SCEPTER: AI Framework for Evidence-Based Clinical Recommendations
Researchers propose SCEPTER (Single-Case Evidence-driven PubMed-To-rEcommendation Reasoner), a framework that transforms clinical case descriptions into evidence-based recommendations by combining PubMed retrieval, PubMedBERT semantic ranking, LLM-based claim extraction, evidence-level weighting, contradiction detection, consensus analysis, and multi-objective Pareto claim selection. SCEPTER generates structured evidence syntheses and actionable recommendations, with a Paper Q&A module for interactive exploration. The framework addresses the challenge of reviewing hundreds of publications from PubMed searches under time constraints, introducing a multi-objective reasoning model.
Key facts
- SCEPTER stands for Single-Case Evidence-driven PubMed-To-rEcommendation Reasoner
- Framework combines PubMed retrieval, PubMedBERT, LLM-based claim extraction, evidence weighting, contradiction detection, consensus analysis, and Pareto claim selection
- Generates structured evidence syntheses and actionable recommendations
- Includes a Paper Q&A module for interactive exploration of selected publications
- Addresses time constraints in reviewing PubMed search results for complex clinical cases
- Introduces multi-objective reasoning model
- Published on arXiv with ID 2607.22574
- Study proposes framework for transforming clinical case descriptions into evidence-based recommendations
Entities
Institutions
- arXiv
- PubMed
- PubMedBERT