Search papers, labs, and topics across Lattice.
The paper introduces Routing-to-RAG (R2RAG), a retrieval-augmented generation architecture that dynamically adjusts its retrieval strategy based on query complexity and evidence sufficiency. R2RAG builds upon the G-RAG system and incorporates modules refined through qualitative output analysis. The system achieved the Best Dynamic Evaluation award in the Open Source category of the NeurIPS 2025 MMU-RAG Competition, showcasing its effectiveness and resource efficiency using smaller LLMs on a single consumer-grade GPU.
A lightweight RAG architecture can achieve state-of-the-art results in dynamic evaluation by intelligently routing queries to different retrieval strategies.
This paper presents the award-winning RMIT-ADM+S system for the Text-to-Text track of the NeurIPS~2025 MMU-RAG Competition. We introduce Routing-to-RAG (R2RAG), a research-focused retrieval-augmented generation (RAG) architecture composed of lightweight components that dynamically adapt the retrieval strategy based on inferred query complexity and evidence sufficiency. The system uses smaller LLMs, enabling operation on a single consumer-grade GPU while supporting complex research tasks. It builds on the G-RAG system, winner of the ACM~SIGIR~2025 LiveRAG Challenge, and extends it with modules informed by qualitative review of outputs. R2RAG won the Best Dynamic Evaluation award in the Open Source category, demonstrating high effectiveness with careful design and efficient use of resources.
