SkillNEO: Non-Elitist Evolutionary Optimization of Agent Skills
Agent skills turn procedural knowledge into reusable artifacts, allowing LLM agents to accumulate and refine expertise through experience.
ProofPaper ↗
Key points
- Building on this, skill evolution iteratively improves skills using execution feedback.
- We introduce SkillNEO (Non-Elitist Evolutionary Optimization for Agent Skills), which separately selects a version to use and versions to keep improving.
- Across 86 SkillsBench tasks in eight domains, including software engineering and office work, SkillNEO raises GPT-5.4 Mini's average reward from 49.1% to 66.4%, within GPT-5.5 and GPT-5.6 Luna's original-skill range.
- Together, these findings show that skill evolution can turn a weaker model's partial progress into reusable procedural knowledge for improving more capable agents.
Sources (1)
- [1]SkillNEO: Non-Elitist Evolutionary Optimization of Agent SkillsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 08:57 PM
Agent skills turn procedural knowledge into reusable artifacts, allowing LLM agents to accumulate and refine expertise through experience.
Building on this, skill evolution iteratively improves skills using execution feedback.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 14, 2026ARIA: Runtime Inspection and Amendment for Multi-Agent Systems
- Oct 14, 2026Research@ 2026: 10 ways we're amplifying human ingenuity
- Oct 13, 2026anisayari/sparky: 🔒 Your private AI assistant, 100% local on Windows: chat, vision, documents (RAG), web search, images, videos with sound, voice, code projec
- Oct 13, 2026Plan or Retrieve First? A Scale Study of Tool Discovery Architectures
- Oct 13, 2026Google wants to host your personal vibe-coded apps in the Play Store
- Oct 13, 2026Natural language security queries at Orca Security using Amazon Bedrock