List your company

Model Calibration on MATH, GSM8K, SelfAware, and TruthfulQA combined

0.086ECECARE-GRPO
0.08340.100950.11850.13605Jan 22, 2026

Evaluation results

MethodLinks
CARE-GRPOBackbone=Qwen2.5-7B2026.010.086
RKL-GSPOBackbone=Qwen2.5-7B2026.010.088
Base ModelBackbone=Qwen2.5-7B2026.010.089
RKL-DAPOBackbone=Qwen2.5-7B2026.010.09
RKL-GRPOBackbone=Qwen2.5-7B2026.010.095
CARE-DAPOBackbone=Qwen2.5-7B2026.010.099
CARE-GSPOBackbone=Qwen2.5-7B2026.010.101
GRPO (No Constraint)Backbone=Qwen2.5-7B2026.010.145
GSPO (No Constraint)Backbone=Qwen2.5-7B2026.010.149
DAPO (No Constraint)Backbone=Qwen2.5-7B2026.010.151