Attributed Text Generation on QAMPARI
45.2Correctness Recall-5Think&Cite
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Think&CiteBase LLM=GPT-4o2024.12 | 45.2 | 41.9 | 50.6 | 52.8 | |
| Summary RAG w/ InteractBase LLM=GPT-4o2024.12 | 38.2 | 37.1 | 39.2 | 40.6 | |
| Snippet RAG w/ InteractBase LLM=GPT-4o2024.12 | 37.2 | 34.5 | 38.7 | 39.5 | |
| Snippet RAGBase LLM=GPT-4o2024.12 | 37.1 | 35.2 | 37.2 | 38.3 | |
| Summary RAGBase LLM=GPT-4o2024.12 | 36.2 | 34 | 36.2 | 39.8 | |
| Vanilla RAG w/ ReRankBase LLM=GPT-4o2024.12 | 32.6 | 30.9 | 33.1 | 32.8 | |
| Inline SearchBase LLM=GPT-4o2024.12 | 27.8 | 27.2 | 19.4 | 23.8 | |
| Think&CiteBase LLM=LLaMA2024.12 | 25.7 | 28.1 | 40.5 | 43.1 | |
| Snippet RAGBase LLM=ChatGPT2024.12 | 24.5 | 21.5 | 22.9 | 24.9 | |
| Summary RAGBase LLM=ChatGPT2024.12 | 23.6 | 21.2 | 23.6 | 25.7 | |
| Summary RAG w/ InteractBase LLM=ChatGPT2024.12 | 23.2 | 20.9 | 22.1 | 24.3 | |
| Vanilla RAG w/ ReRankBase LLM=ChatGPT2024.12 | 22.8 | 21.4 | 21.2 | 21.4 | |
| Snippet RAG w/ InteractBase LLM=ChatGPT2024.12 | 22.4 | 20.8 | 21.6 | 23.1 | |
| Vanilla RAGBase LLM=GPT-4o2024.12 | 22.2 | 25 | 25.9 | 27 | |
| Vanilla RAGBase LLM=ChatGPT2024.12 | 20.8 | 20.8 | 20.5 | 20.9 | |
| VTG2024.12 | 20.3 | 22.4 | 43.5 | 46.9 | |
| Inline SearchBase LLM=ChatGPT2024.12 | 17.2 | 20.4 | 14.9 | 14.9 | |
| FG-Reward2024.12 | 16.7 | 19.5 | 19.5 | 20.1 | |
| APO2024.12 | 15.4 | 20.6 | 17.5 | 19.3 |