List your company

Mathematical Reasoning on GSM8K (Accuracy, Exit Position)

78.2AccuracyBackbone
27.34440.54753.7566.953Apr 20, 2026

Evaluation results

MethodLinks
BackboneBackbone Model=Llama3.1 8B (32 layers)2026.0478.2—
River-LLMBackbone Model=Llama3.1 8B (32 layers), Threshold (τ)=0.72026.0475.626.98
River-LLMBackbone Model=Llama3.1 8B (32 layers), Threshold (τ)=0.52026.0474.42.96
BackboneBackbone Model=Llama3.2 1B (16 layers)2026.0433.5—
River-LLMBackbone Model=Llama3.2 1B (16 layers), Threshold (τ)=0.72026.0433.515.05
River-LLMBackbone Model=Llama3.2 1B (16 layers), Threshold (τ)=0.52026.0429.33.79