Back to Agenda
Navigating the Regulatory Labyrinth: How Well Do Language Models Read the Fine Print?
Session Chair(s)
Sam Kay, RAC
VP of Pharmaceutical Strategy
Basil Systems, United States
Comprehensive benchmarking of four major LLMs across 100 FDA approval packages identifies optimal models for regulatory tasks. Study provides critical insights on accuracy, limitations, and implementation strategies for regulatory AI adoption.
Learning Objective : Identify which LLM models demonstrated highest accuracy (Claude vs Gemini vs ChatGPT vs Perplexity) for extracting critical regulatory data from FDA approval packages; Recognize key limitations and failure modes when using LLMs for regulatory intelligence, including hallucination rates and performance degradation patterns; Apply evidence-based benchmarking to select an LLM
Have an account?