List your company

Answer Accuracy on Alpaca

40.6BRT AccuracyNo Attack
31.65633.97836.338.622Oct 22, 2025

Evaluation results

MethodLinks
No AttackStructure=Hierarchy, Backbone=DeepSeek-V32025.1040.699.2
BPDStructure=Hierarchy, Model=GPT-4o2025.1040.298.8
No AttackStructure=Hierarchy, Model=GPT-4o2025.104099.4
No AttackStructure=Flat, Backbone=DeepSeek-V32025.1039.999.2
Single LLMStructure=Single LLM, Backbone=DeepSeek-V32025.1039.897.2
BPDStructure=Hierarchy, Backbone=DeepSeek-V32025.1039.898.8
No AttackStructure=Flat, Model=GPT-4o2025.1039.899.4
BPDStructure=Flat, Model=GPT-4o2025.1039.699.6
BPDStructure=Flat, Backbone=DeepSeek-V32025.1039.499
Single LLMStructure=Single LLM, Model=GPT-4o2025.1039.497.2
ChallengerStructure=Hierarchy, Backbone=DeepSeek-V32025.1038.995
G-SafeguardStructure=Flat, Model=GPT-4o2025.1038.697
G-SafeguardStructure=Hierarchy, Model=GPT-4o2025.1038.497.2
G-SafeguardStructure=Flat, Backbone=DeepSeek-V32025.1038.297
G-SafeguardStructure=Hierarchy, Backbone=DeepSeek-V32025.1038.296.6
AGENTXPOSEDStructure=Hierarchy, Model=GPT-4o2025.1037.996.2
AGENTXPOSEDStructure=Hierarchy, Backbone=DeepSeek-V32025.1037.695.4
AGENTXPOSEDStructure=Flat, Backbone=DeepSeek-V32025.1037.595.2
InspectorStructure=Flat, Backbone=DeepSeek-V32025.1037.291.4
AGENTXPOSEDStructure=Flat, Model=GPT-4o2025.103795.2
InspectorStructure=Hierarchy, Backbone=DeepSeek-V32025.1035.792.6
ChallengerStructure=Flat, Backbone=DeepSeek-V32025.103593.6
AttackStructure=Hierarchy, Backbone=DeepSeek-V32025.1033.692.6
AttackStructure=Flat, Backbone=DeepSeek-V32025.1033.487
AttackStructure=Flat, Model=GPT-4o2025.103286.8
AttackStructure=Hierarchy, Model=GPT-4o2025.103292.6