Tool-selection bias evaluation on Tool-selection bias benchmark
0.249δAPI BiasDeepSeek-V3.2-Exp
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeepSeek-V3.2-Exp2025.09 | 0.249 | 0.504 | 0.377 | |
| ToolLLaMA2025.09 | 0.277 | 0.391 | 0.334 | |
| GPT-3.5-turbo2025.09 | 0.32 | 0.336 | 0.328 | |
| Qwen3 235B2025.09 | 0.33 | 0.168 | 0.249 | |
| GPT-4.1 mini2025.09 | 0.331 | 0.423 | 0.377 | |
| Gemini 2.5 Flash2025.09 | 0.365 | 0.306 | 0.335 | |
| Claude 3.5 Sonnet2025.09 | 0.37 | 0.325 | 0.347 |