Fact-checking of atomic claims on FactScore English
0.34PR-AUCCCP
Evaluation results
| Method | Links | |
|---|---|---|
| CCPLLM=Mistral 7b2024.03 | 0.34 | |
| CCPLLM=Jais 13b2024.03 | 0.33 | |
| Token EntropyLLM=Mistral 7b2024.03 | 0.3 | |
| PerplexityLLM=Mistral 7b2024.03 | 0.27 | |
| Maximum Prob.LLM=Mistral 7b2024.03 | 0.26 | |
| CCPLLM=Vicuna 13b2024.03 | 0.24 | |
| Maximum Prob.LLM=Jais 13b2024.03 | 0.24 | |
| Token EntropyLLM=Jais 13b2024.03 | 0.24 | |
| P(True)LLM=Mistral 7b2024.03 | 0.24 | |
| PerplexityLLM=Jais 13b2024.03 | 0.21 | |
| P(True)LLM=Vicuna 13b2024.03 | 0.21 | |
| P(True)LLM=Jais 13b2024.03 | 0.19 | |
| Maximum Prob.LLM=Vicuna 13b2024.03 | 0.18 | |
| Token EntropyLLM=Vicuna 13b2024.03 | 0.18 | |
| PerplexityLLM=Vicuna 13b2024.03 | 0.17 | |
| P(True)LLM=GPT-3.5-turbo2024.03 | 0.17 | |
| CCPLLM=GPT-3.5-turbo2024.03 | 0.14 | |
| Maximum Prob.LLM=GPT-3.5-turbo2024.03 | 0.13 | |
| Token EntropyLLM=GPT-3.5-turbo2024.03 | 0.12 | |
| PerplexityLLM=GPT-3.5-turbo2024.03 | 0.11 |