Recent studies on Neural Collapse (NC) reveal that, under class-balanced conditions, the class feature means and the classifier weights spontaneously align into a simplex equiangular tight frame (ETF). In long-tailed regimes, however, severe sample imbalance tends to prevent the emergence of the NC phenomenon, resulting in poor generalization performance.Current efforts predominantly seek to recover the ETF geometry by imposing constraints on features or classifier weights, yet overlook a critical problem: There is a pronounced misalignment between the feature and the classifier weight spaces. In this paper, we theoretically quantify the harm of such misalignment through an optimal error exponent analysis.Built on this insight, we propose three explicit alignment strategies that plug-and-play into existing long-tail methods without architectural change. Extensive experiments on the CIFAR-10-LT, CIFAR-100-LT, and ImageNet-LT datasets consistently boost examined baselines and achieve the state-of-the-art performances.

Current methods for editing pre-trained models face significant challenges, primarily high computational costs and limited scalability. Task arithmetic has recently emerged as a promising solution, using simple arithmetic operations—addition and negation—based on task vectors which are the differences between fine-tuned and pre-trained model weights, to efficiently modify model behavior. However, the full potential of task arithmetic remains underexplored, primarily due to limited mechanisms for overcoming optimization stagnation. To address this challenge, we introduce the notion of difference vector, a generalized form of task vectors derived from the historical movements during optimization. Using difference vectors as directed perturbations, we proposed the Difference Vector-based Anisotropic Scaling Iterative algorithm (DV-BASI) to enable a continuous optimization process for task arithmetic methods without relying on any additional modules or components. Notably, by leveraging escapability and directional advantages of difference vectors, the average performance on different tasks of the multi-task model merged by DV-BASI may even outperform models individually fine-tuned. Based on this observation, we extend the application of difference vectors to a feasible fine-tuning method for single-task models. On the practical side, DV-BASI allows expressive searching directions with few learnable parameters and forms a scalable framework. We also integrate DV-BASI with task arithmetic methods and advanced optimization techniques to achieve state-of-the-art performance on both supervised and unsupervised evaluation protocols.

We developed a deep learning approach to detect and segment nuclear pores in large high-resolution 3D FIB-SEM images. The approach extracts small blocks containing nuclear pore regions from different parts of a cell nucleus and applies data augmentation techniques combined with Random Block Sampling (RBS) to optimize training. By fine tuning the neural network and reducing the batch size to one, we enhanced the model’s ability to learn from individual image features, leading to more precise detection. Our Dynamic Cyclical Data Augmentation (DCDA) adapts in real-time with an automatic termination feature, reducing augmentation time by over 97.4% in our experiments, significantly improving efficiency without compromising model performance. Experimental results consistently demonstrate that our approach can quickly and accurately identify nuclear pores, achieving F1 scores from 0.7119 to 0.8555, with precision ranging from 0.7379 to 0.8516. Training time was also significantly reduced. These substantial improvements in both accuracy and efficiency underscore the method’s effectiveness for large-scale cellular analysis.

The historical, cultural, and intellectual importance of archiving the web has been widely recognized. Today, all countries with high Internet penetration rate have established high-profile archiving initiatives to crawl and archive the fast-disappearing web content for long-term use. As web technologies evolve, established web archiving techniques face challenges. This paper focuses on the potential impact of the relaxed consistency web design on crawler driven web archiving. Relaxed consistent websites may disseminate, albeit ephemerally, inaccurate and even contradictory information. If captured and preserved in the web archives as historical records, such information will degrade the overall archival quality. To assess the extent of such quality degradation, we build a simplified feed-following application and simulate its operation with synthetic workloads. The results indicate that a non-trivial portion of a relaxed consistency web archive may contain observable inconsistency, and the inconsistency window may extend significantly longer than that observed at the data store. We discuss the nature of such quality degradation and propose a few possible remedies.