RESEARCH · RESEARCH · #2326
SPEM: structure-aware procedural memory for efficient skill use in LLM agents (arXiv preprint)
The paper introduces SPEM, a lightweight, structure-aware procedural memory framework that stores Atomic Skill Memory for individual skills and Skill-DAG Memory for reusable skill compositions; it uses Skill-GNN for state- and structure-aware selection and writes successful executions back as new Skill-DAGs. On ALFWorld and SkillsBench, SPEM achieves an 87.1% ALFWorld success rate and 38.38% SkillsBench reward while reducing token and runtime costs compared to strong baselines.
KEY POINTS
- The paper introduces SPEM, a lightweight, structure-aware procedural memory framework that stores Atomic Skill Memory for individual skills and Skill-DAG Memory for reusable skill compositions; it uses Skill-GNN for state- and structure-aware selection and writes successful executions back as new Skill-DAGs.
- On ALFWorld and SkillsBench, SPEM achieves an 87.1% ALFWorld success rate and 38.38% SkillsBench reward while reducing token and runtime costs compared to strong baselines.
- SPEM offers a scalable way to reuse and compose skills in LLM agents with lower token/runtime overhead, improving practical efficiency and effectiveness of agent behavior.
WHY IT MATTERS
SPEM offers a scalable way to reuse and compose skills in LLM agents with lower token/runtime overhead, improving practical efficiency and effectiveness of agent behavior.