OmniScience: A Domain-Specialized LLM for Scientific Reasoning and Discovery
Vignesh Prabhakar, Md Amirul Islam, Adam Atanas, Yao-Ting Wang, Joah Han, Aastha Jhunjhunwala, Rucha Apte, Robert Clark, Kang Xu, Zihan Wang, Kai Liu · 2025
Read OmniScience: A Domain-Specialized LLM for Scientific Reasoning and Discovery by Vignesh Prabhakar, Md Amirul Islam, Adam Atanas, Yao-Ting Wang, Joah Han, Aastha Jhunjhunwala, Rucha Apte, Robert Clark, Kang Xu, Zihan Wang, Kai Liu (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-2503-17604.txt.
Description
A team of eleven authors presents OmniScience, a language model built for science, in this 2025 paper. General models can reason, but a model steeped in science text may do better on hard research questions and lab tasks.
The model was built in three steps: extra pretraining on a curated science corpus, tuning on science task data, and distillation of reasoning skill from a teacher. To show range, the team built a battery agent that ranks molecules as candidate solvents or additives. On a hard science question set and battery tests, the model matched top large models and beat public models of like size. Tests confirmed both extra training steps mattered.
Catalog record
| CREATOR | Vignesh Prabhakar, Md Amirul Islam, Adam Atanas, Yao-Ting Wang, Joah Han, Aastha Jhunjhunwala, Rucha Apte, Robert Clark, Kang Xu, Zihan Wang, Kai Liu |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17604