SEARCH · Searcharxiv

Results for “transformer architecture”

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2Linked to original sources

arXiv2025

Physical models realizing the transformer architecture of large language models

The introduction of the transformer architecture in 2017 marked the most striking advancement in natural language processing. The transformer is a model architecture relying entirely on an attention mechanism to draw global dependencies between input and output. However, we believe there is a gap in our theoretical understanding of what the transformer is, and how it works physically. From a physical perspective on modern chips, such as those chips under 28nm, modern intelligent machines should be regarded as open quantum systems beyond conventional statistical systems. Thereby, in this paper, we construct physical models realizing large language models based on a transformer architecture as open quantum systems in the Fock space over the Hilbert space of tokens. Our physical models underlie the transformer architecture for large language models.

cs.LG↗

arXiv2025

Approximation Rate of the Transformer Architecture for Sequence Modeling

The Transformer architecture is widely applied in sequence modeling applications, yet the theoretical understanding of its working principles remains limited. In this work, we investigate the approximation rate for single-layer Transformers with one head. We consider a class of non-linear relationships and identify a novel notion of complexity measures to establish an explicit Jackson-type approximation rate estimate for the Transformer. This rate reveals the structural properties of the Transformer and suggests the types of sequential relationships it is best suited for approximating. In particular, the results on approximation rates enable us to concretely analyze the differences between the Transformer and classical sequence modeling methods, such as recurrent neural networks.

cs.LG↗

arXiv2025

OT-Transformer: A Continuous-time Transformer Architecture with Optimal Transport Regularization

Transformers have achieved state-of-the-art performance in numerous tasks. In this paper, we propose a continuous-time formulation of transformers. Specifically, we consider a dynamical system whose governing equation is parametrized by transformer blocks. We leverage optimal transport theory to regularize the training problem, which enhances stability in training and improves generalization of the resulting model. Moreover, we demonstrate in theory that this regularization is necessary as it promotes uniqueness and regularity of solutions. Our model is flexible in that almost any existing transformer architectures can be adopted to construct the dynamical system with only slight modifications to the existing code. We perform extensive numerical experiments on tasks motivated by natural language processing, image classification, and point cloud classification. Our experimental results show that the proposed method improves the performance of its discrete counterpart and outperforms relevant comparing models.

cs.LG↗

arXiv2026

TAS-LoRA: Transformer Architecture Search with Mixture-of-LoRA Experts

Transformer architecture search (TAS) discovers optimal vision transformer (ViT) architectures automatically, reducing human effort to manually design ViTs. However, existing TAS methods suffer from the feature collapse problem, where subnets within a supernet fail to learn subnet-specific features, mainly due to the shared weights in a supernet, limiting the performance of individual subnets. To address this, we propose TAS-LoRA, a novel method that introduces parameter-efficient low-rank adaptation (LoRA) to enable subnet-specific feature learning, while maintaining computational efficiency. TAS-LoRA incorporates a Mixture-of-LoRAExperts (MoLE) strategy, where a lightweight router dynamically assigns LoRA experts based on subnet architectures, and introduces a group-wise router initialization technique to encourage diverse feature learning across experts early in training. Extensive experiments on ImageNet and several transfer learning benchmarks, including CIFAR-10/100, Flowers, CARS, and INAT-19, demonstrate that TAS-LoRA mitigates feature collapse effectively, improving performance over state-of-the-art TAS methods significantly.

cs.CV↗

arXiv2019

Efficient Neural Architecture Transformation Searchin Channel-Level for Object Detection

Recently, Neural Architecture Search has achieved great success in large-scale image classification. In contrast, there have been limited works focusing on architecture search for object detection, mainly because the costly ImageNet pre-training is always required for detectors. Training from scratch, as a substitute, demands more epochs to converge and brings no computation saving. To overcome this obstacle, we introduce a practical neural architecture transformation search(NATS)algorithm for object detection in this paper. Instead of searching and constructing an entire network, NATS explores the architecture space on the base of existing network and reusing its weights. We propose a novel neural architecture search strategy in channel-level instead of path-level and devise a search space specially targeting at object detection. With the combination of these two designs, an architecture transformation scheme could be discovered to adapt a network designed for image classification to task of object detection. Since our method is gradient-based and only searches for a transformation scheme, the weights of models pretrained inImageNet could be utilized in both searching and retraining stage, which makes the whole process very efficient. The transformed network requires no extra parameters and FLOPs, and is friendly to hardware optimization, which is practical to use in real-time application. In experiments, we demonstrate the effectiveness of NATSon networks like ResNet and ResNeXt. Our transformed networks, combined with various detection frameworks, achieve significant improvements on the COCO dataset while keeping fast.

cs.CV↗

arXiv2026

MPDiT: Multi-Patch Global-to-Local Transformer Architecture For Efficient Flow Matching and Diffusion Model

Transformer architectures, particularly Diffusion Transformers (DiTs), have become widely used in diffusion and flow-matching models due to their strong performance compared to convolutional UNets. However, the isotropic design of DiTs processes the same number of patchified tokens in every block, leading to relatively heavy computation during training process. In this work, we introduce a multi-patch transformer design in which early blocks operate on larger patches to capture coarse global context, while later blocks use smaller patches to refine local details. This hierarchical design could reduces computational cost by up to 50% in GFLOPs while achieving good generative performance. In addition, we also propose improved designs for time and class embeddings that accelerate training convergence. Extensive experiments on the ImageNet dataset demonstrate the effectiveness of our architectural choices. Code is released at: https://github.com/quandao10/MPDiT

cs.CV↗

arXiv2022

Training-free Transformer Architecture Search

Recently, Vision Transformer (ViT) has achieved remarkable success in several computer vision tasks. The progresses are highly relevant to the architecture design, then it is worthwhile to propose Transformer Architecture Search (TAS) to search for better ViTs automatically. However, current TAS methods are time-consuming and existing zero-cost proxies in CNN do not generalize well to the ViT search space according to our experimental observations. In this paper, for the first time, we investigate how to conduct TAS in a training-free manner and devise an effective training-free TAS (TF-TAS) scheme. Firstly, we observe that the properties of multi-head self-attention (MSA) and multi-layer perceptron (MLP) in ViTs are quite different and that the synaptic diversity of MSA affects the performance notably. Secondly, based on the observation, we devise a modular strategy in TF-TAS that evaluates and ranks ViT architectures from two theoretical perspectives: synaptic diversity and synaptic saliency, termed as DSS-indicator. With DSS-indicator, evaluation results are strongly correlated with the test accuracies of ViT models. Experimental results demonstrate that our TF-TAS achieves a competitive performance against the state-of-the-art manually or automatically design ViT architectures, and it promotes the searching efficiency in ViT search space greatly: from about $24$ GPU days to less than $0.5$ GPU days. Moreover, the proposed DSS-indicator outperforms the existing cutting-edge zero-cost approaches (e.g., TE-score and NASWOT).

cs.CV↗

arXiv2025

Learning Novel Transformer Architecture for Time-series Forecasting

Despite the success of Transformer-based models in the time-series prediction (TSP) tasks, the existing Transformer architecture still face limitations and the literature lacks comprehensive explorations into alternative architectures. To address these challenges, we propose AutoFormer-TS, a novel framework that leverages a comprehensive search space for Transformer architectures tailored to TSP tasks. Our framework introduces a differentiable neural architecture search (DNAS) method, AB-DARTS, which improves upon existing DNAS approaches by enhancing the identification of optimal operations within the architecture. AutoFormer-TS systematically explores alternative attention mechanisms, activation functions, and encoding operations, moving beyond the traditional Transformer design. Extensive experiments demonstrate that AutoFormer-TS consistently outperforms state-of-the-art baselines across various TSP benchmarks, achieving superior forecasting accuracy while maintaining reasonable training efficiency.

cs.LG↗

arXiv2024

HyTAS: A Hyperspectral Image Transformer Architecture Search Benchmark and Analysis

Hyperspectral Imaging (HSI) plays an increasingly critical role in precise vision tasks within remote sensing, capturing a wide spectrum of visual data. Transformer architectures have significantly enhanced HSI task performance, while advancements in Transformer Architecture Search (TAS) have improved model discovery. To harness these advancements for HSI classification, we make the following contributions: i) We propose HyTAS, the first benchmark on transformer architecture search for Hyperspectral imaging, ii) We comprehensively evaluate 12 different methods to identify the optimal transformer over 5 different datasets, iii) We perform an extensive factor analysis on the Hyperspectral transformer search performance, greatly motivating future research in this direction. All benchmark materials are available at HyTAS.

cs.CV↗

arXiv2021

Perspectives and Prospects on Transformer Architecture for Cross-Modal Tasks with Language and Vision

Transformer architectures have brought about fundamental changes to computational linguistic field, which had been dominated by recurrent neural networks for many years. Its success also implies drastic changes in cross-modal tasks with language and vision, and many researchers have already tackled the issue. In this paper, we review some of the most critical milestones in the field, as well as overall trends on how transformer architecture has been incorporated into visuolinguistic cross-modal tasks. Furthermore, we discuss its current limitations and speculate upon some of the prospects that we find imminent.

cs.CV↗

arXiv2024

Reducing the Transformer Architecture to a Minimum

Transformers are a widespread and successful model architecture, particularly in Natural Language Processing (NLP) and Computer Vision (CV). The essential innovation of this architecture is the Attention Mechanism, which solves the problem of extracting relevant context information from long sequences in NLP and realistic scenes in CV. A classical neural network component, a Multi-Layer Perceptron (MLP), complements the attention mechanism. Its necessity is frequently justified by its capability of modeling nonlinear relationships. However, the attention mechanism itself is nonlinear through its internal use of similarity measures. A possible hypothesis is that this nonlinearity is sufficient for modeling typical application problems. As the MLPs usually contain the most trainable parameters of the whole model, their omission would substantially reduce the parameter set size. Further components can also be reorganized to reduce the number of parameters. Under some conditions, query and key matrices can be collapsed into a single matrix of the same size. The same is true about value and projection matrices, which can also be omitted without eliminating the substance of the attention mechanism. Initially, the similarity measure was defined asymmetrically, with peculiar properties such as that a token is possibly dissimilar to itself. A possible symmetric definition requires only half of the parameters. We have laid the groundwork by testing widespread CV benchmarks: MNIST and CIFAR-10. The tests have shown that simplified transformer architectures (a) without MLP, (b) with collapsed matrices, and (c) symmetric similarity matrices exhibit similar performance as the original architecture, saving up to 90% of parameters without hurting the classification performance.

cs.LG↗

arXiv2022

A K-variate Time Series Is Worth K Words: Evolution of the Vanilla Transformer Architecture for Long-term Multivariate Time Series Forecasting

Multivariate time series forecasting (MTSF) is a fundamental problem in numerous real-world applications. Recently, Transformer has become the de facto solution for MTSF, especially for the long-term cases. However, except for the one forward operation, the basic configurations in existing MTSF Transformer architectures were barely carefully verified. In this study, we point out that the current tokenization strategy in MTSF Transformer architectures ignores the token uniformity inductive bias of Transformers. Therefore, the vanilla MTSF transformer struggles to capture details in time series and presents inferior performance. Based on this observation, we make a series of evolution on the basic architecture of the vanilla MTSF transformer. We vary the flawed tokenization strategy, along with the decoder structure and embeddings. Surprisingly, the evolved simple transformer architecture is highly effective, which successfully avoids the over-smoothing phenomena in the vanilla MTSF transformer, achieves a more detailed and accurate prediction, and even substantially outperforms the state-of-the-art Transformers that are well-designed for MTSF.

cs.LG↗

arXiv2024

PointMT: Efficient Point Cloud Analysis with Hybrid MLP-Transformer Architecture

In recent years, point cloud analysis methods based on the Transformer architecture have made significant progress, particularly in the context of multimedia applications such as 3D modeling, virtual reality, and autonomous systems. However, the high computational resource demands of the Transformer architecture hinder its scalability, real-time processing capabilities, and deployment on mobile devices and other platforms with limited computational resources. This limitation remains a significant obstacle to its practical application in scenarios requiring on-device intelligence and multimedia processing. To address this challenge, we propose an efficient point cloud analysis architecture, \textbf{Point} \textbf{M}LP-\textbf{T}ransformer (PointMT). This study tackles the quadratic complexity of the self-attention mechanism by introducing a linear complexity local attention mechanism for effective feature aggregation. Additionally, to counter the Transformer's focus on token differences while neglecting channel differences, we introduce a parameter-free channel temperature adaptation mechanism that adaptively adjusts the attention weight distribution in each channel, enhancing the precision of feature aggregation. To improve the Transformer's slow convergence speed due to the limited scale of point cloud datasets, we propose an MLP-Transformer hybrid module, which significantly enhances the model's convergence speed. Furthermore, to boost the feature representation capability of point tokens, we refine the classification head, enabling point tokens to directly participate in prediction. Experimental results on multiple evaluation benchmarks demonstrate that PointMT achieves performance comparable to state-of-the-art methods while maintaining an optimal balance between performance and accuracy.

cs.CV↗

arXiv2026

Multi-Headed Transformer Architectures as Time-dependent Wasserstein Gradient Flows

In recent years, transformer architectures have revolutionized the field of language processing, opening the door to previously unforeseen possibilities. However, from a theoretical point of view, the mathematical models proposed in the literature often lack direct contact with the actual architectures and depend on strong simplifying assumptions. In this paper, we reduce this gap by modelling the data flow in multi-headed transformer architectures as time-dependent gradient flows for a suitable interaction energy capturing the design of the attention mechanism. The explicit dependence on time allows us to consider different weights for each head and for each layer, without imposing constraints on the initialization method. Moreover, we prove that, under a suitable integrability assumption on the evolution of the weights, each element of the $ω$-limit set of the gradient flows is a stationary point of the interaction energy at a limiting weight distribution. Finally, we analyse the stability of the gradient flows considering perturbations of both the initial data and the weights. Specifically, on the one hand, we study the robustness of the proposed models with respect to noisy inputs, establishing a continuous dependence of the gradient flows on the initial data and uniqueness of the flows. On the other hand, we prove the $Γ$-convergence of the perturbed interaction energy to the unperturbed one, leading to the convergence of the corresponding gradient flows. We complement these theoretical results with numerical experiments that confirm the predicted energy-dissipation identity and clarify the asymptotic behavior of the dynamics in both the autonomous-like (Ornstein--Uhlenbeck) and the genuinely non-autonomous (oscillating-weights) regimes.

cs.LG↗

arXiv2022

Deep Learning Transformer Architecture for Named Entity Recognition on Low Resourced Languages: State of the art results

This paper reports on the evaluation of Deep Learning (DL) transformer architecture models for Named-Entity Recognition (NER) on ten low-resourced South African (SA) languages. In addition, these DL transformer models were compared to other Neural Network and Machine Learning (ML) NER models. The findings show that transformer models substantially improve performance when applying discrete fine-tuning parameters per language. Furthermore, fine-tuned transformer models outperform other neural network and machine learning models on NER with the low-resourced SA languages. For example, the transformer models obtained the highest F-scores for six of the ten SA languages and the highest average F-score surpassing the Conditional Random Fields ML model. Practical implications include developing high-performance NER capability with less effort and resource costs, potentially improving downstream NLP tasks such as Machine Translation (MT). Therefore, the application of DL transformer architecture models for NLP NER sequence tagging tasks on low-resourced SA languages is viable. Additional research could evaluate the more recent transformer architecture models on other Natural Language Processing tasks and applications, such as Phrase chunking, MT, and Part-of-Speech tagging.

cs.CL↗

arXiv2026

Automatic weld seam segmentation for industrial quality control: a comparison of RGB and polarimetric imaging with CNN and transformer architectures

Visual inspection of welded assemblies remains one of the least automated stages in many industrial production processes, still depending largely on the experience of human operators and thus subject to inter-operator variability; the manufacturing of special-purpose machinery cabins, the setting of this study, is one representative case. This work evaluates the feasibility of automatic weld seam segmentation from RGB and polarimetric imagery, comparing controlled laboratory acquisitions with images captured under real, uncontrolled conditions. Convolutional neural network (CNN) architectures and transformer-based architectures are benchmarked under a unified, threshold-independent protocol, training each CNN with three random seeds to separate genuine effects from seed noise. In controlled RGB conditions, CNN models reach a mean mask mAP50 of up to 0.87, but drop to 0.22-0.48 under uncontrolled acquisition, showing that the acquisition setup is a first-order component of the inspection system. Polarimetric imaging with alignment-preserving geometric augmentation localizes previously unseen welds with a mean mask mAP50 up to 0.93: on par with, rather than ahead of, the best controlled-RGB result, but reaching that accuracy on uncontrolled RGB without requiring acquisition control. The clearest architectural finding concerns viewpoint robustness. In-distribution, transformers and CNNs are broadly comparable; but under a test-time viewpoint shift, the transformer models, and RF-DETR in particular, retain high accuracy while every CNN collapses. The gap holds across three seeds and a resolution-matched control, pointing to architecture rather than training resolution. Within the CNN family, capacity brings no reliable in-distribution gain once seed variance is accounted for: small CNNs suffice for fixed viewpoints, transformers for variable ones.

cs.CV↗

arXiv2021

ViTAS: Vision Transformer Architecture Search

Vision transformers (ViTs) inherited the success of NLP but their structures have not been sufficiently investigated and optimized for visual tasks. One of the simplest solutions is to directly search the optimal one via the widely used neural architecture search (NAS) in CNNs. However, we empirically find this straightforward adaptation would encounter catastrophic failures and be frustratingly unstable for the training of superformer. In this paper, we argue that since ViTs mainly operate on token embeddings with little inductive bias, imbalance of channels for different architectures would worsen the weight-sharing assumption and cause the training instability as a result. Therefore, we develop a new cyclic weight-sharing mechanism for token embeddings of the ViTs, which enables each channel could more evenly contribute to all candidate architectures. Besides, we also propose identity shifting to alleviate the many-to-one issue in superformer and leverage weak augmentation and regularization techniques for more steady training empirically. Based on these, our proposed method, ViTAS, has achieved significant superiority in both DeiT- and Twins-based ViTs. For example, with only $1.4$G FLOPs budget, our searched architecture has $3.3\%$ ImageNet-$1$k accuracy than the baseline DeiT. With $3.0$G FLOPs, our results achieve $82.0\%$ accuracy on ImageNet-$1$k, and $45.9\%$ mAP on COCO$2017$ which is $2.4\%$ superior than other ViTs.

cs.CV↗

arXiv2024

Transformer Architecture for NetsDB

Transformers models have become the backbone of the current state-of-the-art models in language, vision, and multimodal domains. These models, at their core, utilize multi-head self-attention to selectively aggregate context, generating dynamic contextual embeddings and modeling long-range dependencies for a clear contextual understanding. Lixi et al. \cite{zhou2022serving} proposed a method to use relational databases for deploying large-scale deep learning models and created an open-source implementation called NetsDB for the same. We build upon the previous work of these authors by creating an end-to-end implementation of the Encoder part of the transformer for model serving in NetsDB. Specifically, we construct a two-block encoder that includes Multi-Head Attention and its accompanying self-attention mechanism, Layer-Norm, Dropout, FeedForward Layers, and the necessary residual connections. We load out weights from our model for distributed processing, deployment, and efficient inferencing. To prove the efficacy of our implementation, we conduct a comprehensive performance analysis by comparing it with existing implementations in PyTorch, Tensorflow, Flax, and MxNet across key metrics such as inference time and model size.

cs.CV↗