Longform Biography Generation on Biographies (test)
71.4FACTSCORECoVe (factor+revise)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CoVe (factor+revise)LLM=Llama 65B2023.09 | 71.4 | 12.3 | |
| CoVe (factored)LLM=Llama 65B2023.09 | 63.7 | 11.7 | |
| Retrieval-basedLLM=PerplexityAI, Note=Reported from previous work2023.09 | 61.6 | 40.8 | |
| CoVe (joint)LLM=Llama 65B2023.09 | 60.8 | 12.8 | |
| Zero-shotLLM=ChatGPT, Note=Reported from previous work2023.09 | 58.7 | 34.7 | |
| Few-shotLLM=Llama 65B2023.09 | 55.9 | 16.6 | |
| Zero-shotLLM=Llama 2 70B Chat2023.09 | 41.3 | 64.9 | |
| Zero-shotLLM=InstructGPT, Note=Reported from previous work2023.09 | 41.1 | 26.3 | |
| CoTLLM=Llama 2 70B Chat2023.09 | 41.1 | 49 |