InstaFlow achieves high-quality one-step text-to-image generation using Rectified Flow, with an FID of 22.4.

Xingchao Liu, Xiwen Zhang, Jianzhu Ma et al.

The study explores adversarial attacks on aligned language models using methods like perplexity detection.

Neel Jain, Avi Schwarzschild, Yuxin Wen et al.

Introduced Epsilon Scaling to reduce exposure bias in diffusion models, achieving 2.17 FID on CIFAR-10.

Mang Ning, Mingxiao Li, Jianlin Su et al.

Nougat uses a Visual Transformer to convert academic PDFs into lightweight markup, significantly improving semantic retention of mathematical expressions.

Lukas Blecher, Guillem Cucurull, Thomas Scialom et al.

OmniQuant employs learnable clipping and transformation to enable high-performance low-bit quantization of LLMs, achieving superior results with minimal training.

Wenqi Shao, Mengzhao Chen, Zhaoyang Zhang et al.

This paper critically evaluates multivariate time series anomaly detection, exposes flaws in point-adjust evaluation protocol, and demonstrates PCA-based baseline surpassing deep learning models on benchmarks.

Mohamed El Amine Sehili, Zonghua Zhang

ExpeL enhances LLM decision-making by learning from experiences in natural language without parameter updates.

Andrew Zhao, Daniel Huang, Quentin Xu et al.

LLM4TS employs a two-stage fine-tuning and multi-scale encoding to enhance pre-trained GPT-2 for efficient time-series forecasting, outperforming SOTA in limited data scenarios.

Ching Chang, Wei-Yao Wang, Wen-Chih Peng et al.

DGDFEM addresses delayed feedback using dynamic graph neural networks, enhancing data freshness and label accuracy.

Xiaolin Zheng, Zhongyu Wang, Chaochao Chen et al.

PDE-Refiner employs multi-step denoising inspired by diffusion models to achieve accurate long-term PDE predictions, outperforming state-of-the-art neural and hybrid models.

Phillip Lippe, Bastiaan S. Veeling, Paris Perdikaris et al.

Proposes a certified multi-fidelity zeroth-order optimization algorithm MFDOO variant with near-optimal cost bounds based on the Sβ,L(f,ε) metric.

Étienne de Montbrun, Sébastien Gerchinovitz

OPTIMIS method improves efficiency by 3.5x and accuracy by 3x in high-dimensional SRAM evaluation.

Yanfang Liu, Guohao Dai, Wei W. Xing

QuIP introduces incoherence-based post-training 2-bit quantization for LLMs, with theoretical guarantees and state-of-the-art empirical results.

Jerry Chee, Yaohui Cai, Volodymyr Kuleshov et al.

WebAgent combines HTML-T5 planning with Flan-U-PaLM code synthesis, raising real-world web-task success by over 50%.

Izzeddin Gur, Hiroki Furuta, Austin Huang et al.

The Android in the Wild (AITW) dataset offers 715k device interaction examples across multiple Android versions and device types.

Christopher Rawles, Alice Li, Daniel Rodriguez et al.

FlashAttention-2 achieves 2× speedup over FlashAttention by optimizing work partitioning, reaching 50-73% of GPU FLOPs utilization during training GPT-style models.

Tri Dao

Ecosystem analysis reveals systemic failures and homogeneous biases across deployed ML models on 11 datasets, highlighting limited improvements on systemic failures over time.

Connor Toups, Rishi Bommasani, Kathleen A. Creel et al.

Proposes <projektor> framework combining optimal transport and neural scaling laws for performance prediction and data selection, improving accuracy and efficiency.

Feiyang Kang, Hoang Anh Just, Anit Kumar Sahu et al.

Sparse Model Soups (SMS) improves pruning by averaging models, enhancing generalization and OOD performance while maintaining sparsity.

Max Zimmer, Christoph Spiegel, Sebastian Pokutta

This review summarizes recent advances in Optimal Transport (2012-2023), covering theory, algorithms, and applications in machine learning.

Eduardo Fernandes Montesuma, Fred Ngolè Mboula, Antoine Souloumiac