Na minha lista:
Detalhes bibliográficos
Main Authors: Manzari, Omid Nejati, Asgariandehkordi, Hojat, Koleilat, Taha, Xiao, Yiming, Rivaz, Hassan
Formato: Preprint
Publicado em: 2026
Assuntos:
Acesso em linha:https://arxiv.org/abs/2604.01310
Tags: Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
_version_ 1866915906730328064
author Manzari, Omid Nejati
Asgariandehkordi, Hojat
Koleilat, Taha
Xiao, Yiming
Rivaz, Hassan
author_facet Manzari, Omid Nejati
Asgariandehkordi, Hojat
Koleilat, Taha
Xiao, Yiming
Rivaz, Hassan
contents 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 training further leads to catastrophic forgetting. To address these challenges, we propose MedQwen, a parameter-efficient medical VLM that couples a spectrally routed Mixture-of-Experts (MoE) with a theoretically grounded scaling rule that aligns low-rank updates with a full-rank, fully fine-tuned MoE, without changing the base architecture. Concretely, we initialize each expert from non-overlapping singular value decomposition (SVD) segments of the pretrained weight and introduce a residual compensation and scaling scheme to enable stable expert specialization and consistent routing under distribution shift. Across 23 medical datasets covering visual question answering, report generation, radiology classification, and hallucination mitigation, MedQwen achieves strong, reliable performance: it approaches full fine-tuning on zero-shot classification with 339$\times$ fewer trainable parameters, and reduces sequential forgetting to $\sim$5\% where strong baselines degrade by $>$20-50\%.
format Preprint
id arxiv_https___arxiv_org_abs_2604_01310
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Sparse Spectral LoRA: Routed Experts for Medical VLMs
Manzari, Omid Nejati
Asgariandehkordi, Hojat
Koleilat, Taha
Xiao, Yiming
Rivaz, Hassan
Computer Vision and Pattern Recognition
Large vision-language models (VLMs) excel on general benchmarks but often lack robustness in medical imaging, where heterogeneous supervision induces cross-dataset interference and sensitivity to data regime (i.e., how the supervisory signals are mixed). In realistic clinical workflows, data and tasks arrive sequentially, so naive continual training further leads to catastrophic forgetting. To address these challenges, we propose MedQwen, a parameter-efficient medical VLM that couples a spectrally routed Mixture-of-Experts (MoE) with a theoretically grounded scaling rule that aligns low-rank updates with a full-rank, fully fine-tuned MoE, without changing the base architecture. Concretely, we initialize each expert from non-overlapping singular value decomposition (SVD) segments of the pretrained weight and introduce a residual compensation and scaling scheme to enable stable expert specialization and consistent routing under distribution shift. Across 23 medical datasets covering visual question answering, report generation, radiology classification, and hallucination mitigation, MedQwen achieves strong, reliable performance: it approaches full fine-tuning on zero-shot classification with 339$\times$ fewer trainable parameters, and reduces sequential forgetting to $\sim$5\% where strong baselines degrade by $>$20-50\%.
title Sparse Spectral LoRA: Routed Experts for Medical VLMs
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.01310