A new framework allows frozen VLMs and LLMs to incorporate new clinical evidence and guidelines without weight fine-tuning. It uses external expertise via Skills and Knowledge modules to update reasoning and tool use while avoiding overfitting common in parameter-free methods.
HOW THIS AFFECTS YOU
●
researcherYou can implement parameter-free updates for vision-language models in specialized domains.
●
healthThis provides a way to update medical AI with new clinical guidelines without expensive retraining.