Tag
The paper introduces Retrieval-Invoked Actual-Use Effect (RAE) to evaluate whether retrieved skills actually help LLM agents, showing that aggregate metrics can mislead by hiding negative effects on specific tasks.
This paper introduces SkillZip, a contract-preserving graph compression framework that compresses agent skill libraries at the section level, improving retrieval and compression efficiency while preserving procedural contracts and dependency closures.
SkillSight is a training-free retrieval framework that calibrates shared background in skill descriptions to improve skill retrieval accuracy for LLM agents, achieving up to 20.21 percentage point improvement in Recall@10 over dense retrievers.
Proposes SkillReranker, an inference-time reranking framework that decomposes tasks and skills to build a directed acyclic execution graph and uses a cross-encoder for scoring, improving task performance and efficiency in agent skill selection.
This paper introduces Group of Skills (GoSkills), a retrieval method that organizes atomic skills into role-labeled execution contexts to improve agent performance within limited context budgets.
This paper introduces SkillRet, a large-scale benchmark for evaluating skill retrieval in LLM agents, addressing the challenge of selecting relevant skills from large libraries. It provides a dataset of over 17,000 skills and demonstrates that task-specific fine-tuning significantly improves retrieval performance.