List your company

Mathematical Reasoning on GSM8K (Robustness/Adversarial Evaluation)

72.5Clean AccuracyAGLS
30.941.752.563.3May 26, 2025

Evaluation results

MethodLinks
AGLSModel=llama3.3-70B2025.0572.517.575.869.45
AGLSModel=gemma2-2B2025.05505901.49
AGLSModel=llama3.1-70B2025.0547.51568.423.44
AGLSModel=qwen2.5-0.5B2025.0532.52.592.311.84