Honghao Lin, David P. Woodruff, Yuan Deng, Jieming Mao, Song Zuo, Vahab Mirrokni
Stellar Colosseum is a many-agent harness for long-horizon research in mathematics and theoretical computer science. It allocates inference across research problems and generates candidates in parallel. Colosseum achieves 71.0% accuracy on TCS-Bench using Gemini 3.1 Pro and Gemini 3.7 Flash.
The Colosseum workflow has the potential to accelerate progress in mathematics and theoretical computer science by automating long-horizon research problems. This can lead to breakthroughs in various fields, including artificial intelligence and computer science. The integration with Google Antigravity's Teamwork framework also enables collaborative research.
Thank you to arXiv for use of its open access interoperability. arxiv.org. aibrief.ai is not affiliated with or endorsed by arXiv.