Searcharxiv⌕ Search

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 19 recordsLinked to original sources
arXiv2019

Response evolution of mechanical metamaterials under architectural transformations

Architectural transformations play a key role in the evolution of complex systems, from design algorithms for metamaterials to flow and plasticity of disordered media. Here, we develop a general framework for the evolution of the linear mechanical response of network structures under discrete architectural transformations via sequential removal and addition of elastic elements. We focus on a class of spatially complex metamaterials, consisting of triangular building blocks. Rotations of these building blocks, corresponding to removing and adding elastic elements, introduce (topological) architectural defects. We show that the metamaterials' states of self stress play a crucial role, and that the mutually exclusive self stress states between two different network architectures span the difference in their mechanical response. For our class of metamaterials, we identify a localized representation of these states of self stress, which allows us to capture the evolving response. We use our insights to understand the unusual stress-steering behaviour of topological defects.

cond-mat.soft↗
arXiv2026

Hierarchical Fault Detection and Diagnosis for Transformer Architectures

Transformers now underpin critical AI systems across industry and research. Yet their faults can silently alter model behavior without runtime errors, and existing techniques offer little support for tracing these failures to their component and root cause. Such faults evade detection because loss and numerical values stay normal, and the visible symptom rarely identifies the component responsible. We present DEFault++, a hierarchical learning-based technique that first detects a fault, then identifies the affected component, and finally the cause within it, helping developers effectively debug transformer models. DEFault++ organizes component-level runtime measurements with a Fault Propagation Graph (FPG), a structural prior over the architecture's dependency paths, and reports the evidence behind each diagnosis. To train and evaluate it, we construct DEFault-bench, a benchmark of 5,556 labeled runs from mutation testing across seven models, nine tasks, and both encoder and decoder architectures. DEFault++ improves fault detection over four prior techniques, reaching an F1 of 0.826--0.909, and in a developer study with 21 participants, it raises repair accuracy from 57.1% to 83.3%. These results show that transformer fault diagnosis benefits from component-level measurements and architecture-aware reasoning rather than model-level behavior alone, and DEFault-bench provides a foundation for further research on transformer fault diagnosis.

cs.SE↗
arXiv2020

Differentiable Neural Architecture Transformation for Reproducible Architecture Improvement

Recently, Neural Architecture Search (NAS) methods are introduced and show impressive performance on many benchmarks. Among those NAS studies, Neural Architecture Transformer (NAT) aims to improve the given neural architecture to have better performance while maintaining computational costs. However, NAT has limitations about a lack of reproducibility. In this paper, we propose differentiable neural architecture transformation that is reproducible and efficient. The proposed method shows stable performance on various architectures. Extensive reproducibility experiments on two datasets, i.e., CIFAR-10 and Tiny Imagenet, present that the proposed method definitely outperforms NAT and be applicable to other models and datasets.

cs.LG↗
arXiv2026

Hybrid Dual-Path Linear Transformations for Efficient Transformer Architectures

Standard Transformer architectures rely heavily on dense linear transformations, treating feature projection as a monolithic, full-rank operation. We argue that this formulation is inefficient and lacks the structural inductive bias necessary for distinguishing between local feature preservation and global context integration. To address this, we introduce the Hybrid Dual-Path Linear (HDPL) operator, which decomposes the affine transformation into two topologically distinct pathways: a sparse block-diagonal component for high-rank local processing, and a low-rank Variational Autoencoder (VAE) bottleneck for global context regularization. By "surgically" replacing specific projections (Query, Key, Value, Gate, Up) with HDPL operators while retaining standard dense layers for aggregation (Output, Down), we achieve a superior balance of efficiency and representational power. Experiments on the FineWeb-Edu dataset demonstrate that the HDPL architecture outperforms a standard Llama-style baseline, reducing validation loss while simultaneously reducing parameter count by 6.8%. Beyond immediate performance gains, we discuss how the explicit materialization of a probabilistic latent space within the Transformer backbone serves as a vital architectural affordance, offering new pathways for inference-time or hypernetwork induced control, continual adaptation, interpretability, and cross-model or cross-modal synchronization. The code is available at https://github.com/VladimerKhasia/HDPL

cs.LG↗
arXiv2025

Dynamic Reasoning Chains through Depth-Specialized Mixture-of-Experts in Transformer Architectures

Contemporary transformer architectures apply identical processing depth to all inputs, creating inefficiencies and limiting reasoning quality. Simple factual queries are subjected to the same multilayered computation as complex logical problems, wasting resources while constraining deep inference. To overcome this, we came up with a concept of Dynamic Reasoning Chains through Depth Specialised Mixture of Experts (DS-MoE), a modular framework that extends the Mixture of Experts paradigm from width-based to depth specialised computation. DS-MoE introduces expert modules optimised for distinct reasoning depths, shallow pattern recognition, compositional reasoning, logical inference, memory integration, and meta-cognitive supervision. A learned routing network dynamically assembles custom reasoning chains, activating only the necessary experts to match input complexity. The dataset on which we trained and evaluated DS-MoE is on The Pile, an 800GB corpus covering diverse domains such as scientific papers, legal texts, programming code, and web content, enabling systematic assessment across reasoning depths. Experimental results demonstrate that DS-MoE achieves up to 16 per cent computational savings and 35 per cent faster inference compared to uniform-depth transformers, while delivering 2.8 per cent higher accuracy on complex multi-step reasoning benchmarks. Furthermore, routing decisions yield interpretable reasoning chains, enhancing transparency and scalability. These findings establish DS-MoE as a significant advancement in adaptive neural architectures, demonstrating that depth-specialised modular processing can simultaneously improve efficiency, reasoning quality, and interpretability in large-scale language models.

cs.CL↗
arXiv2021

Towards Accurate and Compact Architectures via Neural Architecture Transformer

Designing effective architectures is one of the key factors behind the success of deep neural networks. Existing deep architectures are either manually designed or automatically searched by some Neural Architecture Search (NAS) methods. However, even a well-designed/searched architecture may still contain many nonsignificant or redundant modules/operations. Thus, it is necessary to optimize the operations inside an architecture to improve the performance without introducing extra computational cost. To this end, we have proposed a Neural Architecture Transformer (NAT) method which casts the optimization problem into a Markov Decision Process (MDP) and seeks to replace the redundant operations with more efficient operations, such as skip or null connection. Note that NAT only considers a small number of possible transitions and thus comes with a limited search/transition space. As a result, such a small search space may hamper the performance of architecture optimization. To address this issue, we propose a Neural Architecture Transformer++ (NAT++) method which further enlarges the set of candidate transitions to improve the performance of architecture optimization. Specifically, we present a two-level transition rule to obtain valid transitions, i.e., allowing operations to have more efficient types (e.g., convolution->separable convolution) or smaller kernel sizes (e.g., 5x5->3x3). Note that different operations may have different valid transitions. We further propose a Binary-Masked Softmax (BMSoftmax) layer to omit the possible invalid transitions. Extensive experiments on several benchmark datasets show that the transformed architecture significantly outperforms both its original counterpart and the architectures optimized by existing methods.

cs.CV↗
arXiv2025

The Power of Architecture: Deep Dive into Transformer Architectures for Long-Term Time Series Forecasting

Transformer-based models have recently become dominant in Long-term Time Series Forecasting (LTSF), yet the variations in their architecture, such as encoder-only, encoder-decoder, and decoder-only designs, raise a crucial question: What Transformer architecture works best for LTSF tasks? However, existing models are often tightly coupled with various time-series-specific designs, making it difficult to isolate the impact of the architecture itself. To address this, we propose a novel taxonomy that disentangles these designs, enabling clearer and more unified comparisons of Transformer architectures. Our taxonomy considers key aspects such as attention mechanisms, forecasting aggregations, forecasting paradigms, and normalization layers. Through extensive experiments, we uncover several key insights: bi-directional attention with joint-attention is most effective; more complete forecasting aggregation improves performance; and the direct-mapping paradigm outperforms autoregressive approaches. Furthermore, our combined model, utilizing optimal architectural choices, consistently outperforms several existing models, reinforcing the validity of our conclusions. We hope these findings offer valuable guidance for future research on Transformer architectural designs in LTSF. Our code is available at https://github.com/HALF111/TSF_architecture.

cs.LG↗
arXiv2022

Spatio-Temporal Wind Speed Forecasting using Graph Networks and Novel Transformer Architectures

This study focuses on multi-step spatio-temporal wind speed forecasting for the Norwegian continental shelf. The study aims to leverage spatial dependencies through the relative physical location of different measurement stations to improve local wind forecasts. Our multi-step forecasting models produce either 10-minute, 1- or 4-hour forecasts, with 10-minute resolution, meaning that the models produce more informative time series for predicted future trends. A graph neural network (GNN) architecture was used to extract spatial dependencies, with different update functions to learn temporal correlations. These update functions were implemented using different neural network architectures. One such architecture, the Transformer, has become increasingly popular for sequence modelling in recent years. Various alterations have been proposed to better facilitate time series forecasting, of which this study focused on the Informer, LogSparse Transformer and Autoformer. This is the first time the LogSparse Transformer and Autoformer have been applied to wind forecasting and the first time any of these or the Informer have been formulated in a spatio-temporal setting for wind forecasting. By comparing against spatio-temporal Long Short-Term Memory (LSTM) and Multi-Layer Perceptron (MLP) models, the study showed that the models using the altered Transformer architectures as update functions in GNNs were able to outperform these. Furthermore, we propose the Fast Fourier Transformer (FFTransformer), which is a novel Transformer architecture based on signal decomposition and consists of two separate streams that analyse the trend and periodic components separately. The FFTransformer and Autoformer were found to achieve superior results for the 10-minute and 1-hour ahead forecasts, with the FFTransformer significantly outperforming all other models for the 4-hour ahead forecasts.

cs.LG↗
arXiv2020

NAT: Neural Architecture Transformer for Accurate and Compact Architectures

Designing effective architectures is one of the key factors behind the success of deep neural networks. Existing deep architectures are either manually designed or automatically searched by some Neural Architecture Search (NAS) methods. However, even a well-searched architecture may still contain many non-significant or redundant modules or operations (e.g., convolution or pooling), which may not only incur substantial memory consumption and computation cost but also deteriorate the performance. Thus, it is necessary to optimize the operations inside an architecture to improve the performance without introducing extra computation cost. Unfortunately, such a constrained optimization problem is NP-hard. To make the problem feasible, we cast the optimization problem into a Markov decision process (MDP) and seek to learn a Neural Architecture Transformer (NAT) to replace the redundant operations with the more computationally efficient ones (e.g., skip connection or directly removing the connection). Based on MDP, we learn NAT by exploiting reinforcement learning to obtain the optimization policies w.r.t. different architectures. To verify the effectiveness of the proposed strategies, we apply NAT on both hand-crafted architectures and NAS based architectures. Extensive experiments on two benchmark datasets, i.e., CIFAR-10 and ImageNet, demonstrate that the transformed architecture by NAT significantly outperforms both its original form and those architectures optimized by existing methods.

cs.LG↗
arXiv2024

ALBERTA: ALgorithm-Based Error Resilience in Transformer Architectures

Vision Transformers are being increasingly deployed in safety-critical applications that demand high reliability. It is crucial to ensure the correctness of their execution in spite of potential errors such as transient hardware errors. We propose a novel algorithm-based resilience framework called ALBERTA that allows us to perform end-to-end resilience analysis and protection of transformer-based architectures. First, our work develops an efficient process of computing and ranking the resilience of transformers layers. We find that due to the large size of transformer models, applying traditional network redundancy to a subset of the most vulnerable layers provides high error coverage albeit with impractically high overhead. We address this shortcoming by providing a software-directed, checksum-based error detection technique aimed at protecting the most vulnerable general matrix multiply (GEMM) layers in the transformer models that use either floating-point or integer arithmetic. Results show that our approach achieves over 99% coverage for errors that result in a mismatch with less than 0.2% and 0.01% computation and memory overheads, respectively. Lastly, we present the applicability of our framework in various modern GPU architectures under different numerical precisions. We introduce an efficient self-correction mechanism for resolving erroneous detection with an average of less than 2% overhead per error.

cs.CR↗
arXiv2025

Efficient Language Modeling for Low-Resource Settings with Hybrid RNN-Transformer Architectures

Transformer-based language models have recently been at the forefront of active research in text generation. However, these models' advances come at the price of prohibitive training costs, with parameter counts in the billions and compute requirements measured in petaflop/s-decades. In this paper, we investigate transformer-based architectures for improving model performance in a low-data regime by selectively replacing attention layers with feed-forward and quasi-recurrent neural network layers. We test these architectures on the standard Enwik8 and Wikitext-103 corpora. Our results show that our reduced architectures outperform existing models with a comparable number of parameters, and obtain comparable performance to larger models while significantly reducing the number of parameters.

cs.CL↗
arXiv2026

In Transformer We Trust? A Perspective on Transformer Architecture Failure Modes

Transformer architectures have revolutionized machine learning across a wide range of domains, from natural language processing to scientific computing. However, their growing deployment in high-stakes applications, such as computer vision, natural language processing, healthcare, autonomous systems, and critical areas of scientific computing including climate modeling, materials discovery, drug discovery, nuclear science, and robotics, necessitates a deeper and more rigorous understanding of their trustworthiness. In this work, we critically examine the foundational question: \textitHow trustworthy are transformer models?} We evaluate their reliability through a comprehensive review of interpretability, explainability, robustness against adversarial attacks, fairness, and privacy. We systematically examine the trustworthiness of transformer-based models in safety-critical applications spanning natural language processing, computer vision, and science and engineering domains, including robotics, medicine, earth sciences, materials science, fluid dynamics, nuclear science, and automated theorem proving; highlighting high-impact areas where these architectures are central and analyzing the risks associated with their deployment. By synthesizing insights across these diverse areas, we identify recurring structural vulnerabilities, domain-specific risks, and open research challenges that limit the reliable deployment of transformers.

cs.LG↗
arXiv2020

On Layer Normalization in the Transformer Architecture

The Transformer is widely used in natural language processing tasks. To train a Transformer however, one usually needs a carefully designed learning rate warm-up stage, which is shown to be crucial to the final performance but will slow down the optimization and bring more hyper-parameter tunings. In this paper, we first study theoretically why the learning rate warm-up stage is essential and show that the location of layer normalization matters. Specifically, we prove with mean field theory that at initialization, for the original-designed Post-LN Transformer, which places the layer normalization between the residual blocks, the expected gradients of the parameters near the output layer are large. Therefore, using a large learning rate on those gradients makes the training unstable. The warm-up stage is practically helpful for avoiding this problem. On the other hand, our theory also shows that if the layer normalization is put inside the residual blocks (recently proposed as Pre-LN Transformer), the gradients are well-behaved at initialization. This motivates us to remove the warm-up stage for the training of Pre-LN Transformers. We show in our experiments that Pre-LN Transformers without the warm-up stage can reach comparable results with baselines while requiring significantly less training time and hyper-parameter tuning on a wide range of applications.

cs.LG↗
arXiv2025

Disentangling and Integrating Relational and Sensory Information in Transformer Architectures

Relational reasoning is a central component of generally intelligent systems, enabling robust and data-efficient inductive generalization. Recent empirical evidence shows that many existing neural architectures, including Transformers, struggle with tasks requiring relational reasoning. In this work, we distinguish between two types of information: sensory information about the properties of individual objects, and relational information about the relationships between objects. While neural attention provides a powerful mechanism for controlling the flow of sensory information between objects, the Transformer lacks an explicit computational mechanism for routing and processing relational information. To address this limitation, we propose an architectural extension of the Transformer framework that we call the Dual Attention Transformer (DAT), featuring two distinct attention mechanisms: sensory attention for directing the flow of sensory information, and a novel relational attention mechanism for directing the flow of relational information. We empirically evaluate DAT on a diverse set of tasks ranging from synthetic relational benchmarks to complex real-world tasks such as language modeling and visual processing. Our results demonstrate that integrating explicit relational computational mechanisms into the Transformer architecture leads to significant performance gains in terms of data efficiency and parameter efficiency.

cs.LG↗
arXiv2021

UTNet: A Hybrid Transformer Architecture for Medical Image Segmentation

Transformer architecture has emerged to be successful in a number of natural language processing tasks. However, its applications to medical vision remain largely unexplored. In this study, we present UTNet, a simple yet powerful hybrid Transformer architecture that integrates self-attention into a convolutional neural network for enhancing medical image segmentation. UTNet applies self-attention modules in both encoder and decoder for capturing long-range dependency at different scales with minimal overhead. To this end, we propose an efficient self-attention mechanism along with relative position encoding that reduces the complexity of self-attention operation significantly from $O(n^2)$ to approximate $O(n)$. A new self-attention decoder is also proposed to recover fine-grained details from the skipped connections in the encoder. Our approach addresses the dilemma that Transformer requires huge amounts of data to learn vision inductive bias. Our hybrid layer design allows the initialization of Transformer into convolutional networks without a need of pre-training. We have evaluated UTNet on the multi-label, multi-vendor cardiac magnetic resonance imaging cohort. UTNet demonstrates superior segmentation performance and robustness against the state-of-the-art approaches, holding the promise to generalize well on other medical image segmentations.

cs.CV↗
arXiv2024

Circuit Complexity Bounds for RoPE-based Transformer Architecture

Characterizing the express power of the Transformer architecture is critical to understanding its capacity limits and scaling law. Recent works provide the circuit complexity bounds to Transformer-like architecture. On the other hand, Rotary Position Embedding ($\mathsf{RoPE}$) has emerged as a crucial technique in modern large language models, offering superior performance in capturing positional information compared to traditional position embeddings, which shows great potential in application prospects, particularly for the long context scenario. Empirical evidence also suggests that $\mathsf{RoPE}$-based Transformer architectures demonstrate greater generalization capabilities compared to conventional Transformer models. In this work, we establish a circuit complexity bound for Transformers with $\mathsf{RoPE}$ attention. Our key contribution is that we show that unless $\mathsf{TC}^0 = \mathsf{NC}^1$, a $\mathsf{RoPE}$-based Transformer with $\mathrm{poly}(n)$-precision, $O(1)$ layers, hidden dimension $d \leq O(n)$ cannot solve the Arithmetic formula evaluation problem or the Boolean formula value problem. This result significantly demonstrates the fundamental limitation of the expressivity of the $\mathsf{RoPE}$-based Transformer architecture, although it achieves giant empirical success. Our theoretical result not only establishes the complexity bound but also may instruct further work on the $\mathsf{RoPE}$-based Transformer.

cs.LG↗
arXiv2021

EEG-Transformer: Self-attention from Transformer Architecture for Decoding EEG of Imagined Speech

Transformers are groundbreaking architectures that have changed a flow of deep learning, and many high-performance models are developing based on transformer architectures. Transformers implemented only with attention with encoder-decoder structure following seq2seq without using RNN, but had better performance than RNN. Herein, we investigate the decoding technique for electroencephalography (EEG) composed of self-attention module from transformer architecture during imagined speech and overt speech. We performed classification of nine subjects using convolutional neural network based on EEGNet that captures temporal-spectral-spatial features from EEG of imagined speech and overt speech. Furthermore, we applied the self-attention module to decoding EEG to improve the performance and lower the number of parameters. Our results demonstrate the possibility of decoding brain activities of imagined speech and overt speech using attention modules. Also, only single channel EEG or ear-EEG can be used to decode the imagined speech for practical BCIs.

cs.HC↗
arXiv2022

Two-Stream Transformer Architecture for Long Video Understanding

Pure vision transformer architectures are highly effective for short video classification and action recognition tasks. However, due to the quadratic complexity of self attention and lack of inductive bias, transformers are resource intensive and suffer from data inefficiencies. Long form video understanding tasks amplify data and memory efficiency problems in transformers making current approaches unfeasible to implement on data or memory restricted domains. This paper introduces an efficient Spatio-Temporal Attention Network (STAN) which uses a two-stream transformer architecture to model dependencies between static image features and temporal contextual features. Our proposed approach can classify videos up to two minutes in length on a single GPU, is data efficient, and achieves SOTA performance on several long video understanding tasks.

cs.CV↗