AI Scientist v2 Produces First AI-Generated Paper Accepted at Peer-Reviewed Workshop

AI Scientist v2 Produces First AI-Generated Paper Accepted at Peer-Reviewed Workshop

Sakana AI published AI Scientist v2, an end-to-end system for automated scientific discovery. The key milestone: a paper fully generated by the system was accepted at a peer-reviewed machine learning workshop.

How It Works

AI Scientist v2 follows the full research pipeline. It formulates hypotheses, designs experiments, executes them, analyzes and visualizes the results, and writes a complete manuscript. A dedicated experiment manager agent coordinates the process using a progressive agentic tree search methodology.

Key Improvements Over v1

The original AI Scientist required human-authored code templates. Version 2 eliminates this dependency. It generalizes across diverse machine learning domains without needing researchers to provide starting code. The system also uses a vision-language model feedback loop to refine figures and visual content in the paper.

The Peer Review Milestone

This is the first time an AI-generated paper has passed peer review at a recognized venue. While it was a workshop paper (not a full conference paper), it demonstrates that AI systems can produce research that meets minimum quality standards set by human reviewers.

Open Source

The full system is available on GitHub. Researchers can use it to accelerate their own work or study how automated discovery systems behave.

Why It Matters

Automated scientific discovery has been a long-term goal for AI research. AI Scientist v2 shows that the gap between AI-assisted and AI-generated research is narrowing. The implications extend beyond machine learning to any field where experiments can be run programmatically.