Research

Paper

AI LLM February 24, 2026

RMIT-ADM+S at the MMU-RAG NeurIPS 2025 Competition

Authors

Kun Ran, Marwah Alaofi, Danula Hettiachchi, Chenglong Ma, Khoi Nguyen Dinh Anh, Khoi Vo Nguyen, Sachin Pathiyan Cherumanal, Lida Rashidi, Falk Scholer, Damiano Spina, Shuoqi Sun, Oleg Zendel

Abstract

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.

Metadata

arXiv ID: 2602.20735
Provider: ARXIV
Primary Category: cs.IR
Published: 2026-02-24
Fetched: 2026-02-25 06:05

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.20735v1</id>\n    <title>RMIT-ADM+S at the MMU-RAG NeurIPS 2025 Competition</title>\n    <updated>2026-02-24T09:58:25Z</updated>\n    <link href='https://arxiv.org/abs/2602.20735v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.20735v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>This paper presents the award-winning RMIT-ADM+S system for the Text-to-Text\n  track of the NeurIPS~2025 MMU-RAG Competition. We introduce Routing-to-RAG\n  (R2RAG), a research-focused retrieval-augmented generation (RAG)\n  architecture composed of lightweight components that dynamically adapt the\n  retrieval strategy based on inferred query complexity and evidence\n  sufficiency. The system uses smaller LLMs, enabling operation on a single\n  consumer-grade GPU while supporting complex research tasks. It builds on the\n  G-RAG system, winner of the ACM~SIGIR~2025 LiveRAG Challenge, and extends it\n  with modules informed by qualitative review of outputs. R2RAG won the Best\n  Dynamic Evaluation award in the Open Source category, demonstrating high\n  effectiveness with careful design and efficient use of resources.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.IR'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-02-24T09:58:25Z</published>\n    <arxiv:comment>MMU-RAG NeurIPS 2025 winning system</arxiv:comment>\n    <arxiv:primary_category term='cs.IR'/>\n    <author>\n      <name>Kun Ran</name>\n    </author>\n    <author>\n      <name>Marwah Alaofi</name>\n    </author>\n    <author>\n      <name>Danula Hettiachchi</name>\n    </author>\n    <author>\n      <name>Chenglong Ma</name>\n    </author>\n    <author>\n      <name>Khoi Nguyen Dinh Anh</name>\n    </author>\n    <author>\n      <name>Khoi Vo Nguyen</name>\n    </author>\n    <author>\n      <name>Sachin Pathiyan Cherumanal</name>\n    </author>\n    <author>\n      <name>Lida Rashidi</name>\n    </author>\n    <author>\n      <name>Falk Scholer</name>\n    </author>\n    <author>\n      <name>Damiano Spina</name>\n    </author>\n    <author>\n      <name>Shuoqi Sun</name>\n    </author>\n    <author>\n      <name>Oleg Zendel</name>\n    </author>\n  </entry>"
}