Tool-use instruction following on ToolBench I1-Category
67Pass RateDFSDT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DFSDTBackbone=GPT-4, Retriever=Oracle2023.07 | 67 | 66.5 | |
| DFSDTBackbone=ToolLLaMA, Retriever=Oracle2023.07 | 62 | 54.5 | |
| DFSDTBackbone=ChatGPT, Retriever=Oracle2023.07 | 60.5 | 57.3 | |
| DFSDTBackbone=ToolLLaMA, Retriever=API Retriever2023.07 | 60.5 | 55 | |
| ReACTBackbone=GPT-4, Retriever=Oracle2023.07 | 53.5 | 63.5 | |
| DFSDTBackbone=Text-Davinci-003, Retriever=Oracle2023.07 | 46 | 46.8 | |
| ReACTBackbone=ChatGPT, Retriever=Oracle2023.07 | 44.5 | — | |
| ReACTBackbone=ToolLLaMA, Retriever=Oracle2023.07 | 33 | 47.5 | |
| ReACTBackbone=Text-Davinci-003, Retriever=Oracle2023.07 | 20 | 31 | |
| DFSDTBackbone=Claude-2, Retriever=Oracle2023.07 | 18.5 | 43.3 | |
| ReACTBackbone=Claude-2, Retriever=Oracle2023.07 | 5.5 | 33.8 | |
| ReACT & DFSDTBackbone=Vicuna, Retriever=Oracle2023.07 | 0 | 0 | |
| ReACT & DFSDTBackbone=Alpaca, Retriever=Oracle2023.07 | 0 | 0 |