Reinforcement Learning Environment for LLMs in Cancer Diagnosis from Pathology Slides

Summary

A developer from Aluna (YC S24) presents a reinforcement learning environment for large language models focused on cancer diagnosis. The environment allows frontier LLMs to navigate digitized pathology slides by zooming and panning to identify relevant regions for diagnosis. The project demonstrates how LLMs can be trained to perform medical diagnostic tasks using interactive tools, with videos showing the LLM performing diagnosis on pathology slides. The work involves collaboration with diagnostic labs to build datasets and evaluations for oncology tasks.