Reproduce “FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents” 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

Jia Deng, Yimeng Chen, Xiaoqing Xiang, Ziyang Zeng, Shuo Tang, Zhengbo Zhang, Wayne Xin Zhao, Feng Chang and 5 more

FORT develops tasks that make search agents gather evidence across sources rather than rely on retrieval shortcuts. It selects obscure entities, combines facts from multiple sources, obscures searchable names or values, and tests drafts with an agent, revising shortcut-prone or ambiguous questions. In evaluation, FORT required more retrieval and exposed answers later; these measures are proxies, not minimum route costs.

Summary by alphaXiv.

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