BAAI's DisCo Distills 1,000 GitHub Repos Into 5,000 Reusable ML Skills, Lifts MLE-Bench 134%

Beijing Academy of AI's Repo-To-Skill paper introduces DisCo, a two-mode framework that turns GitHub repos and papers into ready-to-load operational-knowledge 'skills' at roughly $40 per repo. The published AREX-Skill library carries 5,000+ verified skills from 1,000 repos across 20 areas and 178 capability families. Applied to a baseline Codex agent, skills lifted MLE-bench from 31.11% to 72.89% (+134.3%), PaperBench from 29.45% to 39.59%, and cut token/step usage vs stronger unaided baselines.