Stellar Colosseum is a model-agnostic harness designed to allocate inference across long-horizon research in mathematics and theoretical computer science, addressing the unreliability of language models on complex problems. The system explores alternative strategies before proof construction, uses a readiness gate to determine when a route is mature enough for decomposition, and represents the proof plan as interdependent section-level subproblems.

  • Colosseum generates candidates in parallel, attacks them with targeted falsification, and combines results through overlapping random-sample tree aggregation.
  • The workflow is integrated into Google Antigravity's Teamwork framework as the Long Proof pattern.
  • On TCS-Bench, a benchmark of research-level theorem-proving tasks from FOCS, STOC, and SODA papers, Colosseum achieves 71.0% accuracy using Gemini 3.1 Pro and Gemini 3.7 Flash.
  • In a Codeforces evaluation with Gemini 3.1 Pro, the proof-oriented pipeline with execution feedback solves 218 of 222 problems.

The authors demonstrate that this approach addresses open problems from top venues like FOCS and JMLR through open-ended research and benchmark evaluations.