Tag
The paper introduces MATE, a deterministic post-retrieval memory adaptation procedure that converts retrieved trajectories into execution-oriented memory for embodied agents, achieving 81.3% and 93.3% success on 134 ALFWorld tasks with Qwen2.5-14B and 72B while using about one-tenth the tokens of raw trajectories.
This paper identifies post-retrieval reuse as a bottleneck for long-horizon agent memory and proposes query-conditioned reuse (QCR), a simple target-bound note format, showing improved success and token efficiency across WebArena, WorkArena, and AppWorld.