Large Language Models (LLMs) have shown strong potential in medical question answering, but their reliance on static parametric knowledge limits their ability to handle rapidly evolving clinical evidence. The retrieval system mitigates this issue by incorporating external medical knowledge, but flat retrieval can return lexically relevant but clinically misaligned evidence. We propose Med-Chain, a training-free medical evidence retrieval framework that reformulates retrieval as inspectable and repairable clinical evidence paths. Med-Chain first routes each question through a clinical topic tree to progressively narrow retrieval from broad domains to specific guideline sections. It then verifies whether the retrieved path matches the clinical intent and performs level-aware refinement to repair mismatches. Experiments on seven medical QA benchmarks show that Med-Chain consistently improves backbone LLMs and outperforms existing medical Retrieval-augmented generation baselines, highlighting the value of clinically structured and repairable evidence retrieval.

Practical deployment of medical vision-language models requires a single model that interprets heterogeneous clinical visual inputs: 2D images, 3D volumes, and medical videos. Yet how to train such unified models remains underexplored, as these modalities differ in spatial-temporal structure, token length, and data availability. We present OmniV-Med, a scalable training recipe built on four empirical findings. First, broad modality coverage requires balancing modality diversity with task-level supervision, motivating OmniV-Med-Instruct, a 252K corpus spanning 14 medical modalities and 11 clinical tasks. Second, we identify a modality-ordering effect: introducing 3D and video too early destabilizes optimization and weakens 2D grounding, motivating a progressive curriculum that first builds 2D alignment before extending to spatial-temporal inputs. Third, we observe an emergent spatiotemporal transfer effect: once strong 2D patch representations are in place, lightweight rotary position adaptation extrapolates them to 3D slices and video frames, letting a single 2D ViT replace modality-specific encoder stacks. Fourth, the imbalance in sequence length across modalities and the redundancy of long volumes and videos hinder scaling; our Redundancy-Guided Token Pruning (RGTP) exploits inter-slice and inter-frame redundancy to cut visual tokens by 60% while preserving performance. Across nine 2D, 3D, and video medical benchmarks, OmniV-Med achieves robust performance. Extensive component-wise ablations systematically refine the design space, yielding a simple yet efficient and scalable recipe for training high-performance unified Med-VLMs. Code is released at https://github.com/ZJUI-AI4H/Hulu-Med.

Vision–language models (VLMs) have demonstrated impressive capabilities in general medical visual question answering, yet it remains unclear whether their predictions reflect evidence-grounded clinical reasoning or reliance on spurious priors. We introduce Med-R2 Bench, a hierarchical benchmark aligned with the clinical workflow to evaluate adversarial robustness with visual grounding. We design stepwise QA tasks to assess whether reasoning chains are strictly grounded in visual evidence across the four clinical stages, and employ adversarial perturbations to test robustness against misleading cues. Med-R2 comprises 42,432 images, 31 task categories, and 110,406 QA pairs. Evaluation across 14 VLMs reveals a sequential performance degradation along the four-stage clinical workflow. Adversarial experiments show that models rely heavily on correct prompts to guess answers. Even when provided with explicit visual cues, the models struggle to accurately align textual descriptions. Finally, we demonstrate that stepwise fine-tuning using our hierarchical data significantly improves reasoning robustness, highlighting its potential to drive future improvements in evidence-based medical AI.