Introducing The AI Scientist-v2, which produced the 1st fully AI-generated paper to pass peer review at a workshop level (at #ICLR2025) ‼️
Tech Report: https://pub.sakana.ai/ai-scientist-v2/paper/
GitHub: https://github.com/SakanaAI/AI-Scientist-v2
This work is a proud collaboration between Sakana AI, UBC, and Oxford University.
Tech Report: https://pub.sakana.ai/ai-scientist-v2/paper/
GitHub: https://github.com/SakanaAI/AI-Scientist-v2
This work is a proud collaboration between Sakana AI, UBC, and Oxford University.
Comments
The AI-Scientist-v2 adds:
🧠 Agentic Tree Search (open-ended idea exploration)
💡 More Generality (reduced template reliance)
👀 VLM Reviewer (better figures and understanding)
⚡ Parallel Execution
https://github.com/SakanaAI/AI-Scientist-v2
This “Agentic Tree Search” enables deeper, more systematic exploration of scientific hypotheses by harnessing principles from the field of open-endedness:
We submitted three v2-generated papers to an #ICLR2025 workshop for standard double-blind peer review. One passed the acceptance threshold. The papers were 100% AI-generated (idea, code, experiments, analysis, writing)! 🤖⚡💡
Blog: https://sakana.ai/ai-scientist-first-publication/
https://bsky.app/profile/sakanaai.bsky.social/post/3lk5kxcndmc26
This experiment had full cooperation from #ICLR2025 leadership, workshop, IRB approval from UBC. As pre-agreed, the accepted paper was withdrawn post-review to allow community discussion before publishing AI-only work.
Read more: https://sakana.ai/ai-scientist-first-publication/#importance-of-transparency-and-ethical-code-of-conduct
The project is open-sourced on GitHub: https://github.com/SakanaAI/AI-Scientist-v2
What's the future of AI in science?
We would love to hear your thoughts❗