List your company

Adversarial Robustness Evaluation on MMLU (TSR, ASR)

75.25Task Success Rate (TSR)Qwen-Plus
27.6740.022552.37564.7275May 27, 2026

Evaluation results

MethodLinks
Qwen-PlusAttack Scenario=Group Collusive2026.0575.2520.75
Qwen-PlusAttack Scenario=Independent2026.0573.7521.5
Qwen-PlusAttack Scenario=Cooperative2026.057224.25
GPT-5-NanoAttack Scenario=Independent2026.0553.543.75
GPT-5-NanoAttack Scenario=Group Collusive2026.0546.2552
GPT-5-NanoAttack Scenario=Cooperative2026.0539.2559
GPT-3.5-TurboAttack Scenario=Independent2026.053266
GPT-3.5-TurboAttack Scenario=Group Collusive2026.0531.2567.5
GPT-3.5-TurboAttack Scenario=Cooperative2026.0529.569