Tool Use on APIBench OOD - TensorHub
2.04Hallucination RateGorilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GorillaTraining Setting=Retrieval-aware (RS), Retriever=Oracle2023.07 | 2.04 | 94.16 | |
| GorillaTraining Setting=Retrieval-aware (RS), Retriever=BM252023.07 | 2.77 | 41.9 | |
| ToolLLAMARetriever=Our Retriever2023.07 | 6.48 | 40.59 | |
| ToolLLAMARetriever=Oracle2023.07 | 7.44 | 88.62 | |
| GorillaTraining Setting=Zero-shot (ZS), Retriever=Oracle2023.07 | 12.99 | 83.21 | |
| GorillaTraining Setting=Zero-shot (ZS), Retriever=BM252023.07 | 20.58 | 34.31 |