RESEARCH · RESEARCH · #2090
SkillNEO: non-elitist evolutionary optimization of agent skills (arXiv:2610.20926v1)
arXiv:2610.20926v1 presents SkillNEO (Non-Elitist Evolutionary Optimization for Agent Skills), which decouples which skill version is executed from which versions are retained for further edits so that low-scoring intermediates can be revised rather than discarded. On 86 SkillsBench tasks across eight domains (e.g., software engineering and office work), SkillNEO raises GPT-5.4 Mini's average reward from 49.1% to 66.4%; reused modified skills improve GPT-5.6 Luna by 10.5 percentage points on 41 modified tasks and place GPT-5.5 10.6 points above its published original-skill reference.
KEY POINTS
- arXiv:2610.20926v1 presents SkillNEO (Non-Elitist Evolutionary Optimization for Agent Skills), which decouples which skill version is executed from which versions are retained for further edits so that low-scoring intermediates can be revised rather than discarded.
- On 86 SkillsBench tasks across eight domains (e.g., software engineering and office work), SkillNEO raises GPT-5.4 Mini's average reward from 49.1% to 66.4%; reused modified skills improve GPT-5.6 Luna by 10.5 percentage points on 41 modified tasks and place GPT-5.5 10.6 points above its published original-skill reference.
- Non-elitist evolution preserves useful partial progress and produces reusable procedural knowledge that can raise stronger agents' performance, affecting how agent skills can be autonomously developed and transferred.
WHY IT MATTERS
Non-elitist evolution preserves useful partial progress and produces reusable procedural knowledge that can raise stronger agents' performance, affecting how agent skills can be autonomously developed and transferred.