List your company

STEM Reasoning on GaokaoCloze

25.9AccuracyAGPO
18.51620.43322.3524.267May 20, 2026

Evaluation results

MethodLinks
AGPOclipping=adaptive, ATS=true2026.0525.9
AGPOclipping=adaptive, ATS=false2026.0521.2
GRPOATS=true2026.0520.5
GRPOclipping=fixed ε2026.0520.3
Adaptive-KL PPO2026.0519.6
PPO2026.0518.9
DPO2026.0518.8