AI RESEARCH

Sparse Spectral LoRA: Routed Experts for Medical VLMs

arXiv CS.CV

ArXi:2604.01310v1 Announce Type: new Large vision-language models (VLMs) excel on general benchmarks but often lack robustness in medical imaging, where heterogeneous supervision induces cross-dataset interference and sensitivity to data regime (i.e., how the supervisory signals are mixed). In realistic clinical workflows, data and tasks arrive sequentially, so naive continual