Hippo：免训练的智能体自进化框架，借助海马体回放机制学习经验
Hippo: Training-Free Agent Self-Evolution via Hippocampal Replay
AI 导读
研究人员提出 Hippo，一种受海马体选择性编码与回放启发的记忆框架，让 LLM 智能体无需更新模型参数即可从任务尝试中学习。该框架通过对比同一任务的多次尝试提炼可复用决策，并按适用条件合并重叠、解决冲突经验。在 SWE-bench Verified、WebArena 和 Mind2Web 三个基准上均取得提升，且保留经验数远少于积累基线。
正文 · 原文
Abstract:Motivated by recursive self-improvement (RSI), we study how LLM agents can learn through external memory without updating model parameters. Extracting useful experience from a single attempt is difficult: flawed reasoning can produce success, while poor execution can obscure a sound strategy. Preserving its value over time is also difficult: our accumulation baseline without consolidation initially improves but eventually performs below a no-memory baseline. Our key idea is to compare attempts at the same task to identify reusable decisions and refine when they apply. We introduce Hippo, a memory framework inspired by selective hippocampal encoding and replay that learns from these comparisons when shortcomings are diagnosed. It compares the resulting experience with existing memory to merge overlaps and resolve conflicts through explicit applicability conditions. This allows prior guidance to be revised without storing every lesson as a separate entry. Experiments on SWE-bench Verified, WebArena, and Mind2Web show performance gains in all three settings and substantially fewer retained experiences than the accumulation baseline. The gains extend across the evaluated base models of varying capabilities, indicating that memory-based adaptation remains useful even for highly capable models.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.24453 [cs.AI] |
| (or arXiv:2610.24453v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.24453 arXiv-issued DOI via DataCite |
Submission history
From: Yutong Song [view email]
[v1]
Sat, 17 Oct 2026 04:43:26 UTC (873 KB)
[v2]
Thu, 29 Oct 2026 17:38:48 UTC (873 KB)
来源：arXiv cs.AI · arxiv.org