Tool-use instruction following on ToolBench I3-Instruction
71Pass RateDFSDT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DFSDTBackbone=GPT-4, Retriever=Oracle2023.07 | 71 | 84 | |
| DFSDTBackbone=ToolLLaMA, Retriever=Oracle2023.07 | 66 | 69 | |
| DFSDTBackbone=ToolLLaMA, Retriever=API Retriever2023.07 | 65 | 73 | |
| DFSDTBackbone=ChatGPT, Retriever=Oracle2023.07 | 62 | 69 | |
| ReACTBackbone=GPT-4, Retriever=Oracle2023.07 | 47 | 78 | |
| DFSDTBackbone=Text-Davinci-003, Retriever=Oracle2023.07 | 46 | 63 | |
| DFSDTBackbone=Claude-2, Retriever=Oracle2023.07 | 28 | 65 | |
| ReACTBackbone=ToolLLaMA, Retriever=Oracle2023.07 | 25 | 55 | |
| ReACTBackbone=Text-Davinci-003, Retriever=Oracle2023.07 | 24 | 45 | |
| ReACTBackbone=ChatGPT, Retriever=Oracle2023.07 | 22 | — | |
| ReACTBackbone=Claude-2, Retriever=Oracle2023.07 | 14 | 47.5 | |
| ReACT & DFSDTBackbone=Vicuna, Retriever=Oracle2023.07 | 0 | 0 | |
| ReACT & DFSDTBackbone=Alpaca, Retriever=Oracle2023.07 | 0 | 0 |