Question Answering on NQ DPR 3610 questions
50EMRankRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| RankRAGBackbone=Llama-3, Model size=70B2024.07 | 50 | |
| RankRAGBackbone=Llama-2, Model size=70B2024.07 | 48.7 | |
| RankRAGBackbone=Llama-2, Model size=13B2024.07 | 46.2 | |
| RankRAGBackbone=Llama-3, Model size=8B2024.07 | 46.1 | |
| ChatQA-1.5Backbone=Llama-3, Model size=70B2024.07 | 46 | |
| ChatQA-1.0Backbone=Llama-2, Model size=70B2024.07 | 45 | |
| ChatQA-1.5Backbone=Llama-3, Model size=8B2024.07 | 44.1 | |
| ChatQA-1.0Backbone=Llama-2, Model size=13B2024.07 | 43.9 | |
| RankRAGBackbone=Llama-2, Model size=7B2024.07 | 42.4 | |
| GPT-3.5-0613 RAGModel Family=OpenAI GPT, RAG=true2024.07 | 42.3 | |
| GPT-4-turbo-2024-0409Model Family=OpenAI GPT2024.07 | 38.3 | |
| Llama-2-ChatBackbone=Llama-2, Configuration=Llama-2 RAG 70B, RAG=true2024.07 | 37.7 | |
| Llama-3-InstructBackbone=Llama-3, Model size=70B, RAG=true2024.07 | 37.3 | |
| GPT-4-0613Model Family=OpenAI GPT2024.07 | 37.2 | |
| ChatQA-1.0Backbone=Llama-2, Model size=7B2024.07 | 37 | |
| GPT-4-turbo-2024-0409 RAGModel Family=OpenAI GPT, RAG=true2024.07 | 36.3 | |
| GPT-4-0613 RAGModel Family=OpenAI GPT, RAG=true2024.07 | 36.2 | |
| GPT-3.5-0613Model Family=OpenAI GPT2024.07 | 35.2 | |
| Llama-3-InstructBackbone=Llama-3, Model size=8B, RAG=true2024.07 | 27.6 |