{ "eval_file": "data/zero_overlap/eval.jsonl", "episodes": 48, "candidate_count": 24, "router": "checkpoints\\router_replay_v7_v2_128\\memory_router_v2.pt", "results": { "prior=1.00 blend=1.00": { "router": "prior=1.00 blend=1.00", "cases": 48, "answerable_cases": 40, "unknown_cases": 8, "accuracy_pct": 62.5, "wrong_attribute_pct": 37.5, "unknown_leak_pct": 87.5, "mean_records_selected": 6.25, "records_active_min": 23, "records_active_max": 24, "stop_reasons": { "evidence_found": 48 }, "top_score_answerable": { "n": 40, "min": 0.68, "p10": 1.68, "p50": 2.99, "p90": 8.52, "max": 8.52 }, "top_score_unknown": { "n": 8, "min": 1.72, "p10": 1.72, "p50": 2.99, "p90": 8.51, "max": 8.51 }, "seconds": 303.2 } }, "rows": { "prior=1.00 blend=1.00": [ { "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": 2.0630308857713966, "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": 1.2078188586235044, "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": 3.1921160952918166, "need_memory": true }, { "id": "zero-overlap-000246", "attribute": "常住城市", "answerable": false, "expected": "VAL-6BWJYDE4", "reply": "关于您小时候待的地方,当前长期记忆中未包含此信息。已知信息如下:user;常住城市;VAL-6BWJYDE4;团队名称;VAL-B5WGYYND;默认语言;VAL-Z7", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 2.031137963995817, "need_memory": true }, { "id": "zero-overlap-000204", "attribute": "阅读工具", "answerable": true, "expected": "VAL-ZRLSPU5V", "reply": "我翻书时依赖的阅读工具是 user;阅读工具;VAL-ZRLSPU5V。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 3.0003122906791813, "need_memory": true }, { "id": "zero-overlap-000078", "attribute": "常住城市", "answerable": true, "expected": "VAL-TQBY65U8", "reply": "您的办公城市是 VAL-JZ94C3K2。", "correct": false, "wrong_attribute": true, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7171694326400755, "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": 1.6311666893959043, "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": "我闹钟设定的时刻是 VAL-8CT2M285。", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 3.1921160952918166, "need_memory": true }, { "id": "zero-overlap-000091", "attribute": "运动习惯", "answerable": true, "expected": "VAL-92MQHEDL", "reply": "关于您每天练习的具体内容,当前长期记忆中未包含相关事实。已知信息如下:user;课程名称;VAL-6VKAPDJL;主管姓名;VAL-6GH64SXW;默认语言;VAL", "correct": false, "wrong_attribute": true, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 0.6810586022399366, "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": "关于您服务的对象,当前长期记忆中未包含相关信息,无法回答。", "correct": false, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7549998807907103, "need_memory": true }, { "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 VAL-VDL8H7Y9。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": 2.1584236472344247, "need_memory": true }, { "id": "zero-overlap-000184", "attribute": "项目代号", "answerable": true, "expected": "VAL-3H5TQEC4", "reply": "您的项目代号是 VAL-3H5TQEC4。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 3.750910937735413, "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": 3.0001864602687434, "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": 1.6799999999999997, "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(宿舍)的排号信息,当前长期记忆中仅包含宿舍楼号 VAL-QVYMWTLK,未包含具体的排号数据,因此无法回答。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 2.9856021845087284, "need_memory": true }, { "id": "zero-overlap-000006", "attribute": "档案标识", "answerable": true, "expected": "VAL-8EFN3KGR", "reply": "您的档案标识是 VAL-8EFN3KGR。user;档案标识;VAL-8EFN3KGR", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 3.2538799031045267, "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": "您随身那台设备的末端数字是 Y。\n设备型号:VAL-NARXVSSY", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 1, "top_score": 8.513663491707547, "need_memory": true }, { "id": "zero-overlap-000000", "attribute": "出生城市", "answerable": true, "expected": "VAL-PHB8FNF9", "reply": "您的出生城市是 VAL-PHB8FNF9。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 2.269420659434381, "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": 2.0410478210603027, "need_memory": true }, { "id": "zero-overlap-000051", "attribute": "项目代号", "answerable": false, "expected": "VAL-KG4RZYH5", "reply": "您的办公城市是 VAL-W9KGRD9U。", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7158904361724852, "need_memory": true }, { "id": "zero-overlap-000061", "attribute": "常住城市", "answerable": true, "expected": "VAL-3SUNPNRD", "reply": "您的办公城市是 VAL-XYRY8BUQ。", "correct": false, "wrong_attribute": true, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7166983771324156, "need_memory": true }, { "id": "zero-overlap-000214", "attribute": "客户名称", "answerable": true, "expected": "VAL-MV382EEF", "reply": "您的服务对象是 VAL-MV382EEF。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.755, "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(宿舍床位)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": 2.9856145822748417, "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(宿舍)的排号信息,当前长期记忆中仅包含宿舍楼号 VAL-RMDSB297,未包含具体的排号数据,因此无法回答。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 2.985639735434937, "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": 1.755, "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": 1.9360580488325603, "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。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.9360580484272303, "need_memory": true }, { "id": "zero-overlap-000168", "attribute": "常住城市", "answerable": true, "expected": "VAL-8TQ8737C", "reply": "您的办公城市是 VAL-5RQ9PMSR。", "correct": false, "wrong_attribute": true, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7161377358436583, "need_memory": true }, { "id": "zero-overlap-000175", "attribute": "客户名称", "answerable": true, "expected": "VAL-QJNRRE8X", "reply": "我服务的对象是 VAL-QJNRRE8X。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.755, "need_memory": true }, { "id": "zero-overlap-000144", "attribute": "入职年份", "answerable": false, "expected": "VAL-GLHQMRN6", "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;办公城市;VAL-F4A9ADMW;常住城市;VAL-M7H3ZUK5;工位", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 2.985480233405518, "need_memory": true }, { "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": 1.1604914593696591, "need_memory": true }, { "id": "zero-overlap-000171", "attribute": "常住城市", "answerable": true, "expected": "VAL-KZA246K9", "reply": "您的办公城市是 VAL-Z8QBE5HQ,常住城市是 VAL-KZA246K9,出生城市是 VAL-N8SWTE6N,邮箱域名是 VAL-S7EXABKA,宿舍楼号", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.7173039603233335, "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": "您的工位楼层是 VAL-87M7HVC7,宿舍楼号是 VAL-JTSHA7PM。", "correct": true, "wrong_attribute": false, "leaked": false, "records_written": 24, "records_active": 24, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 3.193222147133863, "need_memory": true }, { "id": "zero-overlap-000083", "attribute": "办公城市", "answerable": true, "expected": "VAL-YBVYV97X", "reply": "您的白天上班地点(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": 1.9353806868678216, "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": 2.281658059250894, "need_memory": true }, { "id": "zero-overlap-000180", "attribute": "客户名称", "answerable": false, "expected": "VAL-3F87H6UT", "reply": "您随身那台设备的末端数字是 VAL-X5VZ48QF。", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 1, "top_score": 8.513663491707547, "need_memory": true }, { "id": "zero-overlap-000025", "attribute": "紧急联系人姓氏", "answerable": true, "expected": "VAL-STECQ3PX", "reply": "根据长期记忆证据,您的关键联系信息如下:user;默认语言;VAL-PRVTUCP7;起床时间;VAL-PJC3D3W9;课程名称;VAL-93ME7BJ5;", "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": "您的办公城市是 VAL-SZ4R5GT3。", "correct": false, "wrong_attribute": false, "leaked": true, "records_written": 24, "records_active": 23, "stop_reason": "evidence_found", "records_selected": 8, "top_score": 1.717385559082031, "need_memory": true } ] } }