Long term memory faults hurt ai agent task correctness substantially

Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress Test

Artificial IntelligenceSoftware Engineering

Summary

Large language model (LLM) agents store memories to help complete tasks. The authors found that sometimes the system controlling which memories an agent uses gets 'corrupted,' causing the agent to rely on memories incorrectly, even when those memories are correctly saved. They tested multiple LLM providers and showed this kind of error happens often and significantly reduces task accuracy. Fixing this memory control state restored performance almost fully. This work highlights a new type of fault that should be checked alongside normal memory retrieval accuracy.

What this means in practice

  • For ai system engineers: Detect and correct memory control state faults to improve accuracy in LLM-based agents using long-term memory.
  • For software reliability teams: Test memory governance integrity alongside retrieval correctness to ensure dependable LLM agent behavior over time.

Authors

Bowen Liu, Youting Wang, Bonan Shen, Dingyan Shang

Abstract

An LLM agent can retrieve a memory correctly yet use it incorrectly after its validity state changes. We study governance-state corruption: faults that alter which retained records enter model-visible context while stored content and tasks remain fixed. Across six providers, 731 benchmark-grounded bases yield 2,080 matched fault pairs. In offline control-plane checks, corruption changes the operative projection in 96.25% of pairs, while oracle-backed enforcement restores it in 100%. For downstream evaluation, a deterministically selected, non-probability LongMemEval/Memora panel resolves 298 of 300 matched triplets (238 distinct bases). Clean-minus-corrupted contrasts are 40.54 percentage points in LongMemEval binary correctness (95% base-cluster interval [30.32, 50.71]) and 30.60 points in rescaled Memora FAMA ([23.31, 37.69]); both sources also show recovery. OAKS Novel native option parity holds for all 6,714 resolved outputs. Full-cohort frozen-scorer corruption/recovery effects (-27.55/+27.51 points) remain descriptive proxies, not validated native semantic population effects. These bounded findings motivate evaluating memory control-state integrity alongside retrieval accuracy.