Memory access adaptation improves long term agent retrieval accuracy
Adaptive Memory Index Evolution for Agents
Machine Learning
Summary
Agents that remember past experiences sometimes struggle to find the right memory when they need it. The authors introduce a new method called AMIE that changes how an agent chooses which memories to use without altering the memories themselves. This method learns from example decisions about which memories to keep or suppress and adjusts future memory access accordingly. Surprisingly, even a simple approach using how often memories are retrieved works well. Their experiments show better accuracy and retrieval quality across many systems and tasks.
What this means in practice
- •For software engineers: Improve AI agents’ ability to recall relevant past information by adding an adaptive layer that optimizes memory access decisions over time.
- •For customer support teams: Enhance retrieval of previous customer interactions in AI chat systems by evolving memory access policies to prioritize useful past conversations.
Authors
Yuntong Hu, Tongli Su, Yuanfu Sun, Shuo Ma, Qiaoyu Tan, Liang Zhao
Abstract
Long-horizon agents continually accumulate experience, yet retaining a memory does not guarantee access when it is needed. Existing systems can adapt both memory content and access; we isolate persistent access updates as a separately controlled adaptation layer. We formulate this target as Memory Index Evolution and instantiate it with Adaptive Memory Index Evolution (AMIE), a lightweight plug-in layer that evolves per-memory access state from deployment retrievals without modifying stored content or retraining the host memory system. AMIE distils a semantic teacher's keep/suppress judgments into a retrieval-conditioned policy and accumulates its selective actions into bounded state that reweights subsequent host retrievals. Surprisingly, even a zero-call policy that simply accumulates retrieval frequency can match or outperform semantic policies in several settings, revealing retrieval traffic itself as a strong signal for access adaptation, while learned selectivity improves robustness when frequently retrieved memories remain useful. Across five retrieval-based hosts and three benchmarks, trained AMIE improves downstream F1 or accuracy in all 15 comparisons, by 0.70--3.37 percentage points (up to 25% relative), and improves retrieval in all 10 comparisons. Controlled analyses show that persistent competitor suppression can expose previously unseen evidence, while suppressing repeatedly useful memories creates the complementary failure mode. Together, these results establish persistent memory access as a practical adaptation layer, distinct from the content it serves, in retrieval-based agent memory systems.