OpenBioLink/ThoughtSource
Open-source framework providing chain-of-thought reasoning datasets and dataloaders for training and evaluating large language models.

ThoughtSource is a centralized open resource focused on chain-of-thought reasoning data for LLMs. It provides standardized dataloaders that format multiple datasets (commonsense QA, biomedical reasoning, etc.) with reasoning chains in Hugging Face datasets format. The repository also includes workflow tools and a tutorial notebook for working with reasoning data, targeting research into trustworthy and robust LLM reasoning capabilities.
Frequently asked
- What is OpenBioLink/ThoughtSource?
- Open-source framework providing chain-of-thought reasoning datasets and dataloaders for training and evaluating large language models.
- Is ThoughtSource open source?
- Yes — OpenBioLink/ThoughtSource is open source, released under the MIT license.
- What language is ThoughtSource written in?
- OpenBioLink/ThoughtSource is primarily written in Jupyter Notebook.
- How popular is ThoughtSource?
- OpenBioLink/ThoughtSource has 1k stars on GitHub.
- Where can I find ThoughtSource?
- OpenBioLink/ThoughtSource is on GitHub at https://github.com/OpenBioLink/ThoughtSource.