List your company

Language Understanding on MMLU Low tier

66.7Reward+RL Free
39.24446.37253.560.628Jun 5, 2026

Evaluation results

MethodLinks
+RL FreeBackbone=Qwen3-4B, RL strategy=Free2026.0666.765
+RL GateBackbone=Qwen3-4B, RL strategy=Gate, Protocol=confidence-gated GSPO2026.0661.551
+RL FlatBackbone=Qwen3-4B, RL strategy=Flat2026.0641.91
+RL OTCBackbone=Qwen3-4B, RL strategy=OTC2026.0641.22
BaselineBackbone=Qwen3-4B2026.0640.32