Verily’s Post

View organization page for Verily, graphic

163,837 followers

A unique feature of Verily Site CTMS, our clinical trial management system, is protocol digitization. That means taking a study protocol in PDF format and creating a structured, digital version. The use of AI can further expedite the digitization process, with the goal of enabling site research operations teams to increase the productivity and revenue of their research programs. In this study, Verily researchers demonstrated that a fine-tuned LLM (PaLM-2) model can achieve near optimal performance without requiring manual labeling, a process that is both expensive and time-consuming. Our results show that using high confidence labels generated by LLM models such as Gemini-pro can be an efficient strategy for improving LLM performance on specialized tasks such as protocol digitization. Next week at the Machine Learning for Healthcare 2024 Conference we will present this work in the research track. https://veri.ly/46ERPQK #CTMS #LLM #AIInHealthcare #ClinicalTrials

  • Part of a schematic illustrating the approach to fine-tuning PaLM-2

To view or add a comment, sign in

Explore topics