paper-with-me

홈 › Papers

Noise Stability of Transformer Models

2026-02-09 · Themistoklis Haris, Zihan Zhang, Yuichi Yoshida arxiv

Understanding simplicity biases in deep learning offers a promising path toward developing reliable AI. A common metric for this, inspired by Boolean function analysis, is average sensitivity, which captures a model's robustness to single-token perturbations. We argue that average sensitivity has two key limitations: it lacks a natural generalization to real-valued domains and fails to explain the "junta-like" input dependence we empirically observe in modern LLMs. To address these limitations, we propose noise stability as a more comprehensive simplicity metric. Noise stability expresses a model's robustness to correlated noise applied to all input coordinates simultaneously. We provide a theoretical analysis of noise stability for single-layer attention and ReLU MLP layers and tackle the multi-layer propagation problem with a covariance interval propagation approach. Building on this theory, we develop a practical noise stability regularization method. Experiments on algorithmic and next-token-prediction tasks show that our regularizer consistently catalyzes grokking and accelerates training by approximately $35\%$ and $75\%$ respectively. Our results sculpt a new connection between signal propagation in neural networks and interpretability, with noise stability emerging as a powerful tool for understanding and improving modern Transformers.

📄 PDF Abstract BibTeX arXiv:2602.08287

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Reconstructing Richtmyer-Meshkov instabilities from noisy radiographs using low dimensional features and attention-based neural networks

2024-08-02 · Daniel A. Serino, Marc L. Klasky, Balasubramanya T. Nadiga, Xiaojian Xu 외

A trained attention-based transformer network can robustly recover the complex topologies given by the Richtmyer-Meshkoff instability from a sequence of hydrodynamic features derived from radiographic images corrupted wi…

DeepKoopFormer: A Koopman Enhanced Transformer Based Architecture for Time Series Forecasting

2025-08-04 · Ali Forootani, Mohammad Khosravi, Masoud Barati arxiv

Time series forecasting plays a vital role across scientific, industrial, and environmental domains, especially when dealing with high-dimensional and nonlinear systems. While Transformer-based models have recently achie…

Time Series Forecasting

Deep learning based on Transformer architecture for power system short-term voltage stability assessment with class imbalance

2023-10-18 · Yang Li, Jiting Cao, Yan Xu, Lipeng Zhu 외

Most existing data-driven power system short-term voltage stability assessment (STVSA) approaches presume class-balanced input data. However, in practical applications, the occurrence of short-term voltage instability fo…

ClusteringGenerative Adversarial NetworkSynthetic Data Generation

Most Transformer Modifications Still Do Not Transfer at 1-3B: A 2020-2026 Update to Narang et al. (2021) with Downstream Evaluation and a Noise Floor

2026-05-20 · Yang Zhao, Jiahao Lu, Bin Huang, Guhua Zhang 외 arxiv

Narang et al. (2021) evaluated 40+ Transformer modifications at T5-base scale and concluded that most did not transfer. Five years later, the typical working regime has moved to 1-3B parameters, downstream evaluation has…

Stability of Transformers under Layer Normalization

2025-10-10 · Kelvin Kan, Xingjian Li, Benjamin J. Zhang, Tuhin Sahai 외 arxiv

Despite their widespread use, training deep Transformers can be unstable. Layer normalization, a standard component, improves training stability, but its placement has often been ad-hoc. In this paper, we conduct a princ…