Utility Evaluation on AgentDojo
78.4UtilityGLM-4.5
Evaluation Results
| Method | Links | |
|---|---|---|
| GLM-4.5Prompt=InjecPrompt, Technique=default2025.09 | 78.4 | |
| GLM-4.5Prompt=Multi-turn, Technique=default2025.09 | 75.8 | |
| Kimi-K2Prompt=Multi-turn, Technique=default2025.09 | 72 | |
| Kimi-K2Prompt=InjecPrompt, Technique=default2025.09 | 71.5 | |
| Kimi-K2Prompt=Multi-turn, Technique=ChatInject2025.09 | 69.9 | |
| GLM-4.5Prompt=InjecPrompt, Technique=ChatInject + tool2025.09 | 68.1 | |
| GLM-4.5Prompt=InjecPrompt, Technique=ChatInject2025.09 | 67.9 | |
| GLM-4.5Prompt=Multi-turn, Technique=ChatInject2025.09 | 67.9 | |
| GLM-4.5Prompt=InjecPrompt, Technique=ChatInject + think2025.09 | 65.7 | |
| Qwen-3Prompt=Multi-turn, Technique=default2025.09 | 52.4 | |
| Qwen-3Prompt=InjecPrompt, Technique=default2025.09 | 50.9 | |
| Grok-2Prompt=InjecPrompt, Technique=default2025.09 | 41.7 | |
| GPT-ossPrompt=Multi-turn, Technique=default2025.09 | 38.3 | |
| Kimi-K2Prompt=InjecPrompt, Technique=ChatInject + tool2025.09 | 35.2 | |
| Kimi-K2Prompt=InjecPrompt, Technique=ChatInject2025.09 | 35 | |
| Grok-2Prompt=Multi-turn, Technique=default2025.09 | 33.9 | |
| Grok-2Prompt=Multi-turn, Technique=ChatInject2025.09 | 31.9 | |
| Grok-2Prompt=InjecPrompt, Technique=ChatInject2025.09 | 29.8 | |
| Qwen-3Prompt=InjecPrompt, Technique=ChatInject2025.09 | 28.3 | |
| Qwen-3Prompt=Multi-turn, Technique=ChatInject2025.09 | 27.5 | |
| Qwen-3Prompt=InjecPrompt, Technique=ChatInject + think2025.09 | 24.4 | |
| Qwen-3Prompt=InjecPrompt, Technique=ChatInject + tool2025.09 | 22.9 | |
| GPT-ossPrompt=InjecPrompt, Technique=default2025.09 | 19.6 | |
| GPT-ossPrompt=InjecPrompt, Technique=ChatInject2025.09 | 18.8 | |
| Llama-4Prompt=Multi-turn, Technique=default2025.09 | 18.5 | |
| Llama-4Prompt=InjecPrompt, Technique=default2025.09 | 16.5 | |
| Llama-4Prompt=Multi-turn, Technique=ChatInject2025.09 | 16.2 | |
| Llama-4Prompt=InjecPrompt, Technique=ChatInject2025.09 | 15.9 | |
| Llama-4Prompt=InjecPrompt, Technique=ChatInject + tool2025.09 | 14.7 | |
| GPT-ossPrompt=InjecPrompt, Technique=ChatInject + think2025.09 | 11.1 | |
| GPT-ossPrompt=InjecPrompt, Technique=ChatInject + tool2025.09 | 9 | |
| GPT-ossPrompt=Multi-turn, Technique=ChatInject2025.09 | 8 |