List your company

Open-ended arithmetic reasoning on MATH (test)

96.9TDRCritical-CoT
84.10887.42990.7594.071Apr 12, 2026

Evaluation results

MethodLinks
Critical-CoTModel=LLaMA-2-13B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0496.998.638.1780
Critical-CoTModel=Qwen3-14B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0496.897.992.8683
Critical-CoTModel=LLaMA-2-13B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0495.9—32.44220
Critical-CoTModel=GPT-OSS-20B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0492.793.891.7689
Critical-CoTModel=Qwen3-14B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0492.7—77.1106
Critical-CoTModel=GPT-OSS-20B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0484.6—79.6650