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
29 lines
781 B
JSON
29 lines
781 B
JSON
{
|
|
"held_out_attributes": [
|
|
"午餐偏好",
|
|
"客户名称",
|
|
"紧急联系人姓氏",
|
|
"运动习惯",
|
|
"邮箱域名",
|
|
"项目代号"
|
|
],
|
|
"trained_on_attributes": 18,
|
|
"eval_held_out_queries": 56,
|
|
"eval_kept_queries": 194,
|
|
"max_prob_held_out": {
|
|
"p10": 0.1102,
|
|
"p50": 0.1771,
|
|
"p90": 0.1854
|
|
},
|
|
"max_prob_kept": {
|
|
"p10": 0.2125,
|
|
"p50": 0.3445,
|
|
"p90": 0.4934
|
|
},
|
|
"openset_auc_held_out_vs_kept": 0.9629,
|
|
"closed_set_false_claim_pct": 100.0,
|
|
"reject_threshold_from_kept_p05": 0.1752,
|
|
"held_out_rejected_pct_at_threshold": 44.64,
|
|
"kept_retained_pct_at_threshold": 100.0,
|
|
"note": "held-out attributes never appear as training labels, so this is the open-set case a real user bank faces"
|
|
} |