SciClaims: An End-to-End Generative System for Biomedical Claim Analysis

TL;DR

Biomedicine increasingly demands scalable validation of scientific claims. SciClaims proposes an end-to-end, LLM-driven pipeline that extracts claims from text, retrieves evidence from a large PubMed corpus, and verifies claims with explanations, all through a web interface and optimized for a single GPU. The system integrates three modules—claim extraction, evidence retrieval, and claim verification—and employs prompt refinements (CDP/CR) evaluated via SciFact benchmarks, judge-model analyses, and human studies, showing strong claim quality, retrieval relevance, and verification accuracy. Its practical design enables real-time, transparent knowledge discovery in high-stakes domains such as biomedicine and pharmaceuticals, with public accessibility and potential deployment in systematic reviews and patent validation.

Abstract

We present SciClaims, an interactive web-based system for end-to-end scientific claim analysis in the biomedical domain. Designed for high-stakes use cases such as systematic literature reviews and patent validation, SciClaims extracts claims from text, retrieves relevant evidence from PubMed, and verifies their veracity. The system features a user-friendly interface where users can input scientific text and view extracted claims, predictions, supporting or refuting evidence, and justifications in natural language. Unlike prior approaches, SciClaims seamlessly integrates the entire scientific claim analysis process using a single large language model, without requiring additional fine-tuning. SciClaims is optimized to run efficiently on a single GPU and is publicly available for live interaction.

Figures (2)

• Figure 1: System Architecture.
• Figure 2: Screenshot of the SciClaims Demo.