SkillHone: A Harness for Continual Agent Skill Evolution Through Persistent Decision History
cs.LG
Submitted: 2026-06-07
Updated: 2026-10-29
Comments: Accepted to EMNLP 2026 Industry Track. 11 pages, 4 figures, 6 tables. Updated to the published version, including optimization cost and efficiency analysis
Journal ref: Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pp. 1819-1829 (2026)
Code: https://github.com/Tencent/SkillHone
Project page: https://zwlijay.github.io/SkillHone-Project
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Large Language Model based Multi-Agents: A Survey of Progress and Challenges
- SWE-Skills-Bench: Do Agent Skills Actually Help in Real-World Software Engineering?
- DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines
- SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks
- Agent Skills: A Data-Driven Analysis of Claude Skills for Extending Large Language Model Functionality
- SkillForge: Forging Domain-Specific, Self-Evolving Agent Skills in Cloud Technical Support
- Multi-Agent Collaboration: Harnessing the Power of Intelligent LLM Agents
- WebShaper: Agentically Data Synthesizing via Information-Seeking Formalization
- Multi-Agent Collaboration Mechanisms: A Survey of LLMs
- Skill-R1: Agent Skill Evolution via Reinforcement Learning
- Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?
- Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward
- Automated Skill Discovery for Language Agents through Exploration and Iterative Feedback
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis