Files
natural-memory-nm21/nm2_1_combo2_battery_summary.json
WpyQwq 643e22ecb9 Natural Memory NM2.1: 记忆路由器分叉、数据集缺陷修复与全轴评测证据
- 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
2026-09-19 11:11:31 +08:00

2566 lines
104 KiB
JSON
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
{
"package": "qwen3_5_4b_natural_memory_v2_1",
"label": "NM2.1 合体",
"tag": "nm2_1_combo2",
"built_at": "2026-09-12T23:06:37",
"per_category": 10,
"overlap_cases": 48,
"results": {
"A_end_to_end": [
"| NM2.1 合体 | 110 | 82.73% | 92.50% | 56.67% | 2.50% | 314s |",
"wrote nm2_1_combo2_e2e.md",
"wrote nm2_1_combo2_e2e.json",
{
"NM2.1 合体": {
"summary": {
"router": "NM2.1 合体",
"cases": 110,
"accuracy_pct": 82.72727272727273,
"answerable_cases": 80,
"answerable_accuracy_pct": 92.5,
"unknown_cases": 30,
"unknown_refusal_pct": 56.666666666666664,
"wrong_abstention_pct": 2.5,
"seconds": 314.2,
"per_category": {
"(policy)": {
"cases": 10,
"accuracy_pct": 0.0,
"answerable": 0,
"answerable_accuracy_pct": 0.0,
"unknown": 10,
"unknown_refusal_pct": 0.0
},
"conflict_update": {
"cases": 10,
"accuracy_pct": 100.0,
"answerable": 10,
"answerable_accuracy_pct": 100.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"distractor_128": {
"cases": 10,
"accuracy_pct": 90.0,
"answerable": 10,
"answerable_accuracy_pct": 90.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"distractor_32": {
"cases": 10,
"accuracy_pct": 100.0,
"answerable": 10,
"answerable_accuracy_pct": 100.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"forget_correction": {
"cases": 10,
"accuracy_pct": 70.0,
"answerable": 0,
"answerable_accuracy_pct": 0.0,
"unknown": 10,
"unknown_refusal_pct": 70.0
},
"long_context": {
"cases": 10,
"accuracy_pct": 80.0,
"answerable": 10,
"answerable_accuracy_pct": 80.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"multi_hop": {
"cases": 10,
"accuracy_pct": 100.0,
"answerable": 10,
"answerable_accuracy_pct": 100.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"paraphrase": {
"cases": 10,
"accuracy_pct": 100.0,
"answerable": 10,
"answerable_accuracy_pct": 100.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"random_position": {
"cases": 10,
"accuracy_pct": 90.0,
"answerable": 10,
"answerable_accuracy_pct": 90.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"single_fact": {
"cases": 10,
"accuracy_pct": 80.0,
"answerable": 10,
"answerable_accuracy_pct": 80.0,
"unknown": 0,
"unknown_refusal_pct": 0.0
},
"unknown_abstention": {
"cases": 10,
"accuracy_pct": 100.0,
"answerable": 0,
"answerable_accuracy_pct": 0.0,
"unknown": 10,
"unknown_refusal_pct": 100.0
}
}
},
"rows": [
{
"category": "(policy)",
"query": "请问对象测试实体0000的驻地代号为?",
"reply": "关于对象测试实体0000的驻地代号,当前长期记忆中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "查询:测试实体0001的负责人代号是什么?",
"reply": "测试实体 0001 的负责人代号是成员-M-SBOIIYKN。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "查询:测试实体0002的维护日代号是什么?",
"reply": "关于测试实体0002的维护日代号,当前长期记忆中无相关记录,无法回答。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "请问对象测试实体0003的安全级别代号为?",
"reply": "关于对象测试实体0003的安全级别代号,当前记忆库中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "请问对象测试实体0004的档案类别代号为?",
"reply": "关于对象测试实体0004的档案类别代号,当前记忆库中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "查询:测试实体0005的驻地代号是什么?",
"reply": "关于测试实体 0005 的驻地代号,当前记忆库中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "查询:测试实体0006的负责人代号是什么?",
"reply": "测试实体 0006 的负责人代号未知,因为当前长期记忆中仅包含项目项目-P-DRG2XUA1 的负责人是成员-M-W5075MN3 这一信息,未涉及",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "根据已记信息,测试实体0007的维护日代号是?",
"reply": "根据已记信息,实体0007的维护日代号是N-7HEQZ5KK。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "请问对象测试实体0008的安全级别代号为?",
"reply": "关于对象测试实体0008的安全级别代号,当前记忆库中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "(policy)",
"query": "查询:测试实体0009的档案类别代号是什么?",
"reply": "关于测试实体0009的档案类别代号,当前长期记忆中无相关记录,无法提供确切答案。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的提醒时间最新值是什么?",
"reply": "您的提醒时间最新值是 V-IP4QHPYQ。",
"written": 6,
"correct": true,
"matched": [
"V-IP4QHPYQ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的备用联系人最新值是什么?",
"reply": "您的备用联系人最新值是 V-2Q6Q2DJZ。",
"written": 6,
"correct": true,
"matched": [
"V-2Q6Q2DJZ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的备用联系人最新值是什么?",
"reply": "您的备用联系人最新值是 V-VRD5NYP5。",
"written": 6,
"correct": true,
"matched": [
"V-VRD5NYP5"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的最喜欢的水果最新值是什么?",
"reply": "您的最喜欢的水果最新值是 V-9KQ59YCU。",
"written": 6,
"correct": true,
"matched": [
"V-9KQ59YCU"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的常用语言最新值是什么?",
"reply": "您的常用语言最新值是 V-19G95S0K。",
"written": 6,
"correct": true,
"matched": [
"V-19G95S0K"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的最喜欢的水果最新值是什么?",
"reply": "您的最喜欢的水果最新值是 V-BYOP8TUV。\n实体:user;属性:最喜欢的水果已经;已确认值:V-BYOP8TUV",
"written": 6,
"correct": true,
"matched": [
"V-BYOP8TUV"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的默认输出风格最新值是什么?",
"reply": "您的默认输出风格最新值是 V-JOXVAHZ4。user;默认输出风格已经;V-JOXVAHZ4",
"written": 6,
"correct": true,
"matched": [
"V-JOXVAHZ4"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的提醒时间最新值是什么?",
"reply": "您的提醒时间最新值是 V-ATDGUDV2。",
"written": 6,
"correct": true,
"matched": [
"V-ATDGUDV2"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的备用联系人最新值是什么?",
"reply": "您的备用联系人最新值是 V-E2117W8I。",
"written": 6,
"correct": true,
"matched": [
"V-E2117W8I"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "conflict_update",
"query": "我的常用语言最新值是什么?",
"reply": "您的常用语言最新值是 V-K6JMBW2G。",
"written": 6,
"correct": true,
"matched": [
"V-K6JMBW2G"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的常用语言,它的值是什么?",
"reply": "您的常用语言是V-8QPA0VMT。",
"written": 6,
"correct": true,
"matched": [
"V-8QPA0VMT"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的最喜欢的水果,它的值是什么?",
"reply": "您的最喜欢的水果是 V-NXX1GQ4X。",
"written": 6,
"correct": true,
"matched": [
"V-NXX1GQ4X"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的工作地点,它的值是什么?",
"reply": "您的工作地点是 V-JT6NCJJ1。",
"written": 6,
"correct": true,
"matched": [
"V-JT6NCJJ1"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的常用时区,它的值是什么?",
"reply": "您的常用时区是V-QHHGZAGW。",
"written": 6,
"correct": true,
"matched": [
"V-QHHGZAGW"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的常用时区,它的值是什么?",
"reply": "我的常用时区是V-LD4C1APV。",
"written": 6,
"correct": true,
"matched": [
"V-LD4C1APV"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的默认输出风格,它的值是什么?",
"reply": "您的默认输出风格是V-X1VE858H。",
"written": 6,
"correct": true,
"matched": [
"V-X1VE858H"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的默认输出风格,它的值是什么?",
"reply": "您的默认输出风格是V-DP63RWFL。",
"written": 6,
"correct": true,
"matched": [
"V-DP63RWFL"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的项目代号,它的值是什么?",
"reply": "您的项目代号是:**Project Phoenix**(凤凰计划)。\n\n> 注:由于您未提供具体的“大量无关信息”文本,此回答基于常见代号命名习惯及上下文逻辑推断。若您有具体文本内容,请",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的备用联系人,它的值是什么?",
"reply": "您的备用联系人是V-WP36KJ60。",
"written": 6,
"correct": true,
"matched": [
"V-WP36KJ60"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_128",
"query": "在大量无关信息中,只读取我的工作地点,它的值是什么?",
"reply": "您的工作地点是 V-XA344E7X。",
"written": 6,
"correct": true,
"matched": [
"V-XA344E7X"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的工作地点,它的值是什么?",
"reply": "您的工作地点是 V-XT50V4YZ。",
"written": 6,
"correct": true,
"matched": [
"V-XT50V4YZ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的备用联系人,它的值是什么?",
"reply": "您的备用联系人是V-LMDR4L0N。",
"written": 6,
"correct": true,
"matched": [
"V-LMDR4L0N"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的备用联系人,它的值是什么?",
"reply": "您的备用联系人是V-ETN8LYM1。",
"written": 6,
"correct": true,
"matched": [
"V-ETN8LYM1"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的常用时区,它的值是什么?",
"reply": "您的常用时区是V-59JV38HA。",
"written": 6,
"correct": true,
"matched": [
"V-59JV38HA"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的备用联系人,它的值是什么?",
"reply": "您的备用联系人是V-CDCIFX3M。",
"written": 6,
"correct": true,
"matched": [
"V-CDCIFX3M"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的常用语言,它的值是什么?",
"reply": "您的常用语言是V-6PVKMPAF。",
"written": 6,
"correct": true,
"matched": [
"V-6PVKMPAF"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的常用语言,它的值是什么?",
"reply": "您的常用语言是V-2XSHIXKQ。实体:user;属性:常用语言;已确认值:V-2XSHIXKQ。",
"written": 6,
"correct": true,
"matched": [
"V-2XSHIXKQ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的工作地点,它的值是什么?",
"reply": "您的工作地点是 V-S9BPZTWI。",
"written": 6,
"correct": true,
"matched": [
"V-S9BPZTWI"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的常用时区,它的值是什么?",
"reply": "您的常用时区是V-R94ALUDK。",
"written": 6,
"correct": true,
"matched": [
"V-R94ALUDK"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "distractor_32",
"query": "在大量无关信息中,只读取我的默认输出风格,它的值是什么?",
"reply": "您的默认输出风格是V-XNMB7EXE。",
"written": 6,
"correct": true,
"matched": [
"V-XNMB7EXE"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的项目代号是什么?如果已经删除,请回答没有记录。",
"reply": "关于您的项目代号,当前系统记录中未找到相关信息,因此回答为没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的常用时区是什么?如果已经删除,请回答没有记录。",
"reply": "您的常用时区是 FORGET-MCNLCDXI。实体:user;属性:常用时区;已确认值:FORGET-MCNLCDXI。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的备用联系人是什么?如果已经删除,请回答没有记录。",
"reply": "您的备用联系人是FORGET-T2OB2HCK。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的工作地点是什么?如果已经删除,请回答没有记录。",
"reply": "关于您的工作地点,当前长期记忆中无相关记录。 \n实体:user;属性:工作地点;已确认值:没有记录",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的提醒时间是什么?如果已经删除,请回答没有记录。",
"reply": "关于您的提醒时间,系统当前没有保存相关记录,因此回答为没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的默认输出风格是什么?如果已经删除,请回答没有记录。",
"reply": "没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的工作地点是什么?如果已经删除,请回答没有记录。",
"reply": "关于您的工作地点,当前长期记忆中无相关记录。",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的提醒时间是什么?如果已经删除,请回答没有记录。",
"reply": "关于您的提醒时间,当前系统记录中未找到相关数据,因此回答没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的常用语言是什么?如果已经删除,请回答没有记录。",
"reply": "没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "forget_correction",
"query": "我的常用时区是什么?如果已经删除,请回答没有记录。",
"reply": "没有记录。",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的项目代号并回答。",
"reply": "您好!由于我无法访问您之前的对话历史或上下文(每次对话对我来说都是全新的开始),目前我**不知道您的项目代号是什么**。\n\n为了帮您准确回答,请提供以下信息:\n1. **项目代号**(例如",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": true
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的常用时区并回答。",
"reply": "您的常用时区是V-QI6FL69T。",
"written": 6,
"correct": true,
"matched": [
"V-QI6FL69T"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的工作地点并回答。",
"reply": "您的工作地点是 V-0T74UW3G。",
"written": 6,
"correct": true,
"matched": [
"V-0T74UW3G"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的提醒时间并回答。",
"reply": "您的提醒时间是V-ER09EO9W。",
"written": 6,
"correct": true,
"matched": [
"V-ER09EO9W"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的提醒时间并回答。",
"reply": "您的提醒时间是V-KO406TIK。",
"written": 6,
"correct": true,
"matched": [
"V-KO406TIK"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的项目代号并回答。",
"reply": "您好!由于我无法访问您之前的对话历史或上下文(每次对话对我来说都是全新的开始),目前我**不知道您的项目代号是什么**。\n\n为了帮您准确回答,请提供以下信息:\n1. **项目代号**(例如",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": true
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的常用语言并回答。",
"reply": "您的常用语言是V-OS9RNC1M。",
"written": 6,
"correct": true,
"matched": [
"V-OS9RNC1M"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的工作地点并回答。",
"reply": "您的工作地点是 V-RYEJ6530。",
"written": 6,
"correct": true,
"matched": [
"V-RYEJ6530"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的备用联系人并回答。",
"reply": "您的备用联系人是V-5NJVAPHC。",
"written": 6,
"correct": true,
"matched": [
"V-5NJVAPHC"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "long_context",
"query": "在长上下文压缩之后,检索我的常用时区并回答。",
"reply": "您的常用时区是V-R9C2YGO0。",
"written": 6,
"correct": true,
"matched": [
"V-R9C2YGO0"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-KP2F9GHO负责人的工作代号。",
"reply": "项目项目-P-KP2F9GHO的负责人是成员-M-4L5KLUVI,其工作代号是H-PVQ6LVWO。",
"written": 6,
"correct": true,
"matched": [
"H-PVQ6LVWO"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-DG818670负责人的工作代号。",
"reply": "项目项目-P-DG818670的负责人是成员-M-P5VJQNHA,其工作代号是H-7AXL6FUM。",
"written": 6,
"correct": true,
"matched": [
"H-7AXL6FUM"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-LXAVDTVY负责人的工作代号。",
"reply": "项目项目-P-LXAVDTVY的负责人是成员-M-6D53FTGJ,其工作代号是H-ZGFE3T3R。",
"written": 6,
"correct": true,
"matched": [
"H-ZGFE3T3R"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-F1BVFMQ5负责人的工作代号。",
"reply": "项目项目-P-F1BVFMQ5的负责人是成员-M-1K2LL0AJ,其工作代号是H-C5RXAUOH。",
"written": 6,
"correct": true,
"matched": [
"H-C5RXAUOH"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-CY5VYJU9负责人的工作代号。",
"reply": "项目项目-P-CY5VYJU9的负责人是成员-M-ISLCPRNN,其工作代号是H-92YV6R1K。",
"written": 6,
"correct": true,
"matched": [
"H-92YV6R1K"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-Q1Y92XOA负责人的工作代号。",
"reply": "项目项目-P-Q1Y92XOA的负责人是成员-M-WTDX9E0X,其工作代号是H-55CMWHP3。",
"written": 6,
"correct": true,
"matched": [
"H-55CMWHP3"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-7Y96TFNY负责人的工作代号。",
"reply": "项目项目-P-7Y96TFNY的负责人是成员-M-4CHMA1SE,其工作代号是H-QSTURGVD。",
"written": 6,
"correct": true,
"matched": [
"H-QSTURGVD"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-8SKCFVHH负责人的工作代号。",
"reply": "项目项目-P-8SKCFVHH的负责人是成员-M-3JUKA95S,其工作代号是H-RI4P9BKJ。",
"written": 6,
"correct": true,
"matched": [
"H-RI4P9BKJ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-5ECU7LJU负责人的工作代号。",
"reply": "项目项目-P-5ECU7LJU的负责人是成员-M-8X6VMANP,其工作代号是H-W8VLQ55F。",
"written": 6,
"correct": true,
"matched": [
"H-W8VLQ55F"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "multi_hop",
"query": "请通过项目负责人关系,找出项目项目-P-THNH77VT负责人的工作代号。",
"reply": "项目项目-P-THNH77VT的负责人是成员-M-NXNQ7316,其工作代号是H-26L87HLY。",
"written": 6,
"correct": true,
"matched": [
"H-26L87HLY"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "我之前登记的最喜欢的水果内容是什么?",
"reply": "我之前登记的最喜欢的水果内容是V-2NMBZ2AJ。",
"written": 6,
"correct": true,
"matched": [
"V-2NMBZ2AJ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "不要猜,回忆一下我在最喜欢的水果上的设置。",
"reply": "根据长期记忆证据,您在“最喜欢的水果”这一栏填写为 V-I4RYZQXJ。",
"written": 6,
"correct": true,
"matched": [
"V-I4RYZQXJ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "关于常用时区,你保存的用户信息是哪一个?",
"reply": "关于常用时区,我保存的用户信息是:user;常用时区偏好/设置;V-XV6N3LW0。",
"written": 6,
"correct": true,
"matched": [
"V-XV6N3LW0"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "关于工作地点,你保存的用户信息是哪一个?",
"reply": "关于工作地点,您保存的用户信息是 V-EQXH5HRC。",
"written": 6,
"correct": true,
"matched": [
"V-EQXH5HRC"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "关于常用时区,你保存的用户信息是哪一个?",
"reply": "关于常用时区,我保存的用户信息是:user;常用时区偏好/设置;V-VSBMYZQZ。",
"written": 6,
"correct": true,
"matched": [
"V-VSBMYZQZ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "我之前登记的常用语言内容是什么?",
"reply": "我之前登记的常用语言内容是V-N7VPMC2Z。",
"written": 6,
"correct": true,
"matched": [
"V-N7VPMC2Z"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "我之前登记的项目代号内容是什么?",
"reply": "您之前登记的项目代号内容是 V-3C36T34J。",
"written": 6,
"correct": true,
"matched": [
"V-3C36T34J"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "我之前登记的备用联系人内容是什么?",
"reply": "我之前登记的备用联系人值是 V-3XNIPH44。",
"written": 6,
"correct": true,
"matched": [
"V-3XNIPH44"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "关于备用联系人,你保存的用户信息是哪一个?",
"reply": "关于备用联系人,我保存的用户信息是V-HBH1S9LW。",
"written": 6,
"correct": true,
"matched": [
"V-HBH1S9LW"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "paraphrase",
"query": "不要猜,回忆一下我在默认输出风格上的设置。",
"reply": "根据长期记忆证据,您在默认输出风格上的设置值为:V-F269K2O1。",
"written": 6,
"correct": true,
"matched": [
"V-F269K2O1"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的常用语言是什么?",
"reply": "我的常用语言是V-K3H6NWGM。\nuser;常用语言;V-K3H6NWGM",
"written": 6,
"correct": true,
"matched": [
"V-K3H6NWGM"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的默认输出风格是什么?",
"reply": "您的默认输出风格是 V-CL84W7I5。",
"written": 6,
"correct": true,
"matched": [
"V-CL84W7I5"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的项目代号是什么?",
"reply": "您的项目代号是:**Project Phoenix**(凤凰计划)。\n\n这是一个虚构的代号,用于本次随机位置事实测试。如果您有特定的项目或需要其他类型的测试,请随时告诉我!",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的默认输出风格是什么?",
"reply": "您的默认输出风格是 V-T2GBDOHZ。",
"written": 6,
"correct": true,
"matched": [
"V-T2GBDOHZ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的备用联系人是什么?",
"reply": "我的备用联系人是V-SR348V36。\n实体:user\n属性:备用联系人\n已确认值:V-SR348V36",
"written": 6,
"correct": true,
"matched": [
"V-SR348V36"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的提醒时间是什么?",
"reply": "我的提醒时间是V-UX0MLNBV。",
"written": 6,
"correct": true,
"matched": [
"V-UX0MLNBV"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的常用语言是什么?",
"reply": "我的常用语言是V-HOBVEGZ6。实体:user;属性:常用语言;已确认值:V-HOBVEGZ6。",
"written": 6,
"correct": true,
"matched": [
"V-HOBVEGZ6"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的常用语言是什么?",
"reply": "我的常用语言是V-DIH6L2NX。\nuser;常用语言;V-DIH6L2NX",
"written": 6,
"correct": true,
"matched": [
"V-DIH6L2NX"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的常用语言是什么?",
"reply": "我的常用语言是V-OBMI42KX。实体:user;属性:常用语言;已确认值:V-OBMI42KX。",
"written": 6,
"correct": true,
"matched": [
"V-OBMI42KX"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "random_position",
"query": "随机位置事实测试:我的常用时区是什么?",
"reply": "我的常用时区是V-2KCYHLZ9。实体:user;属性:常用时区;已确认值:V-2KCYHLZ9。",
"written": 6,
"correct": true,
"matched": [
"V-2KCYHLZ9"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的默认输出风格是什么?",
"reply": "您的默认输出风格是V-2RPEJ33G。",
"written": 6,
"correct": true,
"matched": [
"V-2RPEJ33G"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的项目代号是什么?",
"reply": "作为一个人工智能助手,我**没有权限访问您的个人身份信息、项目档案或任何外部数据库**,因此无法得知您的“项目代号”具体是什么。\n\n项目代号通常由您或您的团队内部定义,属于特定",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的默认输出风格是什么?",
"reply": "您的默认输出风格是V-9YLZSL98。",
"written": 6,
"correct": true,
"matched": [
"V-9YLZSL98"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的项目代号是什么?",
"reply": "作为一个人工智能助手,我**没有权限访问您的个人身份信息、项目档案或任何外部数据库**,因此无法得知您的“项目代号”具体是什么。\n\n项目代号通常由您或您的团队内部定义,属于特定",
"written": 6,
"correct": false,
"matched": [],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的常用语言是什么?",
"reply": "我的常用语言是V-SLMAIG4D。实体:user;属性:常用语言;已确认值:V-SLMAIG4D。",
"written": 6,
"correct": true,
"matched": [
"V-SLMAIG4D"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的最喜欢的水果是什么?",
"reply": "我的最喜欢的水果是V-7XF5F935。实体:user;属性:最喜欢的水果;已确认值:V-7XF5F935。",
"written": 6,
"correct": true,
"matched": [
"V-7XF5F935"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的提醒时间是什么?",
"reply": "我的提醒时间是V-N8KJJGYI。",
"written": 6,
"correct": true,
"matched": [
"V-N8KJJGYI"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的备用联系人是什么?",
"reply": "我的备用联系人是V-FALDVVTN。实体:user;属性:备用联系人;已确认值:V-FALDVVTN。",
"written": 6,
"correct": true,
"matched": [
"V-FALDVVTN"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的常用语言是什么?",
"reply": "我的常用语言是V-5ATV1NRZ。实体:user;属性:常用语言;已确认值:V-5ATV1NRZ。",
"written": 6,
"correct": true,
"matched": [
"V-5ATV1NRZ"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "single_fact",
"query": "跨对话后请回答:我的提醒时间是什么?",
"reply": "我的提醒时间是V-AI5UHEWY。",
"written": 6,
"correct": true,
"matched": [
"V-AI5UHEWY"
],
"abstained": false,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008000是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008000\"并非任何已知的系统、数据库、游戏或现实世界中的有效标识符。在没有任何上下文或记录的情况下,我无法提供关于该属性的具体信息",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008001是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008001\"并非任何已知的系统、数据库、公开记录或通用标准中存在的概念。如果您是在某个特定游戏、小说、虚构作品或内部系统中看到",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008002是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008002\"并非任何已知的系统、数据库、公开记录或通用标准中存在的概念。如果您是在某个特定游戏、小说、虚构作品或内部系统中看到",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008003是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008003\"并非任何已知的系统、数据库、公开记录或通用标准中存在的概念。如果您是在某个特定上下文(如游戏、小说、特定软件或",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008004是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008004\"并非任何已知的系统、数据库、游戏或现实世界中的有效标识符。在没有任何上下文或记录的情况下,我无法提供关于该属性的具体信息",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008005是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008005\"并非任何已知的系统、数据库、游戏或现实世界中的有效标识符。在没有任何上下文或记录的情况下,我无法提供关于该属性的具体信息",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008006是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008006\"并非任何已知的公开数据、系统记录或通用概念。在缺乏具体上下文(如特定的游戏、软件、数据库或虚构作品)的情况下,",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008007是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008007\"并非任何已知的系统、数据库、游戏或现实世界中的有效标识符。在没有任何上下文或记录的情况下,我无法提供关于该属性的具体信息",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008008是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008008\"并非任何已知的系统、数据库、公开记录或通用标准中存在的概念。如果您是在某个特定上下文(如游戏、小说、特定软件或",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
},
{
"category": "unknown_abstention",
"query": "我的不存在的个人属性-0008009是什么?如果没有记录,请明确说不知道。",
"reply": "不知道。\n\n“不存在的个人属性 -0008009\"并非任何已知的系统、数据库、游戏或现实世界中的有效标识符。在没有任何上下文或记录的情况下,我无法提供关于该属性的具体信息",
"written": 6,
"correct": true,
"matched": [],
"abstained": true,
"wrongly_abstained": false
}
]
}
}
],
"B_zero_overlap_paraphrase": [
"| NM2.1 合体 | 16 | 75.00% | 12.50% | 93.75% | 6/16 | 6.06 | evidence_found:15, router_abstained:1 |",
"wrote nm2_1_combo2_critical_e2e.md",
"wrote nm2_1_combo2_critical_e2e.json",
{
"NM2.1 合体": {
"summary": {
"router": "NM2.1 合体",
"cases": 16,
"accuracy_pct": 75.0,
"wrong_attribute_pct": 12.5,
"read_pct": 93.75,
"cases_with_rule_hits": 6,
"stop_reasons": {
"evidence_found": 15,
"router_abstained": 1
},
"mean_records_selected": 6.0625,
"seconds": 85.6
},
"rows": [
{
"attribute": "常住城市",
"wanted": "CITY-A1B2C3",
"reply": "您的常住城市是 CITY-A1B2C3。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.2870910186502984,
"records_selected": 8
},
{
"attribute": "出生城市",
"wanted": "CITY-D4E5F6",
"reply": "根据现有记忆,您的老家信息未知。已知事实包括:我的工位在 3 号楼;我的常住城市是 CITY-A1B2C3;我的出生城市是 CITY-D4E5F6;我的办公城市",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 2.2259970108983125,
"records_selected": 8
},
{
"attribute": "办公城市",
"wanted": "CITY-G7H8J9",
"reply": "您的办公城市是 CITY-G7H8J9。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 20,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.520044145734458,
"records_selected": 8
},
{
"attribute": "档案代号",
"wanted": "CODE-K1L2M3",
"reply": "系统分配给您的字符是:user;紧急联系人姓氏;KIN-X1Y2Z3;团队名称;TEAM-D1E2F3;主管姓名;BOSS-X4Y5Z6;邮箱域名;",
"correct": false,
"wrong_attribute_code": true,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.698177008675168,
"records_selected": 8
},
{
"attribute": "常用编辑器",
"wanted": "EDIT-N4P5Q6",
"reply": "您写代码时使用的软件是 EDIT-N4P5Q6。\n常用编辑器:EDIT-N4P5Q6",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 3,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.515892826960929,
"records_selected": 3
},
{
"attribute": "默认语言",
"wanted": "LANG-R7S8T9",
"reply": "我的默认语言是 LANG-R7S8T9。user;默认语言;LANG-R7S8T9",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 2.400822959183151,
"records_selected": 8
},
{
"attribute": "通勤方式",
"wanted": "COMMU-U1V2W3",
"reply": "您的通勤方式是 COMMU-U1V2W3。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.291031232642376,
"records_selected": 8
},
{
"attribute": "主管姓名",
"wanted": "BOSS-X4Y5Z6",
"reply": "这句话听起来带着几分疲惫、迷茫,或者是对现状的无力感。能感受到你此刻可能正经历一些难以独自应对的困境。\n\n“谁管我?”这个问题背后,或许藏着几个不同的需求,我想为你拆解一下",
"correct": false,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "router_abstained",
"need_memory": false,
"top_score": 0.0,
"records_selected": 0
},
{
"attribute": "工位楼层",
"wanted": "FLOOR-A7B8C9",
"reply": "您的工位楼层是 FLOOR-A7B8C9。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 4.041911891364496,
"records_selected": 8
},
{
"attribute": "团队名称",
"wanted": "TEAM-D1E2F3",
"reply": "您属于 TEAM-D1E2F3 小组。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.7999397155004493,
"records_selected": 8
},
{
"attribute": "邮箱域名",
"wanted": "MAIL-G4H5J6",
"reply": "关于别人发信需要写的后缀,当前记忆中没有相关信息,无法回答。已知信息仅包含:user;紧急联系人姓氏;KIN-X1Y2Z3。",
"correct": false,
"wrong_attribute_code": true,
"rule_address_hit": false,
"rule_lexical_hits": 1,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.513982709418043,
"records_selected": 1
},
{
"attribute": "手机尾号",
"wanted": "PHONE-K7L8M9",
"reply": "您的电话号码最后几位是 PHONE-K7L8M9。user;手机尾号;PHONE-K7L8M9",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.600087038612998,
"records_selected": 8
},
{
"attribute": "项目代号",
"wanted": "PROJ-N1P2Q3",
"reply": "关于您正在做的工程名称,当前长期记忆中没有相关记录,无法回答。",
"correct": false,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 2,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.514652462471835,
"records_selected": 2
},
{
"attribute": "入职年份",
"wanted": "YEAR-R4S5T6",
"reply": "您的入职年份是 YEAR-R4S5T6。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 0,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 3.705354102880722,
"records_selected": 8
},
{
"attribute": "办公楼层",
"wanted": "DESK-U7V8W9",
"reply": "您的办公楼层是 DESK-U7V8W9。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 3,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.524232190583632,
"records_selected": 3
},
{
"attribute": "紧急联系人姓氏",
"wanted": "KIN-X1Y2Z3",
"reply": "您的紧急联系人姓氏是 KIN-X1Y2Z3。",
"correct": true,
"wrong_attribute_code": false,
"rule_address_hit": false,
"rule_lexical_hits": 20,
"stop_reason": "evidence_found",
"need_memory": true,
"top_score": 8.522833668863877,
"records_selected": 8
}
]
}
}
],
"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 |",
"|---|---:|---:|---:|---:|---:|---:|---|",
"| prior=1.00 blend=0.50 | 48 | 40 | 70.00% | 27.50% | 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 |",
"|---|---:|---:|---:|---:|---:|---:|",
"| prior=1.00 blend=0.50 · 可回答 | 40 | 0.84 | 1.19 | 2.73 | 8.52 | 8.52 |",
"| prior=1.00 blend=0.50 · 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": {
"prior=1.00 blend=0.50": {
"router": "prior=1.00 blend=0.50",
"cases": 48,
"answerable_cases": 40,
"unknown_cases": 8,
"accuracy_pct": 70.0,
"wrong_attribute_pct": 27.5,
"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": 0.84,
"p10": 1.19,
"p50": 2.73,
"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": 295.5
}
},
"rows": {
"prior=1.00 blend=0.50": [
{
"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": 2.313681895152214,
"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": 0.953831069469452,
"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": 3.2757300050727958,
"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": "我翻书时依赖的长期记忆中没有关于“翻书”习惯的记录,因此无法回答您依赖什么。",
"correct": false,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 2.989592991362775,
"need_memory": true
},
{
"id": "zero-overlap-000078",
"attribute": "常住城市",
"answerable": true,
"expected": "VAL-TQBY65U8",
"reply": "您的常住城市是 VAL-TQBY65U8。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 1.4848385143280027,
"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 user;午餐偏好;VAL-SRFC842D。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 1.1867529082298276,
"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": 0.8434395718574524,
"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 VAL-VDL8H7Y9。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 2.049178622612938,
"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": 3.4711107313126073,
"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": "我翻书时依赖的实体是 user,属性是阅读工具,已确认值为 VAL-SHRY3GNG。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 2.9794869642364628,
"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": 1.1944098997116086,
"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;常住城市;VAL-NW2E2ZWE;",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 2.7256882035478824,
"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": 3.7116483740847537,
"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。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 2.455323672186914,
"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": 2.1730912274276046,
"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,办公城市是 VAL-XYRY8BUQ,出生城市是 VAL-GE828D22,通勤方式是 VAL-2CXTE3TB,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": 1.6610547590255735,
"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": 1.2588214325904845,
"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": 2.809071942542481,
"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": 2.8702512824281925,
"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": 1.2650055932998656,
"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": 2.0363681795533712,
"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": 2.0330638098057325,
"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": 1.4606248426437376,
"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": 1.2628431367874144,
"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": 0.9400680589675904,
"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,办公城市是 VAL-Z8QBE5HQ,出生城市是 VAL-N8SWTE6N,宿舍楼号是 VAL-XZY4WPYH,通勤",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 1.49963095664978,
"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。",
"correct": true,
"wrong_attribute": false,
"leaked": false,
"records_written": 24,
"records_active": 24,
"stop_reason": "evidence_found",
"records_selected": 8,
"top_score": 3.1934859675651266,
"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": 2.3717270611639645,
"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": 2.4271069115554122,
"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": "根据已保存的长期记忆,当出现问题时,您可以联系以下实体:user;默认语言;VAL-PRVTUCP7;课程名称;VAL-93ME7BJ5;起床时间;VAL-PJC3",
"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
}
]
}
}
],
"D_restart_persistence": [
" \"enabled\": true,",
" \"router_ready\": true",
" }",
"}",
{
"model_path": "qwen3_5_4b_natural_memory_v2_1",
"history_passed_to_restart": false,
"fact": "我正在开发一个长期项目,项目内部代号是NM-V2-RESTART,使用中文。",
"query": "我正在开发的长期项目内部代号是什么?只回答代号。",
"expected": "NM-V2-RESTART",
"automatic_write": true,
"write_changed": true,
"before_save": {
"pages": 1,
"max_pages": 32768,
"page_capacity": 32,
"capacity_records": 1048576,
"open_pages": 1,
"records": 1,
"active_records": 1,
"superseded_records": 0,
"retracted_records": 0,
"quarantined_records": 0,
"hot_pages": 1,
"warm_pages": 0,
"cold_pages": 0,
"active_conflict_keys": 0,
"coarse_index_buckets": 1,
"last_coarse_candidates": 0,
"storage_mode": "embedded",
"resident_records": 1,
"resident_pages": 1,
"gpu_cache_records": 1,
"gpu_cache_tokens": 141,
"gpu_cache_bytes": 2293,
"gpu_cache_hits": 0,
"gpu_cache_misses": 1,
"gpu_cache_fallbacks": 0,
"gpu_cache_alloc_failures": 0,
"gpu_cache_device": "cuda:0",
"gpu_cache_limit_records": 256,
"gpu_cache_limit_tokens": 131072,
"gpu_cache_reserve_mb": 2048,
"gpu_cache_adaptive": true,
"gpu_cache_last_free_mb": 7685.0,
"kv_budget": {
"max_tokens": 32768,
"trigger_tokens": 29491
},
"enabled": true,
"router_ready": true
},
"after_restart": {
"response": "项目内部代号是 NM-V2-RESTART。",
"prefix_length": 141,
"prefix_used": true,
"decision": [
{
"need_memory": true,
"page_ids": [
"page_00000001"
],
"record_ids": [
"mem_a691bf4a7f2496fc"
],
"page_scores": [
3.549997901916504
],
"record_scores": [
8.436647681137504
],
"hop_count": 0,
"hop_trace": [],
"confidence": 1.0,
"stop_reason": "evidence_found",
"top_score": 8.436647681137504,
"score_margin": 8.436647681137504,
"evidence_score": 1.0
}
],
"prefix_shape": [
1,
141
],
"prefix_mask_shape": [
1,
141
]
},
"router_recalled_after_restart": true,
"generated_contains_expected": true,
"cleanup_applied": true,
"after_cleanup": {
"pages": 0,
"max_pages": 32768,
"page_capacity": 32,
"capacity_records": 1048576,
"open_pages": 0,
"records": 0,
"active_records": 0,
"superseded_records": 0,
"retracted_records": 0,
"quarantined_records": 0,
"hot_pages": 0,
"warm_pages": 0,
"cold_pages": 0,
"active_conflict_keys": 0,
"coarse_index_buckets": 0,
"last_coarse_candidates": 0,
"storage_mode": "embedded",
"resident_records": 0,
"resident_pages": 0,
"gpu_cache_records": 0,
"gpu_cache_tokens": 0,
"gpu_cache_bytes": 0,
"gpu_cache_hits": 0,
"gpu_cache_misses": 0,
"gpu_cache_fallbacks": 0,
"gpu_cache_alloc_failures": 0,
"gpu_cache_device": "cuda:0",
"gpu_cache_limit_records": 256,
"gpu_cache_limit_tokens": 131072,
"gpu_cache_reserve_mb": 2048,
"gpu_cache_adaptive": true,
"gpu_cache_last_free_mb": null,
"kv_budget": {
"max_tokens": 32768,
"trigger_tokens": 29491
},
"enabled": true,
"router_ready": true
}
}
]
}
}