Alias-Free ViT: Fractional Shift Invariance via Linear Attention

DeepBlip is a novel neural framework designed to estimate conditional average treatment effects over time using structural nested mean models (SNMMs). This approach allows for the decomposition of treatment sequences into localized, time-specific 'blip effects', enhancing interpretability and enabling efficient evaluation of treatment policies. DeepBlip integrates sequential neural networks like LSTMs and transformers, addressing the limitations of existing methods by allowing simultaneous learning of all blip functions.

A novel approach to enhancing seismic foundation models has been introduced, synergizing multigrid algorithms with vision transformers. This method addresses the unique characteristics of seismic data, which require specialized processing techniques. The proposed adaptive two-grid foundation model training strategy (ADATG) utilizes Hilbert encoding to effectively capture both high- and low-frequency features in seismogram data, improving the efficiency of seismic data analysis and model training.

Recent research highlights the potential of task arithmetic for editing pre-trained closed-vocabulary models, particularly in image classification. This study investigates task addition in closed-vocabulary models, revealing that weight disentanglement is a common outcome of pre-training. The findings suggest that closed-vocabulary vision transformers can be effectively modified using task arithmetic, leading to enhanced multi-task model deployment capabilities.

The paper introduces the attention-indexed model (AIM), a framework for analyzing learning in deep attention layers. AIM captures the emergence of token-level outputs from bilinear interactions over high-dimensional embeddings. It allows full-width key and query matrices, aligning with practical transformers. The study derives predictions for Bayes-optimal generalization error and identifies phase transitions based on sample complexity, model width, and sequence length, proposing a message passing algorithm and demonstrating optimal performance via gradient descent.

CLAReSNet, a new hybrid architecture for hyperspectral image classification, integrates multi-scale convolutional extraction with transformer-style attention through an adaptive latent bottleneck. This model addresses challenges such as high spectral dimensionality, complex spectral-spatial correlations, and limited training samples with severe class imbalance. By combining convolutional networks and transformers, CLAReSNet aims to enhance classification accuracy and efficiency in hyperspectral imaging applications.

Higher-order Neural Additive Models (HONAMs) have been introduced as an advancement over Neural Additive Models (NAMs), which are known for their predictive performance and interpretability. HONAMs address the limitation of NAMs by effectively capturing feature interactions of arbitrary orders, enhancing predictive accuracy while maintaining interpretability, crucial for high-stakes applications. The source code for HONAM is publicly available on GitHub.

Vision-Language Models (VLMs) are emerging models that integrate visual content with natural language. Current methods typically fuse data either early in the encoding process or late through pooled embeddings. This paper introduces a lightweight fusion module utilizing cross-only, bidirectional attention layers to align hidden states from both modalities, enhancing understanding while keeping encoders non-causal. The proposed method aims to improve the performance of VLMs by leveraging the inherent structure of visual and textual data.

Artificial intelligence (AI) in media has seen rapid advancements over the past decade, particularly with the introduction of Generative Adversarial Networks (GANs) and diffusion models, which have enhanced photorealistic image generation. However, these developments have also led to challenges in distinguishing between real and synthetic content, as evidenced by the rise of deepfakes. Many detection models utilizing deep learning methods like Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have been created, but they often struggle with generalization and multimodal data.