This paper tests the assumption of how NVIDIA GPUs handle warp divergence and finds that the cost remains stable and predictable despite architectural changes.

A new reliability metric is proposed for DTW alignment paths based on agreement between DTW and FlexDTW.

This paper argues that the choice of divergence in statistical inference can be justified by requiring invariance to problem reformulations, leading to the selection of the Kullback-Leibler divergence…

This paper introduces Curvature-Weighted Gradient Diversity (CWGD), a geometry-aware measure for optimization noise that reduces the asymptotic optimization error floor by up to a factor of two compar…

The paper introduces Entropic Projection Alignment (EPA), a unified framework that estimates, explains, and improves model performance under distribution shift by aligning source and target distributi…

This paper identifies prediction bias, a failure mode of entropy minimization in test-time adaptation, and proposes Distribution Shift Bias Reduction (DSBR) to stabilize adaptation and prevent model c…

This paper derives the gradient of the Wolkowicz-Styan upper bound on the maximum eigenvalue of the cross-entropy loss Hessian in three-layer NNs to characterize directions leading to flat minima and…

This paper proves that the minimax lower bound of estimation for popular kernel discrepancies is n^-(1/2) on general topological spaces and under mild assumptions on the kernel.

The paper proposes a unified, constrained optimization framework using KL divergence and likelihood constraints to achieve effective and principled unlearning in diffusion models.

The paper introduces MENTIS, a geometry-first framework that measures how preference alignment structurally changes the internal computations of language models, finding that these changes are selecti…

While restricting a model to the theoretical Markov boundary can significantly improve prediction, the practical process of discovering and using this boundary is often computationally infeasible and…

This paper identifies Security-Recall Divergence (SRD), demonstrating that omission constraints (prohibitions) decay significantly in long-context LLM conversations, while commission constraints (requ…

The paper proposes EKSFT, a selective fine-tuning method that masks high-entropy or high-KL divergence tokens during Supervised Fine-Tuning (SFT) to prevent distribution shift and improve subsequent R…

VISReg introduces a novel regularization technique that combines variance control with a Sliced-Wasserstein-based sketching objective to stabilize self-supervised learning, achieving state-of-the-art…

This paper systematically characterises the sensitivity of emergent misalignment (EM) in LLMs to various training choices, finding that the choice of optimiser has the largest effect on misalignment r…

The paper characterizes 'dead-entry' TLB misses in GPUs, which occur when recently evicted translations are immediately re-walked, and proposes DEPOT, a Bloom filter mechanism that significantly reduc…