CoEvo-Mem: Co-Evolving Retrieval Policy and Memory Bank for LLM Agents
2026-08-03 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors propose a new method, CoEvo-Mem, to improve how large language models remember and use past information. Unlike previous methods that either improve how memories are found or how memories are updated, their approach links these two steps in a feedback loop. This means the system learns better which memories to retrieve and how to update them based on task results. They tested CoEvo-Mem on multiple tasks and found it outperforms existing methods by jointly evolving both memory retrieval and updates.
long-term memorylarge language modelsmemory retrievalmemory updatefeedback loopquery rewritingrouting policynon-stationarityclosed-loop learningcontinual learning
Authors
Bowen Ye, Yongchao Xu, Zhijian Li, Xiang Yin, Junkai Ma, Wenzhao Li
Abstract
As memories accumulate across tasks and sessions, the performance of long-term LLM agents depends jointly on query-specific retrieval and continual memory refinement. However, existing methods typically optimize either memory access, through iterative query refinement or adaptive retrieval policies, or memory evolution such as structural update. This separation overlooks a fundamental feedback loop: retrieval determines which memories receive usage signals, while updated memory bank reshape future retrieval. We propose \textbf{CoEvo-Mem}, a closed-loop framework for co-evolving the retrieval policy and memory bank. For each query, a frozen LLM generates route-specific query rewrites and a routing prior, which a lightweight residual router corrects online. The retrieved context serves as the coupling interface between the two learning processes: task outcomes assign credit to routing decisions, while trajectory-conditioned feedback updates memory values and graph relations. These updates alter how memories are ranked and selected for subsequent queries, thereby closing the feedback loop. To mitigate coupling induced non-stationarity, CoEvo-Mem alternates between updating the router with the memory bank fixed and evolving the memory bank with the retrieval policy fixed. Across seven diverse benchmarks, \textbf{CoEvo-Mem} achieves state-of-the-art performance, demonstrating the importance of retrieval-memory coevolution.