Convolutional neural networks (CNNs) trained on object recognition achieve high task performance but continue to exhibit vulnerability under a range of visual perturbations and out-of-domain images, when compared with biological vision. Prior work has demonstrated that coupling a standard CNN with a front-end (VOneBlock) that mimics the primate primary visual cortex (V1) can improve overall model robustness. Expanding on this, we introduce Early Vision Networks (EVNets), a new class of hybrid CNNs that combine the VOneBlock with a novel SubcorticalBlock, whose architecture draws from computational models in neuroscience and is parameterized to maximize alignment with subcortical responses reported across multiple experimental studies. Without being optimized to do so, the assembly of the SubcorticalBlock with the VOneBlock improved V1 alignment across most standard V1 benchmarks, and better modeled extra-classical receptive field phenomena. In addition, EVNets exhibit stronger emergent shape bias and outperform the base CNN architecture by 9.3\% on an aggregate benchmark of robustness evaluations, including adversarial perturbations, common corruptions, and domain shifts. Finally, we show that EVNets can be further improved when paired with a state-of-the-art data augmentation technique, surpassing the performance of the isolated data augmentation approach by 6.2\% on our robustness benchmark. This result reveals complementary benefits between changes in architecture to better mimic biology and training-based machine learning approaches.

The hypothesis that Convolutional Neural Networks (CNNs) are inherently texture-biased has shaped much of the discourse on feature use in deep learning. We revisit this hypothesis by examining limitations in the cue-conflict experiment by Geirhos et al. To address these limitations, we propose a domain-agnostic framework that quantifies feature reliance through systematic suppression of shape, texture, and color cues, avoiding the confounds of forced-choice conflicts. By evaluating humans and neural networks under controlled suppression conditions, we find that CNNs are not inherently texture-biased but predominantly rely on local shape features. Nonetheless, this reliance can be substantially mitigated through modern training strategies or architectures (ConvNeXt, ViTs). We further extend the analysis across computer vision, medical imaging, and remote sensing, revealing that reliance patterns differ systematically: computer vision models prioritize shape, medical imaging models emphasize color, and remote sensing models exhibit a stronger reliance on texture.

The lawsuit, which was filed Thursday, claims that the AI company unlawfully crawls, scrapes, copies, and distributes CNN's content from CNN Digital Platforms and third-party platforms. It also accuses the AI tools of reproducing verbatim copies of its articles, including paywalled stories, in query responses to users. Perplexity's AI tools allegedly have incorrectly attributed hallucinated content to CNN, which the company says in the suit violates its trademark. CNN's lawsuit stands for the proposition that Perplexity, a company valued at tens of billions of dollars, should not be able to steal from entities that create the original content Perplexity exploits, a CNN spokesperson said in a statement to the outlet. The public rely on high quality news journalism reported by human beings to understand their world, which is frequently dangerous and expensive to produce.

The complaint, filed on Thursday, said that Perplexity unlawfully copied thousands of CNN stories, videos and images to power its products and distribute "identical or substantially similar" competing content. CNN is asking for an unspecified amount of monetary damages and a court order blocking Perplexity from violating its intellectual property rights. "CNN's lawsuit stands for the proposition that Perplexity, a company valued at tens of billions of dollars, should not be able to steal from entities that create the original content Perplexity exploits," the Warner Bros-owned news company said in a statement. Anthropic was the first AI company to settle one of these cases last year, agreeing to pay $1.5bn to resolve a class action lawsuit from a group of authors. Perplexity is also facing lawsuits from The New York Times, Reddit and Dow Jones, among others.

Neural Information Processing Systems http://nips.cc/

This paper introduces a concept of layer aggregation to describe how information from previous layers can be reused to better extract features at the current layer. While DenseNet is a typical example of the layer aggregation mechanism, its redundancy has been commonly criticized in the literature. This motivates us to propose a very light-weighted module, called recurrent layer aggregation (RLA), by making use of the sequential structure of layers in a deep CNN. Our RLA module is compatible with many mainstream deep CNNs, including ResNets, Xception and MobileNetV2, and its effectiveness is verified by our extensive experiments on image classification, object detection and instance segmentation tasks. Specifically, improvements can be uniformly observed on CIFAR, ImageNet and MSCOCO datasets, and the corresponding RLA-Nets can surprisingly boost the performances by 2-3% on the object detection task. This evidences the power of our RLA module in helping main CNNs better learn structural information in images.

We extra define the following notations for the proof. In Assumption 3.2, we assume the Lipschitz continuity and smoothness for all the activation functions. In the proof of lemmas, e.g., Lemma B.1 and B.2, we only use the fact that they are Lipschitz continuous and smooth, as well as bounded by a constant 0 > 0 at point 0, hence we use () to denote all the activation functions like what we do in Assumption 3.2 for simplicity. Additionally, in the following we introduce notations of the derivatives, mainly used in the proof of Lemma B.1 and Lemma B.2. By definition of feedforward neural networks in Section 2, different from the standard neural networks such as FCNs and CNNs in which the connection between neurons are generally only in adjacent layers, the neurons in feedforward neural networks can be arbitrarily connected as long as there is no loop.

Shift invariance is a critical property of CNNs that improves performance on classification. However, we show that invariance to circular shifts can also lead to greater sensitivity to adversarial attacks. We first characterize the margin between classes when a shift-invariant linear classifier is used. We show that the margin can only depend on the DC component of the signals. Then, using results about infinitely wide networks, we show that in some simple cases, fully connected and shift-invariant neural networks produce linear decision boundaries. Using this, we prove that shift invariance in neural networks produces adversarial examples for the simple case of two classes, each consisting of a single image with a black or white dot on a gray background. This is more than a curiosity; we show empirically that with real datasets and realistic architectures, shift invariance reduces adversarial robustness. Finally, we describe initial experiments using synthetic data to probe the source of this connection.