This paper presents the award-winning RMIT-ADM+S system for the Text-to-Text track of the NeurIPS2025 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 ACMSIGIR~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.
Figures & tables
Figure 1: Overview of the system. The workflow includes a query classifier that routes each query to either a Vanilla RAG path for simple queries or a Vanilla Agent for complex queries that iteratively issues revised search query variants until sufficient evidence is gathered, followed by an answer generation module.