arXiv：cs.LG· Taha Sabir, Alberto Montero Molina, Thinh Quang Ly, Veronika Dmitrenko, Mohammad Yaqub, Tausifa Jan Saleem·· 6 小时前
基于 Qwen3.5-9B 的任务专属 LoRA 适配器实现多模态医学图像解析
Multimodal Medical Image Parsing with Task-Specific Low-Rank Adapters
AI 导读
研究者用 Qwen3.5-9B 为 FLARE 2026 Task 3 构建任务专属 LoRA 适配器系统，覆盖六类任务、七种成像模态，采用冻结的 4-bit 共享骨干并按输入问题选择适配器。
正文
Abstract:Adapting a single vision-language model to medical image parsing requires training choices that reflect the demands of each task. We study task-specific adaptation with Qwen3.5-9B for FLARE 2026 Task 3, covering six task families and seven imaging modalities. Our system couples a shared, frozen 4-bit backbone with LoRA adapters selected from the input question. We customize adapter ranks and prompts, rebalance rare classification targets, and supplement regression training with dermatology classification examples. On the fixed internal holdout, the task-specific adapters achieve a dataset-macro balanced accuracy of 0.7791 for classification, a dataset-macro micro-F1 of 0.6286 for multi-label classification, a counting mean absolute error (MAE) of 11.73, a dataset-macro detection F1 of 0.8914, a dataset-macro regression MAE of 9.78, and a report-generation CRIMSON score of 0.8204. On the organizer's 1,703-example hidden-validation set, the submitted system achieves 0.8496 classification balanced accuracy, 0.5338 sample-averaged multi-label F1, and 14.5950 regression MAE. The final 30.1 GB container averages 3.1 seconds per case, including model loading, and produces identical outputs across repeated verification runs. The benefits vary across tasks, supporting the use of task-specific adapter configurations and image budgets within a shared medical vision-language model. Code is available at this https URL.
| Comments: | 17 pages, 4 figures. FLARE 2026 Task 3 methodology paper (2D track) |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.19901 [cs.CV] |
| (or arXiv:2610.19901v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19901 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Taha Sabir [view email]
[v1]
Wed, 14 Oct 2026 09:59:17 UTC (2,446 KB)
来源：arXiv：cs.LG · arxiv.org