ml-intern: Open-source AI agent that automates ML post-training, achieving +22 GPQA points in 10 hours
By
Aymeric Roucher
Leave it on the tray for the seagulls.
Summary
An open-source AI agent called ml-intern, featured on Product Hunt, fully automates post-training for machine learning models. It reads arXiv papers, fixes and creates datasets, runs training jobs, debugs failures, and iterates autonomously. The tool achieved +22 points on GPQA in 10 hours and +60% on HealthBench, positioning itself as a significant advancement for automated ML research.
Key quotes
· 3 pulledAn open-source AI agent that fully automates post-training: reads arXiv papers, fixes & creates datasets, runs training jobs, debugs failures, and iterates all by itself.
Results: +22 pts on GPQA in 10h and +60% on HealthBench.
The future of ML research is here.
You might also wanna read
Autonomous AI Research Agents for Single-GPU Nanochat Training Automation
The article describes an AI research automation project called 'autoresearch' that enables autonomous AI agents to conduct machine learning
github.com·2mo ago
Duolingo open-sources AI Slack agent that connects to 200+ engineering tools
Duolingo developed an internal AI Slack app that connects to over 200 tools (GitHub, Jenkins, Sentry, Grafana, etc.) to help engineers triag
Duolingo·11d ago
Agent: Open-Source macOS AI Automation Tool with 17 LLM Providers for Code, Apps, and System Control
Agent is an open-source macOS application that provides AI-powered automation and control of Mac systems. It integrates with 17 different LL
github.com·1mo ago