Reproduce “DeepWeaver: Bridging the Evidence Synthesis Gap in Open-Ended Question Answering” 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
Xujia Wang, Yizhe Zhang, Bin Xu, Lei Hou, Juanzi Li
DeepWeaver addresses research-style question answering by organizing evidence into revisable claims linked to supporting passages. It drafts an answer, searches residual evidence for missing details, then merges and prunes claims before generating cited sections. On LoQA, versus direct-generation RAG, it raises argument sufficiency from 68.0 to 83.5 and detail preservation from 52.9 to 69.5, within the 100-question Chinese water-environment benchmark.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.