TEPA：撤销过时记忆以构建冲突鲁棒的语言智能体
TEPA: Revoking Stale Memories for Conflict-Robust Language Agents
浏览论文内容
中文总结 AI 辅助
该研究针对语言智能体的记忆污染问题，提出TEPA可撤销证据记忆机制，经多场景实验证实其能有效避免过时记忆干扰，性能优于仅追加、最后写入获胜等机制。
中文摘要 AI 辅助
长期记忆使语言智能体能够复用过往事实、偏好和任务经验，但持久性也带来了核心的可证伪性问题：当世界发生变化时，过时记忆仍可被检索并污染提示。我们将这种失效模式定义为记忆污染：由被新的冲突证据取代的活跃记忆导致的性能下降。我们提出TEPA，一种可撤销的证据记忆机制，将有效性设为记忆的显式状态。TEPA将观测表示为带键的先例，当新证据在同一键下与活跃先例冲突时，会撤销该活跃先例，使检索可从当前证据中获取信息，同时保留已撤销的历史用于审计。在受控的隐藏规则漂移、真实文件支持的可执行漂移以及偏好更新流中，撤销操作可防止过时的活跃记忆在反转后仍保留在检索集中。在50个种子的受控漂移实验中，仅追加和最后写入获胜记忆机制在完全反转时的表现低于无记忆机制（仅追加和最后写入获胜均为0.210，无记忆为0.309，TEPA为0.950），在真实文件执行场景中也呈现相同模式（仅追加为0.203，无记忆为0.298，TEPA为0.950）。在干净的MemoryAgentBench SH-6k上，TEPA与性能强劲的最后写入获胜缓存表现相当，证实当前键替换是单跳事实整合的决定性操作。在多跳和超长上下文的MemoryAgentBench设置上的边界测试则揭示了除事实级有效性跟踪之外的检索链和上下文选择瓶颈。这些结果共同确立了生命周期撤销作为智能体的核心记忆操作，适用于必须证伪、审计并后续重新促进演进知识的场景。
英文摘要
Long-term memory enables language agents to reuse past facts, preferences, and task experience. Persistence also creates a central falsifiability problem: when the world changes, stale memories can remain retrievable and pollute the prompt. We characterize this failure mode as memory pollution: degradation caused by active memories that newer conflicting evidence has superseded. We introduce TEPA, a revocable evidence-memory mechanism that makes validity an explicit state of memory. TEPA represents observations as keyed precedents and revokes active precedents when fresh evidence contradicts them under the same key, allowing retrieval to draw from current evidence while preserving revoked history for audit. Across controlled hidden-regime drift, real file-backed executable drift, and preference-update streams, revocation prevents stale active memory from remaining in the retrieval set after reversal. In controlled drift over 50 seeds, append-only and last-write-wins memory fell below no memory during full reversal (append-only and last-write-wins both 0.210, no memory 0.309, TEPA 0.950), and the same pattern reproduced under real file execution (append-only 0.203, no memory 0.298, TEPA 0.950). On clean MemoryAgentBench SH-6k, TEPA matches a strong last-write-wins cache, confirming that current-key replacement is the decisive operation for single-hop fact consolidation. Boundary tests on multi-hop and very long-context MemoryAgentBench settings expose retrieval-chain and context-selection bottlenecks beyond fact-level validity tracking. Together, these results establish lifecycle revocation as a core memory operation for agents that must falsify, audit, and later re-promote evolving knowledge.
