Tool-calling on Stable Toolbench I1-Tool
75.95Pass RateParaTool
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ParaToolBackbone=Qwen2.5-7B-Instruct2026.05 | 75.95 | 82.28 | |
| Context+Docs & ExamplesBackbone=Qwen2.5-7B-Instruct2026.05 | 68.35 | 69.62 | |
| Parameter+ReinvokeBackbone=Qwen2.5-7B-Instruct2026.05 | 67.72 | 70.25 | |
| ParaToolBackbone=Llama3.1-8B-Instruct2026.05 | 67.09 | 69.62 | |
| ToolLLaMABackbone=Qwen2.5-7B-Instruct2026.05 | 65.82 | 71.52 | |
| Parameter+ReinvokeBackbone=Llama3.1-8B-Instruct2026.05 | 63.92 | 66.46 | |
| Context+Docs & ExamplesBackbone=Llama3.1-8B-Instruct2026.05 | 62.66 | 60.76 | |
| ToolLLaMABackbone=Llama3.1-8B-Instruct2026.05 | 61.11 | 64.55 | |
| Parameter+ToolRetrieverBackbone=Llama3.1-8B-Instruct2026.05 | 59.49 | 61.39 | |
| Parameter+ToolRetrieverBackbone=Qwen2.5-7B-Instruct2026.05 | 57.59 | 60.13 | |
| Context+DocsBackbone=Qwen2.5-7B-Instruct2026.05 | 55.09 | — | |
| Parameter+EmbsimBackbone=Llama3.1-8B-Instruct2026.05 | 55.06 | 55.7 | |
| Parameter+EmbsimBackbone=Qwen2.5-7B-Instruct2026.05 | 55.06 | 59.45 | |
| Context+DocsBackbone=Llama3.1-8B-Instruct2026.05 | 53.8 | — | |
| Global ParameterizationBackbone=Qwen2.5-7B-Instruct2026.05 | 37.97 | 34.18 | |
| Global ParameterizationBackbone=Llama3.1-8B-Instruct2026.05 | 36.08 | 30.38 |