跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.09807 · v1

Compaction, decay and reinforcement in a two-tier conversational memory: a paired evaluation against retrieval-augmented generation

两层对话记忆中的压缩、衰减与强化:与检索增强生成的配对评估

首次提交 2026-11-07 01:56
版本更新 2026-11-07 01:56
雷达观测 2026-11-09 13:22
arXiv ↗PDF ↗

摘要

A conversational agent that remembers has to decide what to keep. Retrieval-augmented generation keeps the transcript and searches it; the alternative is to compact each turn into atomic facts and search those. We compare the two on LoCoMo, with both arms answering all 1,985 questions in one window, against the same stores, with the same generator and judge, and both predictions on one row, which licenses McNemar's exact test on the 512 questions where they disagree. On the four categories comparable work reports, compaction scores 71.4 per cent against 64.6, a lead of 6.8 points (p = 5.5e-08) at 34 per cent fewer prompt tokens. On all five, including the adversarial set comparable work drops, the lead falls to 2.0 and stops being significant, because the baseline leads adversarial by 14.6. That loss has a structure. Our lead falls as the reference answer demands more source wording, 9.6 points ahead at one or two reference words and 3.1 behind past eight, with token overlap moving the same way, so the judge is not the cause. Compaction keeps what a fact asserts and discards how it was said, so it wins where the answer is a value and loses where it is a phrasing. We also report four negative results with their power calculations.

作者备注:7 figures. Data and code: https://github.com/BigBrown10/nmafc-paper

AI 论文解读

基于摘要 · 非全文审阅

在LoCoMo上配对比较“逐轮压缩为原子事实”与“保留原文检索”两种对话记忆方案,发现压缩在事实型问题上占优、在措辞依赖型问题上落后。

正在读取账号与研究记录…

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 01:56 · 最新观测版本

Compaction, decay and reinforcement in a two-tier conversational memory: a paired evaluation against retrieval-augmented generation

A conversational agent that remembers has to decide what to keep. Retrieval-augmented generation keeps the transcript and searches it; the alternative is to compact each turn into atomic facts and search those. We compare the two on LoCoMo, with both arms answering all 1,985 questions in one window, against the same stores, with the same generator and judge, and both predictions on one row, which licenses McNemar's exact test on the 512 questions where they disagree. On the four categories comparable work reports, compaction scores 71.4 per cent against 64.6, a lead of 6.8 points (p = 5.5e-08) at 34 per cent fewer prompt tokens. On all five, including the adversarial set comparable work drops, the lead falls to 2.0 and stops being significant, because the baseline leads adversarial by 14.6. That loss has a structure. Our lead falls as the reference answer demands more source wording, 9.6 points ahead at one or two reference words and 3.1 behind past eight, with token overlap moving the same way, so the judge is not the cause. Compaction keeps what a fact asserts and discards how it was said, so it wins where the answer is a value and loses where it is a phrasing. We also report four negative results with their power calculations.

查看此版本解读 →查看此版本原文 ↗