List your company

Prompt Injection on MMLU (random topology)

15.5ASR (k=1)Inspector
14.6220.5626.532.44Jan 21, 2026

Evaluation results

MethodLinks
InspectorBackbone=GPT-4o-mini, Topology=random2026.0115.516.919.264.364.362.7——
INFA-GUARDBackbone=GPT-4o-mini, Topology=random2026.0116.3151583.38585——
G-safeguardBackbone=GPT-4o-mini, Topology=random2026.0116.417.117.583.683.681.7——
ChallengerBackbone=GPT-4o-mini, Topology=random2026.0116.536.244.268.348.350——
INFA-GUARDBackbone=Qwen3-235B-A22B, Topology=random2026.0117.718.32083.381.781.7——
G-safeguardBackbone=Qwen3-235B-A22B, Topology=random2026.0121.42322.78078.378.3——
AgentSafeBackbone=GPT-4o-mini, Topology=random2026.0124.533.235.476.766.761.7——
AgentXposed-GuideBackbone=GPT-4o-mini, Topology=random2026.0127.737.34273.361.756.7——
AgentXposed-KickBackbone=GPT-4o-mini, Topology=random2026.012838.740.773.361.760——
No DefenseBackbone=GPT-4o-mini, Topology=random2026.012937.74171.763.360——
No DefenseBackbone=Qwen3-235B-A22B, Topology=random2026.0133.849.358.368.353.338.3——
AgentXposed-GuideBackbone=Qwen3-235B-A22B, Topology=random2026.0134.34756.77053.346.7——
AgentXposed-KickBackbone=Qwen3-235B-A22B, Topology=random2026.0134.352.56066.74541.7——
InspectorBackbone=Qwen3-235B-A22B, Topology=random2026.0134.843.350.761.75344——
ChallengerBackbone=Qwen3-235B-A22B, Topology=random2026.0136.152616546.735——
AgentSafeBackbone=Qwen3-235B-A22B, Topology=random2026.0137.555.564.2654038.3——
LlamaGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————38.763.3
No DefenseBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————44.356.7
PropGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————12.388.3
Qwen3GuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————2278.3
ThinkGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————28.375
WildGuardBase Model=GPT-4o-mini, Communication Architecture=random2026.05——————33.766.7