zjunlp/OmniThink
OmniThink is a machine writing framework that uses iterative deep thinking and retrieval-augmented generation with large language models to produce knowledge-rich reports.
Not currently ranked — collecting fresh signals.
star history
The project implements a thinking-based approach to machine writing, where language models expand knowledge boundaries through iterative reflection and search. It integrates with various LLMs including deepseek-reasoner, qwen, and GPT models, and supports retrieval-augmented generation using RAGFlow for local document search. The system evaluates machine-written outputs using methods from the associated EMNLP 2025 paper.
Frequently asked
- What is zjunlp/OmniThink?
- OmniThink is a machine writing framework that uses iterative deep thinking and retrieval-augmented generation with large language models to produce knowledge-rich reports.
- Is OmniThink open source?
- Yes — zjunlp/OmniThink is open source, released under the MIT license.
- What language is OmniThink written in?
- zjunlp/OmniThink is primarily written in Python.
- How popular is OmniThink?
- zjunlp/OmniThink has 485 stars on GitHub.
- Where can I find OmniThink?
- zjunlp/OmniThink is on GitHub at https://github.com/zjunlp/OmniThink.