跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.10170 · v1

Where Did the Evidence Go? Diagnosing Proof-Set Loss in Long-Term Conversational Memory

首次提交 2026-11-07 03:47
版本更新 2026-11-07 03:47
雷达观测 2026-11-10 10:11
arXiv ↗PDF ↗

摘要

Long-term conversational question answering can require several separated dialogue turns that support an answer only together. We introduce EviCover, a retrieval approach designed to retain this complementary evidence. It converts questions into observable evidence requirements, checks requirement coverage using retrieved source identifiers, and selects turns jointly. We track annotated support sets - the benchmark-labeled dialogue turns associated with each question - through retrieval, reranking, and final answer-generation context. We evaluate the baseline memory pipeline on generated, human-verified LoCoMo conversations and human-written REALTALK conversations, and examine EviCover on REALTALK's 108 multi-hop questions. Compared with the hybrid lexical/dense baseline, support-turn recall among 60 candidate turns rises from 0.526 to 0.575, and recall in the final ten-turn context rises from 0.349 to 0.379. The evidence funnel identifies separate discovery and selection failures, while paired answer evaluation shows that greater evidence recall does not yet establish a correctness gain. This analysis locates where complementary evidence is lost and distinguishes evidence retrieval from its use in answer generation.

作者备注:10 pages, 2 figures, 8 tables

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-07 03:47 · 最新观测版本

Where Did the Evidence Go? Diagnosing Proof-Set Loss in Long-Term Conversational Memory

Long-term conversational question answering can require several separated dialogue turns that support an answer only together. We introduce EviCover, a retrieval approach designed to retain this complementary evidence. It converts questions into observable evidence requirements, checks requirement coverage using retrieved source identifiers, and selects turns jointly. We track annotated support sets - the benchmark-labeled dialogue turns associated with each question - through retrieval, reranking, and final answer-generation context. We evaluate the baseline memory pipeline on generated, human-verified LoCoMo conversations and human-written REALTALK conversations, and examine EviCover on REALTALK's 108 multi-hop questions. Compared with the hybrid lexical/dense baseline, support-turn recall among 60 candidate turns rises from 0.526 to 0.575, and recall in the final ten-turn context rises from 0.349 to 0.379. The evidence funnel identifies separate discovery and selection failures, while paired answer evaluation shows that greater evidence recall does not yet establish a correctness gain. This analysis locates where complementary evidence is lost and distinguishes evidence retrieval from its use in answer generation.

查看此版本解读 →查看此版本原文 ↗