Grounding accuracy on Fakepedia (FP) (test)
0.92Grounding AccuracyMistral
Evaluation Results
| Method | Links | |
|---|---|---|
| MistralInstruction=With, Size=7B2023.12 | 0.92 | |
| MistralInstruction=Without, Size=7B2023.12 | 0.9 | |
| Llama2Instruction=With, Size=70B2023.12 | 0.9 | |
| Llama2Instruction=With, Size=13B2023.12 | 0.84 | |
| Llama2Instruction=Without, Size=70B2023.12 | 0.73 | |
| Llama2Instruction=Without, Size=13B2023.12 | 0.7 | |
| GPT-3.5 TurboInstruction=With, Snapshot=03/012023.12 | 0.61 | |
| ZephyrInstruction=With, Size=7B2023.12 | 0.58 | |
| GPT-3.5 TurboInstruction=With, Snapshot=06/132023.12 | 0.54 | |
| ZephyrInstruction=Without, Size=7B2023.12 | 0.52 | |
| GPT-3.5 TurboInstruction=With, Snapshot=11/062023.12 | 0.5 | |
| GPT-3.5 TurboInstruction=Without, Snapshot=03/012023.12 | 0.47 | |
| GPT-4 TurboInstruction=With, Snapshot=11/062023.12 | 0.28 | |
| GPT-3.5 TurboInstruction=Without, Snapshot=11/062023.12 | 0.27 | |
| GPT-3.5 TurboInstruction=Without, Snapshot=06/132023.12 | 0.24 | |
| Llama2Instruction=With, Size=7B2023.12 | 0.22 | |
| Llama2Instruction=Without, Size=7B2023.12 | 0.01 | |
| GPT-4 TurboInstruction=Without, Snapshot=11/062023.12 | 0.01 |