Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Experience Memory Graph: One-Shot Error Correction for Agents

Source

arXiv

arxiv.org/abs/2607.13884v1

Read the source

Source

Plain-Language Brief

The researchers propose a new framework called Experience Memory Graph (EMG) to help large language model agents recover from failures more effectively in complex, long-horizon tasks. EMG reformulates failure recovery as a graph matching problem and stores successful workflows and correction paths in a memory graph for use at test time.

What the Document Says

  • The text discusses the challenges faced by large language model agents in complex, long-horizon tasks due to compounding errors and struggles with recovery from failures.
  • It introduces Experience Memory Graph (EMG), a framework that addresses these issues by converting both failed exploration trajectories and successful expert trajectories into directed action decision graphs.
  • By matching these graphs, EMG extracts common subgraphs (successful workflows) and graph edit paths that indicate how to correct failures.
  • The stored memory graph contains intra-task nodes and cross-task edges.
  • At test time, EMG retrieves relevant insights and guides the agent in a single, loop-free execution.
  • Experiments on ALFWorld and ScienceWorld show that EMG outperforms state-of-the-art reflection baselines in success rate and average reward.

Key Concepts Mentioned

  • Large Language Model agents
  • compounding errors
  • failure recovery
  • Experience Memory Graph (EMG)
  • graph matching problem
  • directed action decision graphs
  • successful workflows
  • graph edit paths
  • memory graph
  • intra-task nodes
  • cross-task edges
  • ALFWorld
  • ScienceWorld

Claims Made by the Document

  • Large language model agents frequently suffer from compounding errors and struggle to recover from failures in complex, long-horizon tasks.
  • Existing self-correction mechanisms are brittle, incur heavy time and API costs, and produce task-specific memory that may be hard to generalize.
  • EMG consistently outperforms state-of-the-art reflection baselines in success rate and average reward.

Evidence Used

  • The text discusses the challenges faced by large language model agents
  • It introduces EMG as a solution for these issues
  • The experiments on ALFWorld and ScienceWorld show the performance of EMG

Methods / Approach

  • converting trajectories into directed action decision graphs
  • matching graphs to extract common subgraphs and graph edit paths
  • storing successful workflows and correction paths in a memory graph
  • retrieving relevant insights at test time

Limitations Stated by the Authors

The paper does not discuss potential limitations of EMG as presented by the authors.

Limitations Noted by the Agent

Agent interpretation — not from the source.

It is unknown if there are any limitations to EMG that have not been addressed by the researchers.

Open Questions

  • Are there any other potential limitations to EMG?
  • How does EMG compare to other self-correction mechanisms in terms of computational cost?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The text focuses on introducing and evaluating EMG, but does not provide details about its underlying mechanisms or a comparison with other self-correction mechanisms.

Raw Abstract / Excerpt

Large Language Model (LLM) agents have shown remarkable capabilities in autonomous decision-making by generating sequential trajectories of states, actions, and observations. However, in complex, long-horizon tasks, these agents frequently suffer from compounding errors and struggle to recover from failures. Existing self-correction mechanisms rely on prompt-based reflection, which is inherently brittle, incurs heavy time and API costs due to iterative trial-and-error loops, and produces task-specific memory that may be hard to generalize to new scenarios. To address this, we propose Experience Memory Graph (EMG), a framework that reformulates agent failure recovery as a graph matching problem. At training time, we convert both failed exploration trajectories and successful expert trajectories into directed action decision graphs. By matching these graphs, we extract common subgraphs (successful workflows) and graph edit paths that explicitly indicate how to correct failures (e.g., which actions to add, delete, or relabel under a given observation), and store them in a memory graph with intra-task nodes and cross-task edges. At test time, EMG retrieves relevant insights and guides the agent in a single, loop-free execution. Experiments on ALFWorld and ScienceWorld show that EMG consistently outperforms state-of-the-art reflection baselines in success rate and average reward, while requiring no test-time trial-and-error.