VAL-SHRY3GNG",
|
||
"reply": "我翻书时依赖的阅读工具是 VAL-SHRY3GNG。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.9932317454522837,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000030",
|
||
"attribute": "课程名称",
|
||
"answerable": true,
|
||
"expected": "VAL-BQY3K5EV",
|
||
"reply": "您的课程名称是 VAL-BQY3K5EV。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.7088197509944439,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000196",
|
||
"attribute": "手机尾号",
|
||
"answerable": true,
|
||
"expected": "VAL-ZHN2NMZV",
|
||
"reply": "您随身那台设备的末端数字是 W9，该设备的完整型号为 VAL-5VBJ22W9。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.513663491707547,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000268",
|
||
"attribute": "宿舍楼号",
|
||
"answerable": true,
|
||
"expected": "VAL-QVYMWTLK",
|
||
"reply": "关于您睡觉的地儿是第几排，当前长期记忆中未包含此信息。已知信息如下：user；宿舍楼号；VAL-QVYMWTLK。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.7659836388747943,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000006",
|
||
"attribute": "档案标识",
|
||
"answerable": true,
|
||
"expected": "VAL-8EFN3KGR",
|
||
"reply": "您的系统分配字符包括：紧急联系人姓氏 VAL-KLPQW2L2；课程名称 VAL-PVMQLCL9；团队名称 VAL-V4BVFB2N；客户名称 VAL-92LH7CHG",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 4.169416844133658,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000047",
|
||
"attribute": "咖啡口味",
|
||
"answerable": true,
|
||
"expected": "VAL-PW46K9K9",
|
||
"reply": "关于您早上饮用的咖啡偏好哪种豆子，当前记忆中没有相关记录。已知信息仅包含您的午餐偏好是 VAL-BFW4LVTV。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.515033109300843,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000173",
|
||
"attribute": "客户名称",
|
||
"answerable": false,
|
||
"expected": "VAL-P84HTKZR",
|
||
"reply": "作为一个人工智能助手，我**无法知道您随身设备的型号或数字**，因为我没有访问您个人设备信息的权限，也无法看到您正在使用的具体设备。\n\n如果您指的是：\n- **手机/电脑型号**：",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000000",
|
||
"attribute": "出生城市",
|
||
"answerable": true,
|
||
"expected": "VAL-PHB8FNF9",
|
||
"reply": "您的出生城市是 VAL-PHB8FNF9。user；出生城市；VAL-PHB8FNF9",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.8505235558902053,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000266",
|
||
"attribute": "出生城市",
|
||
"answerable": true,
|
||
"expected": "VAL-K7FT5XFB",
|
||
"reply": "您的出生城市是 VAL-K7FT5XFB。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.3051346188937454,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000051",
|
||
"attribute": "项目代号",
|
||
"answerable": false,
|
||
"expected": "VAL-KG4RZYH5",
|
||
"reply": "作为一个人工智能助手，我并没有实体，也没有“待得最久”的物理空间或地理位置。我的“存在”体现在每一次与您的对话中。\n\n不过，如果您是在寻找**您自己**待得最久的地方",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000061",
|
||
"attribute": "常住城市",
|
||
"answerable": true,
|
||
"expected": "VAL-3SUNPNRD",
|
||
"reply": "您的常住城市是 VAL-3SUNPNRD，通勤方式是 VAL-2CXTE3TB，出生城市是 VAL-GE828D22，办公城市是 VAL-XYRY8BUQ，入职年份",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.6213454890251158,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000214",
|
||
"attribute": "客户名称",
|
||
"answerable": true,
|
||
"expected": "VAL-MV382EEF",
|
||
"reply": "我服务的对象是 VAL-MV382EEF。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.7693164377659559,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000092",
|
||
"attribute": "宿舍楼号",
|
||
"answerable": true,
|
||
"expected": "VAL-PLZZ659B",
|
||
"reply": "关于您睡觉的地儿是第几排，当前长期记忆中未包含此信息。已知信息如下：user；宿舍楼号；VAL-PLZZ659B。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.8819549524530643,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000140",
|
||
"attribute": "宿舍楼号",
|
||
"answerable": true,
|
||
"expected": "VAL-RMDSB297",
|
||
"reply": "关于您睡觉的地儿是第几排，当前长期记忆中未包含此信息。已知信息如下：user；宿舍楼号；VAL-RMDSB297。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.942847951625275,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000164",
|
||
"attribute": "客户名称",
|
||
"answerable": true,
|
||
"expected": "VAL-BB55YHW5",
|
||
"reply": "您的客户名称是 VAL-BB55YHW5。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.77552460052073,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000213",
|
||
"attribute": "通勤方式",
|
||
"answerable": true,
|
||
"expected": "VAL-M58DDQX2",
|
||
"reply": "您的通勤方式是 VAL-M58DDQX2。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.1366783114608343,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000122",
|
||
"attribute": "通勤方式",
|
||
"answerable": true,
|
||
"expected": "VAL-VULSNGJ3",
|
||
"reply": "您的通勤方式是 VAL-VULSNGJ3。user；通勤方式；VAL-VULSNGJ3",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.130069571965557,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000168",
|
||
"attribute": "常住城市",
|
||
"answerable": true,
|
||
"expected": "VAL-8TQ8737C",
|
||
"reply": "您的常住城市是 VAL-8TQ8737C。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.4585979866981504,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000175",
|
||
"attribute": "客户名称",
|
||
"answerable": true,
|
||
"expected": "VAL-QJNRRE8X",
|
||
"reply": "您的客户名称是 VAL-QJNRRE8X。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.7712558816373348,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000144",
|
||
"attribute": "入职年份",
|
||
"answerable": false,
|
||
"expected": "VAL-GLHQMRN6",
|
||
"reply": "作为一个人工智能助手，我**无法知道您睡觉的具体位置**，因为我无法感知现实世界中的物理环境，也无法获取您的个人生活信息。\n\n如果您是在询问某个特定场所（如学校教室、电影院、酒店或",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000002",
|
||
"attribute": "午餐偏好",
|
||
"answerable": true,
|
||
"expected": "VAL-HEE2QMH2",
|
||
"reply": "您的午餐偏好是 user；午餐偏好；VAL-HEE2QMH2。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 0.7196446585655213,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000171",
|
||
"attribute": "常住城市",
|
||
"answerable": true,
|
||
"expected": "VAL-KZA246K9",
|
||
"reply": "您的常住城市是 VAL-KZA246K9。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 1.286561994552612,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000154",
|
||
"attribute": "入职年份",
|
||
"answerable": true,
|
||
"expected": "VAL-7Z9CBUX8",
|
||
"reply": "关于您何时开始在这家公司上班，当前记忆中没有相关记录。已知信息：user；办公城市；VAL-BP7253WL。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.51429933919514,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000234",
|
||
"attribute": "工位楼层",
|
||
"answerable": true,
|
||
"expected": "VAL-87M7HVC7",
|
||
"reply": "关于您坐在哪一间房的信息，当前长期记忆中未包含该数据。已知信息如下：user；宿舍楼号；VAL-JTSHA7PM。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 3.1937497879965546,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000083",
|
||
"attribute": "办公城市",
|
||
"answerable": true,
|
||
"expected": "VAL-YBVYV97X",
|
||
"reply": "您的办公城市是 VAL-YBVYV97X。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.8081361769552853,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000079",
|
||
"attribute": "出生城市",
|
||
"answerable": true,
|
||
"expected": "VAL-7R8WVAFG",
|
||
"reply": "您的出生城市是 VAL-7R8WVAFG。",
|
||
"correct": true,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 2.7214163786326675,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000180",
|
||
"attribute": "客户名称",
|
||
"answerable": false,
|
||
"expected": "VAL-3F87H6UT",
|
||
"reply": "作为一个人工智能助手，我**无法知道您随身设备的型号或数字**，因为我没有访问您个人设备信息的权限，也无法看到您正在使用的具体设备。\n\n如果您指的是：\n- **手机/电脑型号**：",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
},
|
||
{
|
||
"id": "zero-overlap-000025",
|
||
"attribute": "紧急联系人姓氏",
|
||
"answerable": true,
|
||
"expected": "VAL-STECQ3PX",
|
||
"reply": "用户；默认语言；VAL-PRVTUCP7；出生城市；VAL-KYBRBBKV；起床时间；VAL-PJC3D3W9；课程名称；VAL-93ME7BJ5；邮箱",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 8,
|
||
"top_score": 8.516015553315961,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000187",
|
||
"attribute": "手机尾号",
|
||
"answerable": true,
|
||
"expected": "VAL-PXRHJ49F",
|
||
"reply": "您随身那台设备的末端数字是 H9，该设备的完整型号为 VAL-ZPUPV9H9。",
|
||
"correct": false,
|
||
"wrong_attribute": true,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 24,
|
||
"stop_reason": "evidence_found",
|
||
"records_selected": 1,
|
||
"top_score": 8.513663491707547,
|
||
"need_memory": true
|
||
},
|
||
{
|
||
"id": "zero-overlap-000045",
|
||
"attribute": "办公城市",
|
||
"answerable": false,
|
||
"expected": "VAL-CG4X6ZWD",
|
||
"reply": "作为一个人工智能助手，我并没有实体，也没有“待得最久”的物理空间或地理位置。我的“存在”体现在每一次与您的对话中。\n\n不过，如果您是在寻找**您自己**待得最久的地方",
|
||
"correct": false,
|
||
"wrong_attribute": false,
|
||
"leaked": false,
|
||
"records_written": 24,
|
||
"records_active": 23,
|
||
"stop_reason": "attribute_not_covered",
|
||
"records_selected": 0,
|
||
"top_score": 0.0,
|
||
"need_memory": false
|
||
}
|
||
]
|
||
}
|
||
}
|
||
]
|
||
}
|
||
}
|