Tool Calling on BFCL Live
79.01Multiple Success RateParaTool (ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ParaTool (ours)Model=Qwen2.5-7B-Instruct2026.05 | 79.01 | 79.17 | 73.61 | 77.26 | |
| ParaTool (ours)Model=Llama3.1-8B-Instruct2026.05 | 78 | 85.42 | 65.28 | 76.23 | |
| Context+Docs & ExamplesModel=Qwen2.5-7B-Instruct2026.05 | 77.9 | 75 | 69.44 | 74.12 | |
| Parameter+EmbsimModel=Llama3.1-8B-Instruct2026.05 | 77.08 | — | 48.61 | 70.37 | |
| Parameter+EmbsimModel=Qwen2.5-7B-Instruct2026.05 | 77.02 | — | 51.39 | 69.19 | |
| Parameter+ToolRetrieverModel=Qwen2.5-7B-Instruct2026.05 | 76.95 | — | 50 | 68.71 | |
| Parameter+ReinvokeModel=Qwen2.5-7B-Instruct2026.05 | 76.92 | — | 56.94 | 71.01 | |
| Parameter+ReinvokeModel=Llama3.1-8B-Instruct2026.05 | 76.48 | — | 51.39 | 71.1 | |
| Parameter+Tool RetrieverModel=Llama3.1-8B-Instruct2026.05 | 76.45 | — | 43.06 | 68.31 | |
| Context+DocsModel=Qwen2.5-7B-Instruct2026.05 | 75.28 | 58.33 | 62.5 | 65.37 | |
| Context+Docs & ExamplesModel=Llama3.1-8B-Instruct2026.05 | 73.88 | 79.17 | 61.11 | 71.39 | |
| Context+DocsModel=Llama3.1-8B-Instruct2026.05 | 71.23 | 56.25 | 55.56 | 61.01 | |
| ToolLLaMAModel=Llama3.1-8B-Instruct2026.05 | 58.21 | 62.5 | 33.33 | 51.35 | |
| Global ParameterizationModel=Llama3.1-8B-Instruct2026.05 | 57.99 | 60.42 | 30.56 | 49.66 | |
| ToolLLaMAModel=Qwen2.5-7B-Instruct2026.05 | 52.14 | 56.25 | 29.17 | 45.85 | |
| Global ParameterizationModel=Qwen2.5-7B-Instruct2026.05 | 47.04 | 58.33 | 31.94 | 45.77 |