Get started
Work

Can LLMs extract scientific consensus? A case study in high-temperature superconductivity

Year: 2026
Type: preprint
Abstract: Scientific knowledge is increasingly dispersed across vast and heterogeneous scientific literature, where important claims are often implicit, evolving, and internally debated. While large language models (LLMs) have shown impressive performance in information extraction and summarization, their ability to recover latent scientific consensus remains unclear. Here, we investigate this problem in the context of high-temperature superconductivity (HTS), a long-standing and highly debated topic in condensed matter physics, as a challenging testbed. Using near 18,000 highly-cited publications over the past seven decades, we construct a structured knowledge graph linking competing superconducting mechanisms, material families, evidential modalities, and citation relations. We find that LLM-extracted representations recover coherent and physically interpretable structures, including family-dependent mechanism profiles, evidence-specific correlations, and citation-mediated temporal evolution of scientific beliefs. Ablation studies on LLM further show that the global structure remains robust across prompting, decoding, and model variations. Our results suggest that LLMs can indeed serve as scalable tools for deciphering scientific knowledge in domains characterized by competing interpretations and evolving knowledge.
DOI: 10.48550/arxiv.2606.07570
Authors: Cheng, Mouyang, He, Wenhao, Jin, Zhuotao, Yu, Bowen, Li, Ju, Kozinsky, Boris, Wang, Yao, Volkov, Pavel, Deng, Liangzi, Chu, Ching-Wu, Wen, Xiao-Gang, Li, Mingda
Institutions: IIT@MIT, Massachusetts Institute of Technology, Harvard University, Emory University, University of Connecticut, University of Houston
Open Access status: green
Funders: National Science Foundation, U.S. Department of Energy, Robert Bosch, Office of Science, Division of Materials Research, Office of Advanced Cyberinfrastructure, Basic Energy Sciences
Awards: TRACING THE TOPOLOGICAL FINGERPRINT OF WEYL SEMIMETALS USING NEUTRON PROBES, G1369588039, HDR Institute: Institute for Data Driven Dynamical Design, EAGER: SUPER: Electrochemical Protonation to Achieve Superconducting Matter, G4556251948, G5565249710, TRACING THE TOPOLOGICAL FINGERPRINT OF WEYL SEMIMETALS USING NEUTRON PROBES, NSF Convergence Accelerator Track I: Advancing Sustainable Topological Material Prototype Devices for Energy-efficient Applications
OpenAlex ID: W7164059883
Cites:
Cited by:
Related to: