List your company

Scientific Reasoning on MMLU-Redux (ACC, TOK, LAT)

54.64AccuracyInftyThink+
50.199251.352152.50553.6579Feb 6, 2026

Evaluation results

MethodLinks
InftyThink+RL Setting=Task reward only (✓ T), Sampling Temperature=0.7, Number of Samples=322026.0254.644.8130.42
InftyThink+RL Setting=Task and efficiency reward (✓ T+E), Sampling Temperature=0.7, Number of Samples=322026.0253.942.2912.55
InftyThink+RL Setting=Cold start only (✗), Sampling Temperature=0.7, Number of Samples=322026.0251.343.1217.61
VanillaRL Setting=Task reward only (✓ T), Sampling Temperature=0.7, Number of Samples=322026.0251.15.8569.27
VanillaRL Setting=Cold start only (✗), Sampling Temperature=0.7, Number of Samples=322026.0250.373.4528.54