Reproduce “CORE: Coarse-to-Fine Search with Context Management for Deep Research” 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
Hoang Tran Vuong, Ngoc Quan Pham, Hung Bui, Fatih Porikli, Dat Quoc Nguyen
CORE is a training-free retrieval harness for deep-research agents. It searches broadly, then returns a focused passage snippet and lets agents request nearby context or avoid repeated snippets. Late-interaction token matching locates relevant text within retrieved passages. In fixed-corpus tests, it leads accuracy and evidence recall across nine agent–benchmark combinations, not live-web or open-ended research.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.