English
Related papers

Related papers: Initialization is Critical to Whether Transformers…

200 papers

Transformers have become the predominant architecture in foundation models due to their excellent performance across various domains. However, the substantial cost of scaling these models remains a significant concern. This problem arises…

Factorized layers--operations parameterized by products of two or more matrices--occur in a variety of deep learning contexts, including compressed model training, certain types of knowledge distillation, and multi-head self-attention…

Machine Learning · Statistics 2022-10-07 Mikhail Khodak , Neil Tenenholtz , Lester Mackey , Nicolò Fusi

The transformer architecture is central to the success of modern Large Language Models (LLMs), in part due to its surprising ability to perform a wide range of tasks - including mathematical reasoning, memorization, and retrieval - using…

Machine Learning · Computer Science 2025-09-05 Yihe Dong , Lorenzo Noci , Mikhail Khodak , Mufan Li

Although transformer-based models have shown exceptional empirical performance, the fundamental principles governing their training dynamics are inadequately characterized beyond configuration-specific studies. Inspired by empirical…

Machine Learning · Computer Science 2025-10-09 Zheng-An Chen , Tao Luo

State-of-the-art rehearsal-free continual learning methods exploit the peculiarities of Vision Transformers to learn task-specific prompts, drastically reducing catastrophic forgetting. However, there is a tradeoff between the number of…

Computer Vision and Pattern Recognition · Computer Science 2023-08-21 Thomas De Min , Massimiliano Mancini , Karteek Alahari , Xavier Alameda-Pineda , Elisa Ricci

Transformer architectures have achieved remarkable success across language, vision, and multimodal tasks, and there is growing demand for them to address in-context compositional learning tasks. In these tasks, models solve the target…

Machine Learning · Computer Science 2025-11-26 Wei Chen , Jingxi Yu , Zichen Miao , Qiang Qiu

The extent to which neural networks are able to acquire and represent symbolic rules remains a key topic of research and debate. Much current work focuses on the impressive capabilities of large language models, as well as their often…

Machine Learning · Computer Science 2025-06-11 Anna Langedijk , Jaap Jumelet , Willem Zuidema

Overparameterized models like deep neural networks have the intriguing ability to recover target functions with fewer sampled data points than parameters (see arXiv:2307.08921). To gain insights into this phenomenon, we concentrate on a…

Machine Learning · Computer Science 2024-05-24 Jiajie Zhao , Zhiwei Bai , Yaoyu Zhang

To produce accurate predictions, language models (LMs) must balance between generalization and memorization. Yet, little is known about the mechanism by which transformer LMs employ their memorization capacity. When does a model decide to…

Computation and Language · Computer Science 2023-02-14 Adi Haviv , Ido Cohen , Jacob Gidron , Roei Schuster , Yoav Goldberg , Mor Geva

In this paper, I introduce the retrieval problem, a simple yet common reasoning task that can be solved only by transformers with a minimum number of layers, which grows logarithmically with the input size. I empirically show that large…

Machine Learning · Computer Science 2025-10-29 Tiberiu Musat

Finding the right initialisation for neural networks is crucial to ensure smooth training and good performance. In transformers, the wrong initialisation can lead to one of two failure modes of self-attention layers: rank collapse, where…

Machine Learning · Statistics 2026-05-22 Alessio Giorlandino , Sebastian Goldt

This paper studies how the model architecture and data configurations influence the empirical memorization capacity of generative transformers. The models are trained using synthetic text datasets derived from the Systematized Nomenclature…

Computation and Language · Computer Science 2025-06-18 Anton Changalidis , Aki Härmä

Large language models exhibit sophisticated capabilities, yet understanding how they work internally remains a central challenge. A fundamental obstacle is that training selects for behavior, not circuitry, so many weight configurations can…

Machine Learning · Computer Science 2026-02-27 Joshua S. Schiffman

Transformer-based large language models have displayed impressive in-context learning capabilities, where a pre-trained model can handle new tasks without fine-tuning by simply augmenting the query with some input-output examples from that…

Machine Learning · Computer Science 2024-06-18 Hongkang Li , Meng Wang , Songtao Lu , Xiaodong Cui , Pin-Yu Chen

Transformers have achieved remarkable successes across a wide range of applications, yet the theoretical foundation of their model efficiency remains underexplored. In this work, we investigate how the model parameters -- mainly attention…

Machine Learning · Computer Science 2025-10-07 Ruoxi Yu , Haotian Jiang , Jingpu Cheng , Penghao Yu , Qianxiao Li , Zhong Li

Transformers pretrained via next token prediction learn to factor their world into parts, representing these factors in orthogonal subspaces of the residual stream. We formalize two representational hypotheses: (1) a representation in the…

Humans exhibit remarkable compositional reasoning by integrating knowledge from various sources. For example, if someone learns ( B = f(A) ) from one source and ( C = g(B) ) from another, they can deduce ( C=g(B)=g(f(A)) ) even without…

Artificial Intelligence · Computer Science 2025-10-14 Yutong Yin , Zhaoran Wang

Standard Transformers have a fixed computational depth, fundamentally limiting their ability to generalize to tasks requiring variable-depth reasoning, such as multi-hop graph traversal or nested logic. We propose a depth-recurrent…

Machine Learning · Computer Science 2026-03-24 Hung-Hsuan Chen

At present, the mechanisms of in-context learning in Transformers are not well understood and remain mostly an intuition. In this paper, we suggest that training Transformers on auto-regressive objectives is closely related to…

Transformers are widely used in state-of-the-art machine translation, but the key to their success is still unknown. To gain insight into this, we consider three groups of parameters: embeddings, attention, and feed forward neural network…

Computation and Language · Computer Science 2021-09-22 Nikolay Bogoychev