List your company

Math Reasoning on AIME 2024 (Pass@k Suite)

63.3Pass@1Merged (α=0.2)
-2.42814.63631.748.764Oct 4, 2025Nov 4, 2025Dec 6, 2025Jan 7, 2026Feb 7, 2026Mar 11, 2026Apr 12, 2026

Evaluation results

MethodLinks
Merged (α=0.2)Model variant=Merged (α=0.2), Interpolation Weight=0.22026.0463.3————————
Qwen3-4B-Instruct-2507Model variant=Base Model2026.0460————————
SFT No-TestModel variant=SFT No-Test2026.0440————————
THRBase Model=Qwen2.5-Math-7B, p=< 02025.1017.322.628.535.543.251.358.866.373.3
THRBase Model=Qwen2.5-Math-7B2025.1015.721.327.334.743.251.458.463.666.7
GRPOBase Model=Qwen2.5-Math-7B2025.1014.420.727.534.74249.658.167.376.7
THRBase Model=Qwen2.5-Math-7B, p=> 02025.1013.61924.931.839.948.857.364.270
THRBase Model=Qwen2.5-Math-1.5B, p=< 02025.1011.918.224.931.237.945.352.961.270
GRPOBase Model=Qwen2.5-Math-1.5B2025.1011.417.724.330.536.743.4505663.3
THRBase Model=Qwen2.5-Math-1.5B2025.1010.616.723.430.237.244.851.958.563.3
THRBase Model=Qwen2.5-Math-1.5B, p=> 02025.108.413.6202734.743.150.857.663.3
BaseBase Model=Qwen2.5-Math-7B2025.107.513.522324147.953.759.466.7
BaseBase Model=Qwen2.5-Math-1.5B2025.103.36.311.318.527.436.444.349.653.3
GRPOBase Model=Qwen2.5-0.5B-Instruct2025.100.40.81.52.95.41017.227.336.7
THRBase Model=Qwen2.5-0.5B-Instruct2025.100.40.71.52.95.49.715.72226.7
THRBase Model=Qwen2.5-0.5B-Instruct, p=< 02025.100.40.81.52.95.49.414.921.530
THRBase Model=Qwen2.5-0.5B-Instruct, p=> 02025.100.40.71.42.64.78.112.919.930
BaseBase Model=Qwen2.5-0.5B-Instruct2025.100.10.20.40.81.63.15.69.816.7