Tool-calling on Stable Toolbench I3-Inst
68.85Pass RateParaTool
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ParaToolBackbone=Qwen2.5-7B-Instruct2026.05 | 68.85 | 73.77 | |
| ParaToolBackbone=Llama3.1-8B-Instruct2026.05 | 65.57 | 68.85 | |
| Context+Docs & ExamplesBackbone=Qwen2.5-7B-Instruct2026.05 | 65.57 | 68.85 | |
| ToolLLaMABackbone=Qwen2.5-7B-Instruct2026.05 | 65.57 | 70.49 | |
| Context+Docs & ExamplesBackbone=Llama3.1-8B-Instruct2026.05 | 63.93 | 55.74 | |
| ToolLLaMABackbone=Llama3.1-8B-Instruct2026.05 | 60.66 | 57.38 | |
| Parameter+ReinvokeBackbone=Qwen2.5-7B-Instruct2026.05 | 60.66 | 62.3 | |
| Context+DocsBackbone=Qwen2.5-7B-Instruct2026.05 | 58.82 | — | |
| Parameter+ReinvokeBackbone=Llama3.1-8B-Instruct2026.05 | 57.38 | 55.74 | |
| Context+DocsBackbone=Llama3.1-8B-Instruct2026.05 | 52.46 | — | |
| Parameter+ToolRetrieverBackbone=Qwen2.5-7B-Instruct2026.05 | 52.46 | 55.74 | |
| Parameter+ToolRetrieverBackbone=Llama3.1-8B-Instruct2026.05 | 50.82 | 52.46 | |
| Parameter+EmbsimBackbone=Qwen2.5-7B-Instruct2026.05 | 50.82 | 54.1 | |
| Parameter+EmbsimBackbone=Llama3.1-8B-Instruct2026.05 | 47.54 | 44.26 | |
| Global ParameterizationBackbone=Qwen2.5-7B-Instruct2026.05 | 31.33 | 27.87 | |
| Global ParameterizationBackbone=Llama3.1-8B-Instruct2026.05 | 29.25 | 21.31 |