Tag
This paper investigates perturbation-based data selection methods (RSP and RSP-Drop) for adapting medical language models, showing improved Macro-F1 scores over random selection when evaluated with MedGemma-4B on five medical QA datasets.
The author notes a surprising lack of publicly available APIs for medical-oriented LLMs, despite models like MedGemma and BioMistral existing on Hugging Face, and seeks information on any available options.
Google DeepMind released MedGemma 27B Multimodal and MedSigLIP, expanding their open-source Health AI Developer Foundations to include high-performing, privacy-preserving models for medical text and imaging tasks.