List your company

Open-ended arithmetic reasoning on GSM8K (test)

99.3TDRCritical-CoT
78.18883.66989.1594.631Apr 12, 2026

Evaluation results

MethodLinks
Critical-CoTModel=Qwen3-14B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0499.3—91.7590
Critical-CoTModel=GPT-OSS-20B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0498.998.992.24212
Critical-CoTModel=GPT-OSS-20B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0496.9—72.4610
Critical-CoTModel=LLaMA-2-13B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0494.997.941.8580
Critical-CoTModel=Qwen3-14B, Backdoor Attack Type=In-Context Learning-Based Backdoor2026.0493.999.495.9510
Critical-CoTModel=LLaMA-2-13B, Backdoor Attack Type=Fine-Tuning-Based Backdoor2026.0479—38.2320