Citation-aware Question Answering on ALCE ELI5
21.4EM RecallFineRef
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FineRefBase Model=Mistral-Orca-7B2025.11 | 21.4 | 22 | 63.5 | 15.7 | 76.3 | |
| Mistral-Orca-7BProtocol=In-context2025.11 | 20.8 | 20.9 | 29.3 | 12.5 | 43.3 | |
| FineRefBase Model=LLaMA2-7B-chat2025.11 | 19.2 | 21.5 | 60.1 | 13.5 | 68.4 | |
| CaLFBase Model=Mistral-Orca-7B2025.11 | 18.4 | 20.4 | 62.7 | 13.1 | 73.1 | |
| LLaMA2-7B-chatProtocol=Few-shot FT2025.11 | 17.8 | 21.3 | 58.2 | 11.2 | 48.7 | |
| LLaMA2-7B-chatProtocol=In-context2025.11 | 17.7 | 20.5 | 36.2 | 10.8 | 38.2 | |
| Mistral-Orca-7BProtocol=Few-shot FT2025.11 | 17.3 | 19.4 | 60.5 | 10.9 | 57.7 | |
| CaLFBase Model=LLaMA2-7B-chat2025.11 | 17 | 20.8 | 59.6 | 11.9 | 66.5 | |
| GPT-4Model Version=gpt-4-06132025.11 | 14.2 | — | 38.4 | — | 46.9 | |
| ChatGPTModel Version=gpt-3.5-turbo-03012025.11 | 12 | — | 57.2 | — | 50.5 | |
| Self-RAG 7BBase Model=Llama22025.11 | 9.7 | 16.9 | 32.6 | 5.4 | 27.6 | |
| Blueprint (BP)Base Model=T5-3B2025.11 | 5.2 | — | — | — | 60.9 |