Fudan University

Observing that VecSet latent tokens retain recoverable spatial correspondences, the work proposes Position Forcing: during denoising it recovers token positions from the predicted clean latent, quantizes them at progressively finer resolutions according to the noise level, and feeds them back to the diffusion Transformer through 3D RoPE, providing coarse-to-fine spatial guidance for single-stage 3D generation without a separate position generation stage and achieving best or tied-best results on all four ULIP and Uni3D metrics.

Observing that VecSet latent tokens retain recoverable spatial correspondences, the work proposes Position Forcing: during denoising it recovers token positions from the predicted clean latent, quantizes them at progressively finer resolutions according to the noise level, and feeds them back to the diffusion Transformer through 3D RoPE, providing coarse-to-fine spatial guidance for single-stage 3D generation without a separate position generation stage and achieving best or tied-best results on all four ULIP and Uni3D metrics.

The study presents StrokeAgent, a multimodal agentic decision-support framework built around temporal alignment and continuous information synthesis that processes clinical records, laboratory results, raw electrocardiographic (ECG) waveforms and source multimodal CT images to support emergency triage, the initial thrombolysis decision and final reperfusion planning, and in a curated set of 100 patients with multimodal clinical and imaging data enriched for clinically complex reperfusion scenarios it achieved 86% concordance with expert-adjudicated final decisions, exceeding the mean of three tool-free large language models by 13 percentage points and the mean of six resident and attending neurologists by 17 percentage points, with 76% complete pathway concordance, completing the full deci

The study presents StrokeAgent, a multimodal agentic decision-support framework built around temporal alignment and continuous information synthesis that processes clinical records, laboratory results, raw electrocardiographic (ECG) waveforms and source multimodal CT images to support emergency triage, the initial thrombolysis decision and final reperfusion planning, and in a curated set of 100 patients with multimodal clinical and imaging data enriched for clinically complex reperfusion scenarios it achieved 86% concordance with expert-adjudicated final decisions, exceeding the mean of three tool-free large language models by 13 percentage points and the mean of six resident and attending neurologists by 17 percentage points, with 76% complete pathway concordance, completing the full deci

The work proposes UniField, a unified multi-modality, multi-task MRI field-strength enhancement framework that consolidates T1, T2, and FLAIR modalities and cross-field tasks such as 64mT-to-3T and 3T-to-7T into a single model, embeds 3D structural priors by operating in the latent space of the pretrained video super-resolution model FlashVSR with LoRA fine-tuning, and introduces a Field-Aware Spectral Rectification Mechanism (FASRM) that adjusts low-, mid-, and high-frequency loss weights according to the physical properties of the source and target fields; it also organizes and publicly releases a paired multi-field MRI dataset from five institutions that is an order of magnitude larger than existing benchmarks, reporting average improvements of about 1.81 dB in PSNR and 9.

The work proposes UniField, a unified multi-modality, multi-task MRI field-strength enhancement framework that consolidates T1, T2, and FLAIR modalities and cross-field tasks such as 64mT-to-3T and 3T-to-7T into a single model, embeds 3D structural priors by operating in the latent space of the pretrained video super-resolution model FlashVSR with LoRA fine-tuning, and introduces a Field-Aware Spectral Rectification Mechanism (FASRM) that adjusts low-, mid-, and high-frequency loss weights according to the physical properties of the source and target fields; it also organizes and publicly releases a paired multi-field MRI dataset from five institutions that is an order of magnitude larger than existing benchmarks, reporting average improvements of about 1.81 dB in PSNR and 9.

The work proposes BiasRecon, a bias-calibrated adaptation framework grounded in a minimal-intervention principle, which alternates frequency-guided prior calibration (learnable scalars α and β modulating low- and high-frequency components at each U-Net skip layer, adding only 2L parameters), score-based denoising, and adaptive regularization that uses Stein's Unbiased Risk Estimator (SURE) to tune the regularization parameter γ; trained on FastMRI knee (973 volumes) and evaluated without retraining on FastMRI-Brain (cross-anatomy), Stanford Knee (cross-center) and CMRxRecon (combined shift) under 8× Gaussian 1D undersampling, it improves PSNR over DDS by +1.19 dB, +2.06 dB and +1.60 dB respectively, and is best in 10 of 12 sampling/acceleration settings.

The work proposes BiasRecon, a bias-calibrated adaptation framework grounded in a minimal-intervention principle, which alternates frequency-guided prior calibration (learnable scalars α and β modulating low- and high-frequency components at each U-Net skip layer, adding only 2L parameters), score-based denoising, and adaptive regularization that uses Stein's Unbiased Risk Estimator (SURE) to tune the regularization parameter γ; trained on FastMRI knee (973 volumes) and evaluated without retraining on FastMRI-Brain (cross-anatomy), Stanford Knee (cross-center) and CMRxRecon (combined shift) under 8× Gaussian 1D undersampling, it improves PSNR over DDS by +1.19 dB, +2.06 dB and +1.60 dB respectively, and is best in 10 of 12 sampling/acceleration settings.

The work proposes MedQ-Engine, a closed-loop data engine that iterates through evaluate-explore-evolve phases: it clusters model failures on a development set into failure prototypes, uses the visual component of those prototypes as retrieval anchors over a roughly one-million-image pool spanning five modalities, and combines progressive human-in-the-loop annotation with entropy-guided routing and quality-assured fine-tuning, so that with only 10K annotations an 8B-parameter model surpasses GPT-4o by over 13% on medical image quality assessment and narrows the gap to human experts to 4.34%, with more than 4x sample efficiency over random sampling.

The work proposes MedQ-Engine, a closed-loop data engine that iterates through evaluate-explore-evolve phases: it clusters model failures on a development set into failure prototypes, uses the visual component of those prototypes as retrieval anchors over a roughly one-million-image pool spanning five modalities, and combines progressive human-in-the-loop annotation with entropy-guided routing and quality-assured fine-tuning, so that with only 10K annotations an 8B-parameter model surpasses GPT-4o by over 13% on medical image quality assessment and narrows the gap to human experts to 4.34%, with more than 4x sample efficiency over random sampling.

The work proposes SEER, a framework that curates the skill-tagged, image-grounded reasoning-trace dataset SEER-Trace (22,330 multimodal instruction instances from 1,811 cases), extracts anatomical evidence and synthesizes an executable task specification at inference, and uses SEER-Loop to distill high-reward reasoning episodes into reusable skills stored in SEER-Bank, thereby improving accuracy and stability in free-text-promptable 3D medical image segmentation, reporting an 81.94% reduction in performance variance and an 18.60% improvement in worst-case Dice under linguistic perturbations.

The work proposes SEER, a framework that curates the skill-tagged, image-grounded reasoning-trace dataset SEER-Trace (22,330 multimodal instruction instances from 1,811 cases), extracts anatomical evidence and synthesizes an executable task specification at inference, and uses SEER-Loop to distill high-reward reasoning episodes into reusable skills stored in SEER-Bank, thereby improving accuracy and stability in free-text-promptable 3D medical image segmentation, reporting an 81.94% reduction in performance variance and an 18.60% improvement in worst-case Dice under linguistic perturbations.