EMMA: Structured Episodic Memory for Multi-Agent Workflow Synthesis
Ruijia Niu, Kaiwen Liu, Rose Yu and Kristina Klinkner (Amazon, UC San Diego) present EMMA, which treats a multi-agent system's configuration as structured episodic memory and restricts each self-improvement edit to the component responsible for the diagnosed failure.
Ask this paper
What counts as memory. Controller routing memory, specialist prompts that inherit from a generalist prompt, and per-step procedural skills are all treated as editable episodic memory.
Scoped updates. A reflector assigns each diagnosed failure to the scope that owns its correction and proposes one edit there. Guards that do not depend on the backbone keep every accepted edit scoped, incremental and bounded.
Why it matters. Without this scoping, evolution follows the backbone's habits rather than the observed failures: one backbone keeps adding skills while another keeps spawning agents.
Results. On KramaBench, AppWorld and BBEH with five backbones, EMMA is best overall on every benchmark against ACE, GEPA, EvoMAS and ADAS, by up to 32.5% over the strongest baseline, and uses 17 to 38% fewer training executions than the most efficient baseline.
Abstract
Deploying LLM agents in specialized domains often require fine-grained task-specific procedures and knowledge a general model lacks, yet parameter training is costly and non-interpretable, and hand-engineered agent systems shift the cost to expert labor. Episodic memory updates at context level, but single agent adaptation struggles to alleviate cross-task interference, especially with complex tasks such as multi-application navigation or data pool analysis. Although multi-agent evolution alleviates this problem, existing methods lack clear definitions on update rules and rely on the backbone model to decide where and how much the system changes. This ill-scoped evolution makes learning follow the backbone's structural preferences rather than the observed failures: one backbone accumulates skills while another spawns agents. We present EMMA (structured Episodic Memory for Multi-Agent workflow synthesis), which treats the configuration of a multi-agent system (controller routing memory, specialists inheriting a generalist prompt, and per-step procedural skills) as episodic memory. EMMA makes the ownership and admissibility of updates explicit: a reflector assigns each diagnosed failure to the scope owning its correction and proposes one edit within it, and backbone-independent guards keep every accepted edit scoped, incremental and bounded. Across KramaBench, AppWorld and BBEH with five backbones and four baselines (ACE, GEPA, EvoMAS and ADAS), EMMA attains the best overall performance on every benchmark, up to 32.5% over the strongest baseline, maintains consistent improvements across backbones, and uses 17-38% fewer training executions than the most efficient baseline.