List your company

Prompt Injection on MMLU

12ASR@3PropGuard
10.2821.8933.545.11Aug 11, 2025Sep 25, 2025Nov 9, 2025Dec 24, 2025Feb 7, 2026Mar 24, 2026May 8, 2026

Evaluation results

MethodLinks
PropGuardArchitecture=Tree2026.0512—91.7
PropGuardArchitecture=Chain2026.0512.3—90
PropGuardArchitecture=Random2026.0512.3—88.3
PropGuardArchitecture=Star2026.0514.7—86.7
INFA-GUARDBackbone=GPT-4o-mini, Communication Iterations=3, Topology Type=Random2026.0115—85
INFA-GuardArchitecture=Chain2026.0515.3—86.7
G-SafeguardArchitecture=Chain2026.0516.7—83.3
INFA-GuardArchitecture=Star2026.0516.7—85
G-SafeguardTopology=Chain2025.12170.9822—
G-SafeguardTopology=Star, Backbone LLM=GPT-4o-mini2025.0817.330.9867—
G-SafeguardArchitecture=Tree2026.0517.7—83.3
INFA-GuardArchitecture=Tree2026.0517.7—83.3
INFA-GuardArchitecture=Random2026.0517.7—81.7
BlindGuardTopology=Tree2025.12180.8133—
G-SafeguardTopology=Star2025.12180.9911—
G-SafeguardArchitecture=Star2026.0518—81.7
XG-GuardTopology=Chain2025.1218.330.9511—
G-SafeguardTopology=Tree2025.1218.330.9911—
G-SafeguardTopology=Tree, Backbone LLM=GPT-4o-mini2025.0818.330.9867—
G-SafeguardArchitecture=Random2026.0518.7—81.7
G-SafeguardTopology=Random, Backbone LLM=GPT-4o-mini2025.08190.9911—
InspectorBackbone=GPT-4o-mini2026.0119.2—62.7
G-SafeguardTopology=Random2025.1219.330.9956—
G-SafeguardTopology=Chain, Backbone LLM=GPT-4o-mini2025.0819.330.9911—
BlindGuardTopology=Chain, Backbone LLM=GPT-4o-mini2025.0819.330.8578—
DOMINANTTopology=Tree, Backbone LLM=GPT-4o-mini2025.0819.330.6133—
INFA-GUARDBackbone=Qwen3-235B-A22B2026.0120—81.7
BlindGuardTopology=Chain2025.1220.330.84—
BlindGuardTopology=Star, Backbone LLM=GPT-4o-mini2025.0820.330.8622—
XG-GuardTopology=Tree2025.1220.670.92—
BlindGuardTopology=Tree, Backbone LLM=GPT-4o-mini2025.0820.670.8356—
GUARDIANArchitecture=Chain2026.0520.7—81.7
TAMTopology=Tree2025.12210.5556—
XG-GuardTopology=Random2025.1221.330.9289—
TAMTopology=Chain2025.12220.5378—
XG-GuardTopology=Star2025.12220.9289—
TAMTopology=Tree, Backbone LLM=GPT-4o-mini2025.08220.5556—
GUARDIANArchitecture=Star2026.0522—80
GUARDIANArchitecture=Tree2026.0522.3—80
BlindGuardTopology=Random2025.1222.330.8444—
DOMINANTTopology=Tree2025.1222.670.5867—
GUARDIANArchitecture=Random2026.0523—78.3
DOMINANTTopology=Chain2025.1223.330.5467—
PREMTopology=Chain2025.1223.670.4533—
TAMTopology=Chain, Backbone LLM=GPT-4o-mini2025.0823.670.5289—
PREMTopology=Tree, Backbone LLM=GPT-4o-mini2025.08240.4444—
PREMTopology=Tree2025.1224.330.4044—
DOMINANTTopology=Star2025.1224.330.5956—
DOMINANTTopology=Chain, Backbone LLM=GPT-4o-mini2025.0824.670.5956—
PREMTopology=Chain, Backbone LLM=GPT-4o-mini2025.08250.4489—
BlindGuardTopology=Star2025.12260.8311—
DOMINANTTopology=Star, Backbone LLM=GPT-4o-mini2025.0826.670.6267—
BlindGuardTopology=Random, Backbone LLM=GPT-4o-mini2025.0826.670.8533—
TAMTopology=Star2025.12270.6222—
No DefenseTopology=Tree2025.1227.67——
TAMTopology=Star, Backbone LLM=GPT-4o-mini2025.0828.330.6444—
DOMINANTTopology=Random, Backbone LLM=GPT-4o-mini2025.08290.5733—
PREMTopology=Star, Backbone LLM=GPT-4o-mini2025.08310.4311—
DOMINANTTopology=Random2025.1231.330.5867—
No DefenseTopology=Tree, Backbone LLM=GPT-4o-mini2025.0832——
PREMTopology=Star2025.12330.4622—
No DefenseTopology=Chain2025.1234.33——
TAMTopology=Random2025.1234.670.4578—
InspectorArchitecture=Star2026.0535.7—66.7
ChallengerArchitecture=Tree2026.0536.7—70
PREMTopology=Random, Backbone LLM=GPT-4o-mini2025.08370.4889—
ChallengerArchitecture=Chain2026.0537—66.7
AgentXposedArchitecture=Tree2026.0538—63.3
No DefenseTopology=Chain, Backbone LLM=GPT-4o-mini2025.0838.33——
TAMTopology=Random, Backbone LLM=GPT-4o-mini2025.0838.330.4933—
No DefenseTopology=Star, Backbone LLM=GPT-4o-mini2025.0839——
InspectorArchitecture=Chain2026.0539—61.7
InspectorArchitecture=Random2026.0539—63.3
AgentXposedArchitecture=Chain2026.0539.3—61.7
PREMTopology=Random2025.1239.670.4667—
AgentSafeArchitecture=Chain2026.0539.7—68.3
No DefenseTopology=Star2025.1240.66——
No DefenseArchitecture=Chain2026.0540.7—60
No DefenseTopology=Random, Backbone LLM=GPT-4o-mini2025.0841——
InspectorArchitecture=Tree2026.0541—61.7
No DefenseArchitecture=Tree2026.0542.3—61.7
ChallengerArchitecture=Star2026.0544—56.7
No DefenseArchitecture=Random2026.0544.3—56.7
No DefenseTopology=Random2025.1244.67——
AgentSafeArchitecture=Tree2026.0546.7—58.3
AgentSafeArchitecture=Star2026.0547—51.7
AgentXposedArchitecture=Random2026.0548.3—55
ChallengerArchitecture=Random2026.0548.7—50
AgentXposedArchitecture=Star2026.0552—53.3
AgentSafeArchitecture=Random2026.0552—51.7
No DefenseArchitecture=Star2026.0555—46.7