This study systematically evaluates LLMs' ability in time series anomaly detection, revealing they mainly rely on visual understanding with limited reasoning capacity.

Zihao Zhou, Rose Yu

TidalDecode enhances LLM decoding speed by 2.1× using position persistent sparse attention, maintaining near full-attention performance.

Lijie Yang, Zhihao Zhang, Zhuofu Chen et al.

LLaMEA-HPO integrates HPO to reduce LLM query costs, achieving comparable or superior algorithm performance.

Niki van Stein, Diederick Vermetten, Thomas Bäck

Diffusion models enable high-fidelity 3D shape generation via iterative noise addition and removal, outperforming traditional methods.

Zhen Wang, Dongyuan Li, Yaozu Wu et al.

Proposed ActiView benchmark evaluates active perception in multimodal LLMs via view shifting and zooming, revealing significant performance gaps.

Ziyue Wang, Chi Chen, Fuwen Luo et al.

Casablanca created a 48-hour multi-dialect Arabic speech dataset with manual annotations, evaluating multilingual models’ adaptability and performance.

Bashar Talafha, Karima Kadaoui, Samar Mohamed Magdy et al.

SparseVLM uses text-guided, training-free visual token sparsification, reducing FLOPs by 54% with only 1% accuracy drop.

Yuan Zhang, Chun-Kai Fan, Junpeng Ma et al.

Unsupervised framework using persistent entropy and topological preservation to detect topological changes in MNIST data streams.

Sebastian Basterrech

MetricX-24 employs multi-stage training and synthetic data, significantly outperforming MetricX-23 in translation quality assessment.

Juraj Juraska, Daniel Deutsch, Mara Finkelstein et al.

Accelerated RPCholesky algorithm speeds up kernel matrix approximation by 40x using rejection sampling.

Ethan N. Epperly, Joel A. Tropp, Robert J. Webber

Introduces Lévy stochastic integrals to analyze discrete diffusion errors, providing the first KL divergence error bound for τ-leaping schemes.

Yinuo Ren, Haoxuan Chen, Grant M. Rotskoff et al.

Proposes Switchable Sparse-Dense Learning (SSD) to enhance pre-training efficiency, achieving up to 2x faster inference.

Zhengyan Zhang, Chaojun Xiao, Qiujieli Qin et al.

Quantitative bounds on support sparsity and bias in quadratically regularized OT using Minty trick, with rates \(\epsilon^{1/(2+d)}\).

Johannes Wiesel, Xingyu Xu

This paper proves that quadratically regularized optimal transport supports shrink at rate ε^{1/3}, revealing precise sparsity behavior for small regularization parameters.

Alberto González-Sanz, Marcel Nutz

Introduces Sm, a smoothness mechanism based on Dirichlet energy, to enhance local dependency modeling in MIL, significantly improving instance localization metrics.

Francisco M. Castro-Macías, Pablo Morales-Álvarez, Yunan Wu et al.

X-ALMA employs plug-and-play modules and ARPO optimization to achieve top translation quality across 50 languages, outperforming Aya-101 and Aya-23 on FLORES-200 and WMT'23 datasets.

Haoran Xu, Kenton Murray, Philipp Koehn et al.

Dynamic Sparse Training (DST) outperforms dense training in image corruption robustness at 10-50% sparsity, with resource savings.

Boqian Wu, Qiao Xiao, Shunxin Wang et al.

Using Gaussian Process regression to learn transformations like Cole-Hopf, simplifying nonlinear PDEs for easier solutions.

Jonghyeon Lee, Boumediene Hamzi, Yannis Kevrekidis et al.

HELMET offers seven task categories to evaluate long-context language models, improving existing benchmarks.

Howard Yen, Tianyu Gao, Minmin Hou et al.

ICR leverages attention pattern changes in LLMs for efficient zero-shot re-ranking, reducing latency by over 60%.

Shijie Chen, Bernal Jiménez Gutiérrez, Yu Su