The technical report of our winning system at the #NeurIPS2025 MMU-RAG competition is out!🚀
We built an open, reproducible deep research by combining Qwen3 with a fully open search API and trained it via LLM-as-a-judge–based preference tuning.
arXiv: https://t.co/tO6evvpvHM
Our team has been selected as a 🏆winner🏆 in the #NeurIPS2025 MMU-RAG competition (open-source category)!
We built a reproducible, open deep research system based entirely on open models (Qwen3) and an open web corpus (ClueWeb22).
Our technical report will be available soon!
New dataset release: 🌐FineWiki
This is an updated and better extracted version of Wikipedia, covering 325+ languages.
Unlike the old dataset from 2023, we kept all the math content, tables, properly rendered templates, and extracted key facts.
Examples and highlights below.