The authors introduce Stellar Colosseum, a model-agnostic harness designed to allocate inference across long-horizon research in mathematics and theoretical computer science. 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. Integrated into Google Antigravity's Teamwork framework as the Long Proof pattern, it achieved 71.0% accuracy on TCS-Bench using Gemini 3.1 Pro and Gemini 3.7 Flash. Additionally, a proof-oriented pipeline with execution feedback solved 218 of 222 problems in a Codeforces evaluation.
The framework addresses the unreliability of language models on long-horizon research problems by managing sequences of uncertain and interdependent decisions.