Reproduce “Faithful in Steps: Improving Generalization and Citation in RAG via Query Decomposition” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yue Liu, Zhongying Ru, Shimin Di, Jipeng Zhang, Ruiyuan Zhang, Xiaofang Zhou

QDRAG-R improves citation support for multimodal retrieval-augmented question answering by decomposing questions before filtering retrieved evidence. Its reranker scores contexts against the original question and subquestions, while preference training rewards decompositions that recall gold contexts. On out-of-domain InfoSeek, its 7B model scores 72.9 Cite, versus 66.3 without DPO. Cite is an automated aggregate-support judgment, not verification of individual citations.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.