aiqbench analysis
Trace2Skill addresses the scalability bottleneck of LLM agent development by distilling execution trajectories into transferable skills that generalize across model scales and families without requiring parameter updates. The framework effectively compresses recurring failures into standard operating procedures, outperforming both sequential editing and retrieval-based memory banks in diverse domains like math reasoning and vision QA. This approach offers a practical path toward robust, portable agent capabilities by leveraging inductive reasoning over agent experience rather than relying solely on parametric knowledge.
Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
arXiv:2603.25158v5 Announce Type: replace Abstract: Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually authoring such skills does not scale, and skills generated purely from parametric knowledge often miss critical operational pitfalls. We introduce Trace2Skill, a framework that consolidates broad execution trajectories in parallel into a unified skill directory through inductive reasoning over agent experience. Trace2Skill supports both deepening existing human-written skills and creating useful skills from weak LLM-generated drafts. Experiments demonstrate the effectiveness of Trace2Skill across diverse domains, including office workflows, math reasoning, and vision QA. Importantly, the evolved skills are not merely memorized artifacts of the trajectories used to create them: they often transfer across model scales, across model families, and to out-of-distribution settings. For example, skills evolved from Qwen3.5-35B trajectories improve a Qwen3.5-122B agent by up to $57.65$ percentage points on WikiTableQuestions. Further analyses show that Trace2Skill outperforms sequential skill editing and ReasoningBank-style retrieval memories, compresses recurring failures and workarounds into standard operating procedures (SoPs), and yields portable skills that can be reused without parameter updates or test-time retrieval.
Abstract:Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually authoring such skills does not scale, and skills generated purely from parametric knowledge often miss critical operational pitfalls. We introduce Trace2Skill, a framework that consolidates broad execution trajectories in parallel into a unified skill directory through inductive reasoning over agent experience. Trace2Skill supports both deepening existing human-written skills and creating useful skills from weak LLM-generated drafts. Experiments demonstrate the effectiveness of Trace2Skill across diverse domains, including office workflows, math reasoning, and vision QA. Importantly, the evolved skills are not merely memorized artifacts of the trajectories used to create them: they often transfer across model scales, across model families, and to out-of-distribution settings. For example, skills evolved from Qwen3.5-35B trajectories improve a Qwen3.5-122B agent by up to $57.65$ percentage points on WikiTableQuestions. Further analyses show that Trace2Skill outperforms sequential skill editing and ReasoningBank-style retrieval memories, compresses recurring failures and workarounds into standard operating procedures (SoPs), and yields portable skills that can be reused without parameter updates or test-time retrieval.
Work in Progress. May version add more experiments
Artificial Intelligence (cs.AI)
arXiv:2603.25158 [cs.AI]
(or arXiv:2603.25158v5 [cs.AI] for this version)
Full text under the publisher's open licence · 189 words
Related coverage
Socratic Skill Learning: Interleaving Teaching and Practice for Transferable Agent Skills
arXiv · Oct 15, 2026
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
[AINews] DeepSeek v4.1-Flash: 763B-P8B-D16B novel causal Encoder–Decoder architecture with vision marks the Return of the Whale
Latent Space · Sep 12, 2026
[AINews] not much happened today
Latent Space · Oct 3, 2026