Evidence-Guided Clinical Trial Design

2025 – present | Advisor: Reza Abbasi-Asl (UCSF / UC Berkeley) | Related: RECITE benchmark

LLM pipelines and retrieval over 500K+ clinical trials for eligibility criteria analysis and literature–trial matching, plus large-scale studies of how criteria evolve and whether trials adopt published recommendations for broadening patient eligibility.

• RECITE (KDD 2026): benchmark of 11,913 eligibility criteria revisions across 5,735 trials, built from protocol amendments, with 15 LLM configurations evaluated with and without retrieval-augmented generation
• Eligibility criteria tracking across 500K+ trials with 1.4M+ protocol versions
• Adoption gap analysis: 19.6K assessments across 718 recommendations and 5,386 trials
• Paper discovery and extraction from 7,400+ papers (1,188 verified recommendations from 462 papers)
• ClintriaLM: LLM pipeline that matches published literature to ClinicalTrials.gov trials and analyzes eligibility-criteria revisions, using biomedical embedding models for retrieval