Thought-Level Beam Search for Reasoning

Benchmark Model Rank Results
mathematical-reasoning-on-hmmt-feb-2025Gambit (Phi-4-reasoning-plus-14B)#7Accuracy: 75.8
mathematical-reasoning-on-aime-2026Gambit (Phi-4-reasoning-plus-14B)#18Accuracy: 90.0
mathematical-reasoning-on-aime-2025Gambit (Qwen3-4B-Thinking-2507)#23Accuracy: 90.0
multiple-choice-question-answering-mcqa-on-gpqa-diamondGambit (Phi-4-reasoning-plus-14B)#62Accuracy: 77.1