List your company

Logical Inference on LogiQA (TSR, ASR)

76.75Task Success Rate (TSR)G-Safeguard
8.8926.507544.12561.7425May 27, 2026

Evaluation results

MethodLinks
G-SafeguardBackbone=Qwen-Plus, Attack Setting=Independent2026.0576.7516.5
STARBackbone=Qwen-Plus, Attack Setting=Independent2026.057516.5
STARBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0573.516.75
G-SafeguardBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0573.2519
ARGUSBackbone=Qwen-Plus, Attack Setting=Cooperative2026.056429.75
ARGUSBackbone=Qwen-Plus, Attack Setting=Independent2026.0562.2532.5
Blind-GuardBackbone=Qwen-Plus, Attack Setting=Independent2026.0557.7538.25
STARBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0552.534.25
G-SafeguardBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.055238.5
Blind-GuardBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0551.7543.75
STARBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.055133.75
C&IBackbone=Qwen-Plus, Attack Setting=Independent2026.0549.7546.75
G-SafeguardBackbone=GPT-5-Nano, Attack Setting=Independent2026.0547.2544.25
G-SafeguardBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0546.7545
G-SafeguardBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0546.2544.75
C&IBackbone=Qwen-Plus, Attack Setting=Cooperative2026.0546.2549.5
STARBackbone=GPT-5-Nano, Attack Setting=Independent2026.0545.2542
STARBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.054144.5
Blind-GuardBackbone=GPT-5-Nano, Attack Setting=Independent2026.0534.559.25
Blind-GuardBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.0529.2564.5
Blind-GuardBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.0526.565.75
ARGUSBackbone=GPT-5-Nano, Attack Setting=Independent2026.0525.2569.5
C&IBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.052568.5
C&IBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.052471
C&IBackbone=GPT-5-Nano, Attack Setting=Independent2026.0521.2574.5
ARGUSBackbone=GPT-3.5-Turbo, Attack Setting=Cooperative2026.052076.25
ARGUSBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.0516.2579.75
ARGUSBackbone=GPT-3.5-Turbo, Attack Setting=Independent2026.051581.25
Blind-GuardBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.0514.581.75
C&IBackbone=GPT-5-Nano, Attack Setting=Cooperative2026.0511.586.25