TEPA: Revoking Stale Memories for Conflict-Robust Language Agents
2026-08-07 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address a problem where outdated memories in language agents cause errors because they conflict with newer information. They propose TEPA, a method that tags memories with a validity status and revokes old facts when new evidence contradicts them. This helps the agent use only up-to-date memories while still keeping a history for checking past information. Their experiments show TEPA improves accuracy compared to methods that only add or overwrite memories without revocation. However, challenges remain in complex cases requiring multi-step reasoning or very long memory contexts.
long-term memorylanguage agentsmemory pollutionfalsifiabilityTEPAevidence revocationfact consolidationmemory retrievalcontext-selectionpreference update
Authors
Yan Zhou, Yue Ouyang, Kaiyang Zheng, Suncheng Xiang
Abstract
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this failure mode as memory pollution: degradation caused by active memories that newer conflicting evidence has superseded.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.