List your company

Mathematical Reasoning on SAT Math, AMC23, AIME24, and Olympiad Bench Average (out-domain)

65.77Average AccuracyQwen3-4B
61.03862.266563.49564.7235Feb 13, 2026

Evaluation results

MethodLinks
Qwen3-4BBase Model=Qwen3-4B2026.0265.772,472.800
ThinkPrune-500Base Model=Qwen3-4B2026.0264.962,109.80.090.02
ThinkPrune-1kBase Model=Qwen3-4B2026.0264.462,006.90.09-0.01
ACPOBase Model=Qwen3-4B2026.0263.121,675.60.12-0.08
CRTBase Model=Qwen3-4B2026.0262.991,361.40.240.03
O1-Pruner-4.0Base Model=Qwen3-4B2026.0262.661,394.30.2-0.04
ShorterBetterBase Model=Qwen3-4B2026.0261.221,343.40.11-0.23