Reproduce “SciCurateBench: Curating Experimental Data from Heterogeneous Scientific Literature with Verbal-Action 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
Minuk Ma, Kaiyun Guo, Jiarui Ding
LangPoli addresses literature-to-dataset curation by generating paper-specific tools from natural-language instructions and using a critic to identify evidence gaps. In zero-shot tests on the 200-paper life-sciences SciCurateBench, it scores 83.16% classification F1 versus 58.22% for Claude-3.5, and 69.43% retrieval F1 versus 59.13% for ChatCoT. Removing parallel specialist actors lowers retrieval F1 to 62.97%; results are limited to this benchmark.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.