This paper investigates limitations of learning tanh neural networks under finite-precision computations and Lp accuracy guarantees.

The paper introduces a novel, non-deep neural network architecture that achieves the performance of LLMs by finding the global optimum of the loss function in a single, closed-form iteration, eliminat…

This paper addresses the vulnerability of DNNs used in robotic semantic segmentation to adversarial attacks by proposing specialized detection strategies to enhance safety in robotic perception system…

This paper explores how different components of the Transformer feedforward block architecture impact rank preservation across depth during initialization.

This paper characterizes the gap between neural network performance and their neural tangent kernel limit on compositional tasks, attributing it to a mismatch between kernel smoothness bias and target…

This paper compares and optimizes CUDA strategies for a shallow neural network, achieving a 1.41x speedup on a large dataset.

This paper discusses the current understanding of Large Language Models (LLMs), their capabilities, and their relationship to human cognition, with a focus on emerging capabilities and mechanistic imp…

The paper analyzes congruence-based neural architectures for classifying positive-definite matrices, demonstrating that common semi-orthogonality constraints severely limit the model's expressivity.

This paper demonstrates that Concept Bottleneck Models (CBMs), despite their interpretability, are highly vulnerable to targeted adversarial attacks that manipulate semantic concepts, and proposes SPE…

This paper proposes Hierarchical Block-Local Learning (HBLL), a framework for training deep neural networks without full end-to-end backpropagation, achieving $\mathcal{O}(\log N)$ parallel time compl…

The paper analyzes LLM vulnerability detection using mechanistic interpretability, finding that models primarily rely on safety detectors rather than direct vulnerability signature recognition.

This paper analyzes the computational complexity of verifying feedforward neural networks when their weights are restricted to finite-width arithmetic, finding that verification remains NP-complete fo…

An adversarial model trainer can plant statistically undetectable backdoors in deep neural networks, providing access to adversarial examples.

This paper provides the first algorithmic separation between constant-depth and logarithmic-depth networks, identifying a class of Boolean functions that logarithmic-depth networks can learn efficient…

This paper studies the expressivity of shallow polynomial neural networks with monomial activation functions over finite fields, quantifying it by the cardinality of the neuromanifold and deriving low…

The paper establishes that the training process of fully connected deep neural networks (DNNs) on exponential family data is mathematically equivalent to performing a Renormalization Group (RG) calcul…

This paper introduces a mechanistic neuronal network model for multilayer learning, offering biological insights and an alternative to backpropagation.