List your company

General Knowledge Question Answering on MMLU (Adversarial Evaluation)

86Task Success Rate (TSR)STAR
39.251.3563.575.65May 27, 2026

Evaluation results

MethodLinks
STARBackbone=Qwen-Plus, Attack Setting=Independent2026.05868
STARBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0585.757.25
G-SafeguardBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0584.7511.25
G-SafeguardBackbone=Qwen-Plus, Attack Setting=Independent2026.0582.2512.75
ARGUSBackbone=Qwen-Plus, Attack Setting=Independent2026.0581.515
STARBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.058012.25
ARGUSBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0579.2516.5
Blind-GuardBackbone=Qwen-Plus, Attack Setting=Independent2026.0578.2517
STARBackbone=GPT-5-Nano, Attack Setting=Independent2026.0577.7515
STARBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.057711.5
C&IBackbone=Qwen-Plus, Attack Setting=Independent2026.0576.7520.5
STARBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0576.7511.25
Blind-GuardBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0575.2520.5
C&IBackbone=Qwen-Plus, Attack Setting=Cooperative2026.057422.5
G-SafeguardBackbone=GPT-5-Nano, Attack Setting=Independent2026.0573.521.5
G-SafeguardBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.0570.2525.25
Blind-GuardBackbone=GPT-5-Nano, Attack Setting=Independent2026.0567.2526.75
ARGUSBackbone=GPT-5-Nano, Attack Setting=Independent2026.0562.2534.5
Blind-GuardBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0561.532
C&IBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0559.7533.5
Blind-GuardBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0559.2534
G-SafeguardBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.055935.75
C&IBackbone=GPT-5-Nano, Attack Setting=Independent2026.055838.25
C&IBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0557.2536.25
ARGUSBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0556.7540.25
ARGUSBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0555.541.5
G-SafeguardBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0554.7541.25
ARGUSBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.0551.546.25
Blind-GuardBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.054550
C&IBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.054157