RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs
About
Large language models (LLMs) typically utilize the top-k contexts from a retriever in retrieval-augmented generation (RAG). In this work, we propose a novel instruction fine-tuning framework RankRAG, which instruction-tunes a single LLM for the dual purpose of context ranking and answer generation in RAG. In particular, the instruction-tuned LLMs work surprisingly well by adding a small fraction of ranking data into the training blend, and outperform existing expert ranking models, including the same LLM exclusively fine-tuned on a large amount of ranking data. For generation, we compare our model with many strong baselines, including GPT-4-0613, GPT-4-turbo-2024-0409, and ChatQA-1.5, an open-sourced model with the state-of-the-art performance on RAG benchmarks. Specifically, our Llama3-RankRAG significantly outperforms Llama3-ChatQA-1.5 and GPT-4 models on nine knowledge-intensive benchmarks. In addition, it also performs comparably to GPT-4 on five RAG benchmarks in the biomedical domain without instruction fine-tuning on biomedical data, demonstrating its superb capability for generalization to new domains.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Question Answering | ARC Challenge | Accuracy70.6 | 906 | |
| Multi-hop Question Answering | 2WikiMultihopQA | EM38.2 | 387 | |
| Question Answering | OBQA | Accuracy87.5 | 300 | |
| Multi-hop Question Answering | HotpotQA | F1 Score63.6 | 294 | |
| Question Answering | PopQA | Accuracy66.1 | 186 | |
| Question Answering | 2Wiki | F160 | 152 | |
| Multi-hop Question Answering | 2Wiki | -- | 152 | |
| Question Answering | PubMedQA | Accuracy79.8 | 145 | |
| Question Answering | HotpotQA | F155.4 | 128 | |
| Question Answering | PubMedQA (test) | Accuracy65 | 128 |