Tool Use on ToolBench All v1.0
97Tool Call Rate (TCR)NaviAgent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NaviAgentModel=Deepseek-V32025.06 | 97 | 55.2 | 4.6 | |
| NaviAgentModel=Qwen2.5-32B2025.06 | 83.2 | 45.4 | 4.66 | |
| α-UMIModel=Deepseek-V32025.06 | 82.9 | 40.7 | 6.06 | |
| α-UMIModel=Qwen2.5-32B2025.06 | 78.3 | 32.8 | 5.94 | |
| α-UMIModel=Qwen2.5-14B2025.06 | 75.5 | 26.9 | 5.74 | |
| NaviAgentModel=Qwen2.5-14B2025.06 | 61.6 | 35.8 | 4.38 | |
| ReActModel=Deepseek-V32025.06 | 52.5 | 34.5 | 3.54 | |
| ToolLLMModel=Deepseek-V32025.06 | 51.3 | 34.4 | 3.86 | |
| ToolLLMModel=Qwen2.5-14B2025.06 | 51 | 21.3 | 4.03 | |
| ToolLLMModel=Qwen2.5-32B2025.06 | 46.2 | 29.3 | 3.83 | |
| ReActModel=Qwen2.5-32B2025.06 | 33.6 | 23.4 | 3.63 | |
| ReActModel=Qwen2.5-14B2025.06 | 27.1 | 20.6 | 3.61 |