HiFi-RAG: Hierarchical Content Filtering and Two-Pass Generation for Open-Domain RAG
This addresses retrieval and generation challenges in open-domain RAG for applications like question-answering, though it appears incremental as it builds on existing RAG methods with a multi-stage pipeline.
The paper tackles the problem of irrelevant information and misalignment in open-domain Retrieval-Augmented Generation (RAG) by introducing HiFi-RAG, a hierarchical filtering and two-pass generation system, which improved ROUGE-L by 19.6% and DeBERTaScore by 6.2% on a validation set and outperformed a baseline by up to 57.4% on a test set.
Retrieval-Augmented Generation (RAG) in open-domain settings faces significant challenges regarding irrelevant information in retrieved documents and the alignment of generated answers with user intent. We present HiFi-RAG (Hierarchical Filtering RAG), the winning closed-source system in the Text-to-Text static evaluation of the MMU-RAGent NeurIPS 2025 Competition. Our approach moves beyond standard embedding-based retrieval via a multi-stage pipeline. We leverage the speed and cost-efficiency of Gemini 2.5 Flash (4-6x cheaper than Pro) for query formulation, hierarchical content filtering, and citation attribution, while reserving the reasoning capabilities of Gemini 2.5 Pro for final answer generation. On the MMU-RAGent validation set, our system outperformed the baseline, improving ROUGE-L to 0.274 (+19.6%) and DeBERTaScore to 0.677 (+6.2%). On Test2025, our custom dataset evaluating questions that require post-cutoff knowledge (post January 2025), HiFi-RAG outperforms the parametric baseline by 57.4% in ROUGE-L and 14.9% in DeBERTaScore.