Large Language Models (LLMs) hold significant potential for enhancing healthcare applications, yet their deployment is hindered by high computational and memory demands. Model compression techniques offer solutions to reduce these demands, but their impact on medical LLMs remains underexplored. In this paper, we introduce CMedBench, the first comprehensive benchmark for evaluating compressed LLMs in medical contexts. CMedBench assesses five core dimensions: Medical Knowledge Ability, Medical Application Ability, Trustworthiness Maintenance, Compression Cross Combination, and Computational Efficiency. Through extensive empirical studies, we analyze the trade-offs between model efficiency and clinical performance across diverse models, datasets, and compression strategies. Our findings highlight critical limitations in current evaluation practices and provide a robust framework for aligning compression strategies with medical requirements. CMedBench serves as a vital resource for researchers and practitioners, guiding the development of efficient, trustworthy, and clinically effective LLMs for healthcare applications.

Large Vision-Language Models (VLMs) exhibit impressive multi-modal capabilities but suffer from prohibitive computational and memory demands, due to their long visual token sequences and massive parameter sizes. To address these issues, recent works have proposed training-free compression methods. However, existing efforts often suffer from three major limitations: (1) Current approaches do not decompose techniques into comparable modules, hindering fair evaluation across spatial and temporal redundancy. (2) Evaluation confined to simple single-turn tasks, failing to reflect performance in realistic scenarios. (3) Isolated use of individual compression techniques, without exploring their joint potential. To overcome these gaps, we introduce LLMC+, a comprehensive VLM compression benchmark with a versatile, plug-and-play toolkit. LLMC+ supports over 20 algorithms across five representative VLM families and enables systematic study of token-level and model-level compression. Our benchmark reveals that: (1) Spatial and temporal redundancies demand distinct technical strategies. (2) Token reduction methods degrade significantly in multi-turn dialogue and detail-sensitive tasks. (3) Combining token and model compression achieves extreme compression with minimal performance loss. We believe LLMC+ will facilitate fair evaluation and inspire future research in efficient VLM.

Federated learning (FL) enables privacy-preserving collaboration among distributed clients, but practical deployments often face heterogeneous models and non-IID data, leading to degraded communication and personalization. In addition, real-world FL systems frequently encounter newly joined clients that require rapid adaptation and abnormal clients that may upload corrupted updates, further exacerbating instability and hindering global convergence. To address these challenges in image classification, we propose HFedDGHN, a Heterogeneous Federated Dynamic Graph HyperNetwork that jointly models inter-client relations and personalized parameter generation. Specifically, a graph structure learner adaptively captures client correlations to construct a dynamic collaboration graph, while a graph-convolutional hypernetwork generates model parameters for heterogeneous architectures, enabling implicit knowledge transfer without sharing local data or weights. Moreover, the framework naturally supports meta-learning-based generalization, allowing efficient adaptation to newly joined clients. Furthermore, the dynamic graph enhances robustness by isolating abnormal clients, as they tend to be excluded from most neighborhoods during adaptive graph construction. Extensive experiments across multiple benchmarks demonstrate that HFedDGHN achieves superior accuracy compared to state-of-the-art personalized and heterogeneous FL methods, while naturally improving robustness and scalability in real-world deployments.

Filter pruning is one of the most popular approaches for compressing convolutional neural networks (CNNs). The most critical task in pruning is to evaluate the importance of each convolutional filter, such that the less important filters can be removed while the overall model performance is minimally affected. In each layer, some filters may be linearly dependent on each other, which means that they have replaceable information. Redundant information can be removed without significantly affecting information richness and model performance. In this paper, we propose a novel low-rank guided pruning scheme to obtain skeleton neural networks by alternatively training and pruning CNNs. In each step, training is performed with nuclear-norm regularization to low-rank the filters in each layer, followed by filter pruning to maintain the information richness via the maximally linearly independent subsystem. A novel “smaller-norm-and-linearly-dependent-less-important” pruning criterion is proposed to compress the model. The training and pruning processes can be repeated until the model is fully trained. To investigate the performance, we applied the proposed joint training and pruning scheme to train the CNNs for image classification. We considered three benchmark datasets: MNIST, CIFAR-10 and ILSVRC-2012. The proposed method successfully achieved a higher pruning rate and better classification performance compared to state-of-the-art compression methods.