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

Validating key claims in scientific literature, particularly in biomedical research, is essential for ensuring accuracy and advancing knowledge. This process is critical in sectors like the pharmaceutical industry, where rapid scientific progress requires automation and deep domain expertise. However, current solutions have significant limitations. They lack end-to-end pipelines encompassing all claim extraction, evidence retrieval, and verification steps; rely on complex NLP and information retrieval pipelines prone to multiple failure points; and often fail to provide clear, user-friendly justifications for claim verification outcomes. To address these challenges, we introduce SciClaims, an advanced system powered by state-of-the-art large language models (LLMs) that seamlessly integrates the entire scientific claim analysis process. SciClaims outperforms previous approaches in both claim extraction and verification without requiring additional fine-tuning, setting a new benchmark for automated scientific claim analysis.

@article{ortega2025_2503.18526,
  title={ SciClaims: An End-to-End Generative System for Biomedical Claim Analysis },
  author={ Raúl Ortega and José Manuel Gómez-Pérez },
  journal={arXiv preprint arXiv:2503.18526},
  year={ 2025 }
}