The paper presents DisCo, an agent that distills operational knowledge from GitHub repositories into compact, verified skills. These skills are created in two ways: task-agnostic (covering widely used ML repositories) and task-oriented (built for specific tasks). The resulting AREX-Skill Library contains over 5,000 skills from 1,000 repositories, organized into 20 areas and 178 capability families. With a fixed GPT-5.5 backbone and research harness, the skill-equipped agent outperformed the same agent without skills on all four benchmarks: MLE-bench (134.3% higher), PaperBench (34.4% higher), FrontierCS (9.2% higher), and PassNet (14.0% higher). This work is interesting because it shows that the know-how scattered across code repositories can be distilled into reusable units, allowing agents to apply proven procedures instead of rediscovering them through trial and error.
Distilled repository skills improve agents conducting machine-learning research
DisCo converts practical knowledge from GitHub repositories into compact skills that substantially improved a research agent on four benchmarks.
Academic
Jianlyu Chen · Yuyang Hu · Hongjin Qian · Jiawei Liu · Wenqing Wei · Xiaolong Chen · +5 more
Beijing Academy of Artificial Intelligence · University of Science and Technology of China · Renmin University of China · Hong Kong Polytechnic University
Research Digest··2 min read
In one line
Distilling GitHub repositories into reusable skills boosts autonomous ML research agent performance on four benchmarks.
Before you read this
Three ideas this paper builds onagent skillsLearned capabilities that enable AI agents to perform specialized tasks with improved effectiveness.106 papers →agent harnessA system that structures, verifies, monitors, or controls how an AI agent interacts with external tools, environments, or execution runs.150 papers →verificationChecking outputs for correctness before accepting them, often to catch errors or reward hacking.277 papers →
Read it as
The authors built an agent that distills operational know-how from repositories and papers, then retrieves those skills while carrying out machine-learning research tasks.
The authors developed DisCo, a research agent that creates and applies reusable skills encoding operational knowledge—the practical steps required to make methods work.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.