The AI Scientist-v2 : Workshop-Level Automated Scientific Discovery via Agentic Tree Search
Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, David Ha · 2025
Read The AI Scientist-v2 : Workshop-Level Automated Scientific Discovery via Agentic Tree Search by Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, David Ha (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08066.txt.
Description
Yutaro Yamada and co-authors describe a system that can carry out research on its own, in work from 2025. The system, called the AI Scientist-v2, forms ideas, runs tests, studies the data, draws charts, and writes up the results. No human-written code templates guide it, unlike the first version.
A search method that grows step by step, led by a manager agent, steers the work. A vision model also gives feedback on charts so they can be improved. The team sent three fully self-made papers to a peer-reviewed workshop. One scored above the mean bar for human papers, a first for a fully AI-made paper. The code is shared in the open.
Catalog record
| CREATOR | Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, David Ha |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 115 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-08066