Attributed Text Generation on ASQA
50.1Correctness (EM Rec.)Think&Cite
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Think&CiteBase LLM=GPT-4o2024.12 | 50.1 | 89.5 | 87.1 | |
| Summary RAG w/ InteractBase LLM=GPT-4o2024.12 | 48.1 | 73.1 | 72.8 | |
| Summary RAGBase LLM=GPT-4o2024.12 | 46.5 | 70.2 | 67.2 | |
| Snippet RAG w/ InteractBase LLM=GPT-4o2024.12 | 45.2 | 67.8 | 66.7 | |
| Think&CiteBase LLM=LLaMA2024.12 | 45.2 | 82.3 | 80.6 | |
| Snippet RAGBase LLM=GPT-4o2024.12 | 45.1 | 68.9 | 66.5 | |
| Summary RAGBase LLM=ChatGPT2024.12 | 43.3 | 68.9 | 61.8 | |
| Vanilla RAG w/ ReRankBase LLM=GPT-4o2024.12 | 42.1 | 83.4 | 82.3 | |
| VTG2024.12 | 41.5 | 86.7 | 80 | |
| Snippet RAGBase LLM=ChatGPT2024.12 | 41.4 | 65.3 | 57.4 | |
| Vanilla RAGBase LLM=GPT-4o2024.12 | 41.3 | 68.5 | 75.6 | |
| Snippet RAG w/ InteractBase LLM=ChatGPT2024.12 | 41.2 | 64.5 | 57.7 | |
| APO2024.12 | 40.5 | 72.8 | 69.6 | |
| Vanilla RAGBase LLM=ChatGPT2024.12 | 40.4 | 73.6 | 72.5 | |
| Inline SearchBase LLM=GPT-4o2024.12 | 40.3 | 65.7 | 66.9 | |
| Vanilla RAG w/ ReRankBase LLM=ChatGPT2024.12 | 40.2 | 84.8 | 81.6 | |
| FG-Reward2024.12 | 40.1 | 77.8 | 76.3 | |
| Summary RAG w/ InteractBase LLM=ChatGPT2024.12 | 39.1 | 73.4 | 66.5 | |
| Inline SearchBase LLM=ChatGPT2024.12 | 32.4 | 58.3 | 58.2 |