Reproduce “STAIR (STructure Aware Information Retriever): A novel dataset and LLM based retriever for document structure augmentation” 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
Vineet Kumar, Meghanadh Pulivarthi, vishwajeet kumar, Jaydeep Sen, Riyaz Ahmad Bhat, Sachindra Joshi
STAIR uses a document’s table of contents (ToC) to retrieve the leaf section likely to contain an answer, rather than selecting fixed-length text chunks. It is trained on queries paired with the ToC and gold section titles, then generates a leaf title. On SearchTome, it achieves 82.6% Recall@1 versus DSI’s 76.9%; this evaluates gold-section retrieval on structured books with synthetic questions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.