LLM-AutoSciLab : Closed-Loop Scientific Discovery via Active Experimentation with LLMs
Sanchit Kabra, Nikhil Abhyankar, Saaketh Desai, Prasad P. Iyer, Chandan K. Reddy, [0.5em] Virginia Tech Sandia National Laboratories · 2026
Read LLM-AutoSciLab : Closed-Loop Scientific Discovery via Active Experimentation with LLMs by Sanchit Kabra, Nikhil Abhyankar, Saaketh Desai, Prasad P. Iyer, Chandan K. Reddy, [0.5em] Virginia Tech Sandia National Laboratories (2026) 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-2605-24043.txt.
Description
Kabra and four co-authors close the loop on discovery, pairing idea generation with test selection. Their system, LLM-AutoSciLab, offers ideas, picks telling tests to split or refine them, and updates its state from the evidence, rather than fitting fixed data sets. To test live work they build ActiveSciBench, with 57 enzyme tasks and 45 gene-network tasks under test budgets. Across NewtonBench and both ActiveSciBench sets, it beats prior methods, and guided tests prove two to five times more sample-thrifty than the rivals.
Catalog record
| CREATOR | Sanchit Kabra, Nikhil Abhyankar, Saaketh Desai, Prasad P. Iyer, Chandan K. Reddy, [0.5em] Virginia Tech Sandia National Laboratories |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 84066 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.24043