Large-scale pretraining datasets drive the success of large language models (LLMs). However, these web-scale corpora inevitably contain large amounts of noisy data due to unregulated web content or randomness inherent in data. Although LLM pretrainers often speculate that such noise contributes to instabilities in large-scale LLM pretraining and, in the worst cases, loss divergence, this phenomenon remains poorly understood.In this work, we present a systematic empirical study of whether noisy data causes LLM pretraining divergences and how it does so. By injecting controlled synthetic uniformly random noise into otherwise clean datasets, we analyze training dynamics across model sizes ranging from 480M to 5.2B parameters. We show that noisy data indeed induces training loss divergence, and that the probability of divergence depends strongly on the noise type, amount of noise, and model scale. We further find that noise-induced divergences exhibit activation patterns distinct from those caused by high learning rates, and we provide diagnostics that differentiate these two failure modes. Together, these results provide a large-scale, controlled characterization of how noisy data affects loss divergence in LLM pretraining.
Machine learning classifiers rely on loss functions for performance evaluation, often on a private (hidden) dataset. In a recent line of research, label inference was introduced as the problem of reconstructing the ground truth labels of this private dataset from just the (possibly perturbed) cross-entropy loss function values evaluated at chosen prediction vectors (without any other access to the hidden dataset). In this paper, we formally study the necessary and sufficient conditions under which label inference is possible from \emph{any} (noisy) loss function value. Using tools from analyti…
Direct decoding for task-oriented dialogue is known to suffer from the explaining-away effect, manifested in models that prefer short and generic responses. Here we argue for the use of Bayes' theorem to factorize the dialogue task into two models, the distribution of the context given the response, and the prior for the response itself. This approach, an instantiation of the noisy channel model, both mitigates the explaining-away effect and allows the principled incorporation of large pretrained models for the response prior. We present extensive experiments showing that a noisy channel model…
How does self-supervised pretraining improve robustness against noisy labels across various medical image classification datasets?
Center for Imaging Science, RIT（罗切斯特理工学院成像科学中心）；University of Aberdeen（阿伯丁大学）；NepAl Applied Mathematics and Informatics Institute for Research (NAAMII)（尼泊尔应用数学与信息学研究所）；Biomedical Engineering, RIT（罗切斯特理工学院生物医学工程系）
University of Edinburgh（爱丁堡大学）；WS AI Labs（WS人工智能实验室）；New York University（纽约大学）
Advances in neural modeling have achieved state-of-the-art (SOTA) results on public natural language processing (NLP) benchmarks, at times surpassing human performance. However, there is a gap between public benchmarks and real-world applications where noise, such as typographical or grammatical mistakes, is abundant and can result in degraded performance. Unfortunately, works which evaluate the robustness of neural models on noisy data and propose improvements, are limited to the English language. Upon analyzing noise in different languages, we observe that noise types vary greatly across lan…
Prior works have found it beneficial to combine provably noise-robust loss functions e.g., mean absolute error (MAE) with standard categorical loss function e.g. cross entropy (CE) to improve their learnability. Here, we propose to use Jensen-Shannon divergence as a noise-robust loss function and show that it interestingly interpolate between CE and MAE with a controllable mixing parameter. Furthermore, we make a crucial observation that CE exhibit lower consistency around noisy data points. Based on this observation, we adopt a generalized version of the Jensen-Shannon divergence for multiple…
Learning to Forget with Information Divergence Reweighted Objectives for Noisy Labels
Department of Mathematics（数学系）；Texas State University（德克萨斯州立大学）；School of Information Systems and Management（信息系统与管理学院）；University of South Florida（佛罗里达州立大学）
Pseudo-Quantized Actor-Critic Algorithm for Robustness to Noisy Temporal Difference Error
Principles of Informatics Research Division, National Institute of Informatics (NII)（国立信息学研究所信息学原理研究部）；Informatics Program, Graduate Institute for Advanced Studies (SOKENDAI)（综合研究大学院大学信息学项目）
Speaker recognition, recognizing speaker identities based on voice alone, enables important downstream applications, such as personalization and authentication. Learning speaker representations, in the context of supervised learning, heavily depends on both clean and sufficient labeled data, which is always difficult to acquire. Noisy unlabeled data, on the other hand, also provides valuable information that can be exploited using self-supervised training methods. In this work, we investigate how to pretrain speaker recognition models by leveraging dialogues between customers and smart-speaker…
Contrastive pretraining techniques for text classification has been largely studied in an unsupervised setting. However, oftentimes labeled data from related tasks which share label semantics with current task is available. We hypothesize that using this labeled data effectively can lead to better generalization on current task. In this paper, we propose a novel way to effectively utilize labeled data from related tasks with a graph based supervised contrastive learning approach. We formulate a token-graph by extrapolating the supervised information from examples to tokens. Our formulation res…
OpenAI（OpenAI（开放人工智能研究中心））；University of British Columbia（不列颠哥伦比亚大学）
Pretraining on noisy, internet-scale datasets has been heavily studied as a technique for training models with broad, general capabilities for text, images, and other modalities. However, for many sequential decision domains such as robotics, video games, and computer use, publicly available data does not contain the labels required to train behavioral priors in the same way. We extend the internet-scale pretraining paradigm to sequential decision domains through semi-supervised imitation learning wherein agents learn to act by watching online unlabeled videos. Specifically, we show that with …
We show when maximizing a properly defined $f$-divergence measure with respect to a classifier's predictions and the supervised labels is robust with label noise. Leveraging its variational form, we derive a nice decoupling property for a family of $f$-divergence measures when label noise presents, where the divergence is shown to be a linear combination of the variational difference defined on the clean distribution and a bias term introduced due to the noise. The above derivation helps us analyze the robustness of different $f$-divergence functions. With established robustness, this family o…
Robust Physics-Informed Neural Network Approach for Estimating Heterogeneous Elastic Properties from Noisy Displacement Data
Department of Industrial and Systems Engineering, Texas A&M University（工业与系统工程系，德克萨斯大学阿马尔科分校）；Department of Mechanical Engineering, Texas A&M University（机械工程系，德克萨斯大学阿马尔科分校）
Indraprastha Institute of Information Technology Delhi（英迪拉普拉斯塔信息技术学院德里分校）；National Institute of Technology Rourkela（鲁尔克拉国家理工学院）；National Institute of Technology Jamshedpur（詹谢普尔国家理工学院）；University of Central Florida（中佛罗里达大学）
Deep learning has shown remarkable progress in a wide range of problems. However, efficient training of such models requires large-scale datasets, and getting annotations for such datasets can be challenging and costly. In this work, we explore the use of user-generated freely available labels from web videos for video understanding. We create a benchmark dataset consisting of around 2 million videos with associated user-generated annotations and other meta information. We utilize the collected dataset for action classification and demonstrate its usefulness with existing small-scale annotated…
LiLAW: Lightweight Learnable Adaptive Weighting to Learn Sample Difficulty & Improve Noisy Training
Department of Computer Science（计算机科学系）；University of Toronto（多伦多大学）；Department of Mathematics（数学系）；The Hospital for Sick Children（圣·玛利亚医院）；Department of Statistics（统计学系）；UHN KITE Research Institute（UHN KITE研究所）；T-CAIREM；Vector Institute（向量研究所）；Institute of Biomedical Engineering（生物医学工程研究所）；Rehabilitation Sciences Institute（康复科学研究所）
Deep learning with noisy labels is challenging as deep neural networks have the high capacity to memorize the noisy labels. In this paper, we propose a learning algorithm called Co-matching, which balances the consistency and divergence between two networks by augmentation anchoring. Specifically, we have one network generate anchoring label from its prediction on a weakly-augmented image. Meanwhile, we force its peer network, taking the strongly-augmented version of the same image as input, to generate prediction close to the anchoring label. We then update two networks simultaneously by sele…
We use Generative Adversarial Networks (GANs) to design a class conditional label noise (CCN) robust scheme for binary classification. It first generates a set of correctly labelled data points from noisy labelled data and 0.1% or 1% clean labels such that the generated and true (clean) labelled data distributions are close; generated labelled data is used to learn a good classifier. The mode collapse problem while generating correct feature-label pairs and the problem of skewed feature-label dimension ratio ($\sim$ 784:1) are avoided by using Wasserstein GAN and a simple data representation c…
Uncertainty Quantification via Hölder Divergence for Multi-View Representation Learning
Shenzhen MSU-BIT University（深圳北理莫斯科大学）；Minzu University of China（中央民族大学）；Beijing Institute of Technology（北京理工大学）；Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ)（广东省人工智能与数字经济实验室（深圳））
What is the information leakage of an iterative randomized learning algorithm about its training data, when the internal state of the algorithm is \emph{private}? How much is the contribution of each specific training epoch to the information leakage through the released model? We study this problem for noisy gradient descent algorithms, and model the \emph{dynamics} of Rényi differential privacy loss throughout the training process. Our analysis traces a provably \emph{tight} bound on the Rényi divergence between the pair of probability distributions over parameters of models trained on neigh…
FedRG: Unleashing the Representation Geometry for Federated Learning with Noisy Clients
Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）；TMLR Group, Hong Kong Baptist University（香港浸会大学TMLR组）；Beihang University（北京航空航天大学）；The Hong Kong University of Science and Technology（香港科技大学）
Self-Supervised Moving Object Segmentation of Sparse and Noisy Radar Point Clouds
CARIAD SE（CARIAD公司）；TU Dortmund University（杜伊斯堡-艾森大学）；University of Bonn（波恩大学）；Lamarr Institute for Machine Learning and Artificial Intelligence（拉马尔人工智能与机器学习研究所）
University of Macau（澳门大学）；Hong Kong Baptist University（香港浸会大学）；Agency for Science, Technology and Research (A*STAR)（新加坡科技研究局）
Federated learning (FL) collaboratively trains a shared global model depending on multiple local clients, while keeping the training data decentralized in order to preserve data privacy. However, standard FL methods ignore the noisy client issue, which may harm the overall performance of the shared model. We first investigate critical issue caused by noisy clients in FL and quantify the negative impact of the noisy clients in terms of the representations learned by different layers. We have the following two key observations: (1) the noisy clients can severely impact the convergence and perfor…
Non-Linear Strong Data-Processing for Quantum Hockey-Stick Divergences
Department of Mathematics and Illinois Quantum Information Science and Technology (IQUIST) Center, University of Illinois Urbana-Champaign（数学系和伊利诺伊大学量子信息科学与技术中心，伊利诺伊大学厄巴纳-香槟分校）；Centre for Quantum Technologies, National University of Singapore（新加坡国立大学量子技术中心）；Institute for Information Processing (tnt/L3S), Leibniz Universität Hannover（汉诺威莱布尼茨信息处理研究所（tnt/L3S））
Privacy Loss of Noisy Stochastic Gradient Descent Might Converge Even for Non-Convex Losses
McMaster University（麦克马斯特大学）；Instituto de Investigaciones en Matemáticas Aplicadas y en Sistemas, Universidad Nacional Autónoma de México（墨西哥国立自治大学应用数学与系统研究所）
