List your company

Code Reasoning on HumanEval base and extended (out-of-distribution)

0.677AccuracyInftyThink+
0.5660320.5948410.623650.652459Feb 6, 2026

Evaluation results

MethodLinks
InftyThink+RL_Setting=Task reward only (✓ T), Sampling_Times=32, Temperature=0.72026.020.6770.6268.2142.22
InftyThink+RL_Setting=Task and efficiency reward (✓ T+E), Sampling_Times=32, Temperature=0.72026.020.67420.61914.6623.9
VanillaRL_Setting=Task reward only (✓ T), Sampling_Times=32, Temperature=0.72026.020.60440.56278.1790.1
InftyThink+RL_Setting=Cold start only (✗), Sampling_Times=32, Temperature=0.72026.020.59030.54345.0227.5
VanillaRL_Setting=Cold start only (✗), Sampling_Times=32, Temperature=0.72026.020.57030.52526.5865.89