Tool-calling on Stable Toolbench I2-Inst
77.42Pass RateParaTool
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ParaToolBackbone=Qwen2.5-7B-Instruct2026.05 | 77.42 | 79.03 | |
| ParaToolBackbone=Llama3.1-8B-Instruct2026.05 | 74.19 | 75 | |
| ToolLLaMABackbone=Llama3.1-8B-Instruct2026.05 | 70.75 | 67.92 | |
| Parameter+ReinvokeBackbone=Qwen2.5-7B-Instruct2026.05 | 69.35 | 73.39 | |
| Context+Docs & ExamplesBackbone=Llama3.1-8B-Instruct2026.05 | 68.55 | 63.71 | |
| Parameter+ReinvokeBackbone=Llama3.1-8B-Instruct2026.05 | 68.55 | 70.16 | |
| ToolLLaMABackbone=Qwen2.5-7B-Instruct2026.05 | 66.98 | 63.2 | |
| Parameter+ToolRetrieverBackbone=Llama3.1-8B-Instruct2026.05 | 65.32 | 64.52 | |
| Context+Docs & ExamplesBackbone=Qwen2.5-7B-Instruct2026.05 | 65 | 61.29 | |
| Parameter+EmbsimBackbone=Llama3.1-8B-Instruct2026.05 | 62.1 | 64.52 | |
| Parameter+ToolRetrieverBackbone=Qwen2.5-7B-Instruct2026.05 | 59.68 | 62.9 | |
| Context+DocsBackbone=Llama3.1-8B-Instruct2026.05 | 58.06 | — | |
| Parameter+EmbsimBackbone=Qwen2.5-7B-Instruct2026.05 | 54.84 | 62.1 | |
| Context+DocsBackbone=Qwen2.5-7B-Instruct2026.05 | 51 | — | |
| Global ParameterizationBackbone=Qwen2.5-7B-Instruct2026.05 | 44.35 | 47.58 | |
| Global ParameterizationBackbone=Llama3.1-8B-Instruct2026.05 | 41.94 | 38.71 |