List your company

Mathematical Deduction on AIME25 (test)

71.11AccuracyInjectRLOpt
54.698858.959463.2267.4806Feb 12, 2026

Evaluation results

MethodLinks
InjectRLOptbackbone=Qwen3-32B2026.0271.11
InjectCorrectbackbone=Qwen3-32B2026.0270
Baselinebackbone=Qwen3-32B2026.0269.33
InjectRLOptbackbone=Qwen3-14B2026.0267.33
InjectRLOptbackbone=Qwen3-8B2026.0266
InjectCorrectbackbone=Qwen3-14B2026.0264.67
InjectRLOptbackbone=Qwen3-4B2026.0264
InjectCorrectbackbone=Qwen3-8B2026.0261.33
Baselinebackbone=Qwen3-14B2026.0259.33
Baselinebackbone=Qwen3-8B2026.0258.67
InjectCorrectbackbone=Qwen3-4B2026.0257.33
Baselinebackbone=Qwen3-4B2026.0255.33