Claim-level Uncertainty Quantification on FactScore English (test)
71ROC-AUCCCP
Evaluation Results
| Method | Links | |
|---|---|---|
| CCPBase LLM=Jais 13b2024.03 | 71 | |
| CCPBase LLM=Mistral 7b2024.03 | 66 | |
| CCPBase LLM=Vicuna 13b2024.03 | 66 | |
| Maximum Prob.Base LLM=Jais 13b2024.03 | 64 | |
| Token EntropyBase LLM=Jais 13b2024.03 | 63 | |
| PerplexityBase LLM=Jais 13b2024.03 | 61 | |
| P(True)Base LLM=Vicuna 13b2024.03 | 61 | |
| Maximum Prob.Base LLM=Vicuna 13b2024.03 | 60 | |
| Token EntropyBase LLM=Mistral 7b2024.03 | 60 | |
| Token EntropyBase LLM=Vicuna 13b2024.03 | 60 | |
| Maximum Prob.Base LLM=Mistral 7b2024.03 | 59 | |
| CCPBase LLM=GPT-3.5-turbo2024.03 | 58 | |
| PerplexityBase LLM=Mistral 7b2024.03 | 58 | |
| PerplexityBase LLM=Vicuna 13b2024.03 | 58 | |
| P(True)Base LLM=Jais 13b2024.03 | 55 | |
| Maximum Prob.Base LLM=GPT-3.5-turbo2024.03 | 54 | |
| PerplexityBase LLM=GPT-3.5-turbo2024.03 | 53 | |
| Token EntropyBase LLM=GPT-3.5-turbo2024.03 | 53 | |
| P(True)Base LLM=Mistral 7b2024.03 | 53 | |
| P(True)Base LLM=GPT-3.5-turbo2024.03 | 53 |