Superposition Prompting: Improving and Accelerating Retrieval-Augmented Generation
This addresses efficiency and accuracy problems for users of RAG systems, offering a novel prompting method without fine-tuning, though it is incremental as it builds on existing RAG frameworks.
The paper tackles the high computational cost and distraction issues in retrieval-augmented generation (RAG) for large language models by proposing superposition prompting, which processes documents in parallel paths and discards irrelevant ones, achieving a 93x reduction in compute time and 43% accuracy improvement on the NaturalQuestions-Open dataset.
Despite the successes of large language models (LLMs), they exhibit significant drawbacks, particularly when processing long contexts. Their inference cost scales quadratically with respect to sequence length, making it expensive for deployment in some real-world text processing applications, such as retrieval-augmented generation (RAG). Additionally, LLMs also exhibit the "distraction phenomenon", where irrelevant context in the prompt degrades output quality. To address these drawbacks, we propose a novel RAG prompting methodology, *superposition prompting*, which can be directly applied to pre-trained transformer-based LLMs *without the need for fine-tuning*. At a high level, superposition prompting allows the LLM to process input documents in parallel *prompt paths*, discarding paths once they are deemed irrelevant. We demonstrate the capability of our method to simultaneously enhance time efficiency across a variety of question-answering benchmarks using multiple pre-trained LLMs. Furthermore, our technique significantly improves accuracy when the retrieved context is large relative the context the model was trained on. For example, our approach facilitates a 93x reduction in compute time while *improving* accuracy by 43% on the NaturalQuestions-Open dataset with the MPT-7B instruction-tuned model over naive RAG.