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 667 records · Page 37Linked to original sources

arXiv2026

SpikeDecoder: Realizing the GPT Architecture with Spiking Neural Networks

The Transformer architecture is widely regarded as the most powerful tool for natural language processing, but due to a high number of complex operations, it inherently faces the issue of high energy consumption. To address this issue, we consider Spiking Neural Networks (SNNs), which are an energy-efficient alternative to conventional Artificial Neural Networks (ANNs) due to their naturally event-driven approach to processing information. However, this inherently makes them difficult to train. Often, many SNN-based models circumvent this issue by converting pre-trained ANNs. More recently, attempts have been made to design directly trainable SNN-based adaptations of the Transformer model structure. Although the results showed great promise, the application field was computer vision. Moreover, the proposed model incorporates only encoder blocks. In this paper, we propose SpikeDecoder, a fully SNN-based implementation of the Transformer decoder block, for applications in natural language processing. In a series of experiments, we analyze the impact of exchanging different blocks of the ANN model with spike-based alternatives to identify trade-offs and significant sources of performance loss. We further investigate the role of residual connections and the selection of SNN-compatible normalization techniques. Besides the work on the model architecture, we formulate and compare different embedding methods to project text data into spikes. Finally, we demonstrate that our proposed SNN-based decoder block reduces the theoretical energy consumption by 87% to 93% compared to the ANN baseline.

cs.NE↗

arXiv2025

Gated Associative Memory: A Parallel O(N) Architecture for Efficient Sequence Modeling

The Transformer architecture, underpinned by the self-attention mechanism, has become the de facto standard for sequence modeling tasks. However, its core computational primitive scales quadratically with sequence length (O(N^2)), creating a significant bottleneck for processing long contexts. In this paper, we propose the Gated Associative Memory (GAM) network, a novel, fully parallel architecture for sequence modeling that exhibits linear complexity (O(N)) with respect to sequence length. The GAM block replaces the self-attention layer with two parallel pathways: a causal convolution to efficiently capture local, position-dependent context, and a parallel associative memory retrieval mechanism to model global, content-based patterns. These pathways are dynamically fused using a gating mechanism, allowing the model to flexibly combine local and global information for each token. We implement GAM from scratch and conduct a rigorous comparative analysis against a standard Transformer model and a modern linear-time baseline (Mamba) on the WikiText-2 benchmark, as well as against the Transformer on the TinyStories dataset. Our experiments demonstrate that GAM is consistently faster, outperforming both baselines on training speed, and achieves a superior or competitive final validation perplexity across all datasets, establishing it as a promising and efficient alternative for sequence modeling.

cs.CL↗

arXiv2024

GPT-2 Through the Lens of Vector Symbolic Architectures

Understanding the general priniciples behind transformer models remains a complex endeavor. Experiments with probing and disentangling features using sparse autoencoders (SAE) suggest that these models might manage linear features embedded as directions in the residual stream. This paper explores the resemblance between decoder-only transformer architecture and vector symbolic architectures (VSA) and presents experiments indicating that GPT-2 uses mechanisms involving nearly orthogonal vector bundling and binding operations similar to VSA for computation and communication between layers. It further shows that these principles help explain a significant portion of the actual neural weights.

cs.LG↗

arXiv2024

Body Transformer: Leveraging Robot Embodiment for Policy Learning

In recent years, the transformer architecture has become the de facto standard for machine learning algorithms applied to natural language processing and computer vision. Despite notable evidence of successful deployment of this architecture in the context of robot learning, we claim that vanilla transformers do not fully exploit the structure of the robot learning problem. Therefore, we propose Body Transformer (BoT), an architecture that leverages the robot embodiment by providing an inductive bias that guides the learning process. We represent the robot body as a graph of sensors and actuators, and rely on masked attention to pool information throughout the architecture. The resulting architecture outperforms the vanilla transformer, as well as the classical multilayer perceptron, in terms of task completion, scaling properties, and computational efficiency when representing either imitation or reinforcement learning policies. Additional material including the open-source code is available at https://sferrazza.cc/bot_site.

cs.RO↗

arXiv2024

Transformer-based Neuro-Animator for Qualitative Simulation of Soft Body Movement

The human mind effortlessly simulates the movements of objects governed by the laws of physics, such as a fluttering, or a waving flag under wind force, without understanding the underlying physics. This suggests that human cognition can predict the unfolding of physical events using an intuitive prediction process. This process might result from memory recall, yielding a qualitatively believable mental image, though it may not be exactly according to real-world physics. Drawing inspiration from the intriguing human ability to qualitatively visualize and describe dynamic events from past experiences without explicitly engaging in mathematical computations, this paper investigates the application of recent transformer architectures as a neuro-animator model. The visual transformer model is trained to predict flag motions at the \emph{t+1} time step, given information of previous motions from \emph{t-n} $\cdots$ \emph{t} time steps. The results show that the visual transformer-based architecture successfully learns temporal embedding of flag motions and produces reasonable quality simulations of flag waving under different wind forces.

cs.CV↗

arXiv2022

LiteTransformerSearch: Training-free Neural Architecture Search for Efficient Language Models

The Transformer architecture is ubiquitously used as the building block of large-scale autoregressive language models. However, finding architectures with the optimal trade-off between task performance (perplexity) and hardware constraints like peak memory utilization and latency is non-trivial. This is exacerbated by the proliferation of various hardware. We leverage the somewhat surprising empirical observation that the number of decoder parameters in autoregressive Transformers has a high rank correlation with task performance, irrespective of the architecture topology. This observation organically induces a simple Neural Architecture Search (NAS) algorithm that uses decoder parameters as a proxy for perplexity without need for any model training. The search phase of our training-free algorithm, dubbed Lightweight Transformer Search (LTS), can be run directly on target devices since it does not require GPUs. Using on-target-device measurements, LTS extracts the Pareto-frontier of perplexity versus any hardware performance cost. We evaluate LTS on diverse devices from ARM CPUs to NVIDIA GPUs and two popular autoregressive Transformer backbones: GPT-2 and Transformer-XL. Results show that the perplexity of 16-layer GPT-2 and Transformer-XL can be achieved with up to 1.5x, 2.5x faster runtime and 1.2x, 2.0x lower peak memory utilization. When evaluated in zero and one-shot settings, LTS Pareto-frontier models achieve higher average accuracy compared to the 350M parameter OPT across 14 tasks, with up to 1.6x lower latency. LTS extracts the Pareto-frontier in under 3 hours while running on a commodity laptop. We effectively remove the carbon footprint of hundreds of GPU hours of training during search, offering a strong simple baseline for future NAS methods in autoregressive language modeling.

cs.LG↗

arXiv2025

Decoupling Knowledge and Reasoning in Transformers: A Modular Architecture with Generalized Cross-Attention

Transformers have achieved remarkable success across diverse domains, but their monolithic architecture presents challenges in interpretability, adaptability, and scalability. This paper introduces a novel modular Transformer architecture that explicitly decouples knowledge and reasoning through a generalized cross-attention mechanism to a globally shared knowledge base with layer-specific transformations, specifically designed for effective knowledge retrieval. Critically, we provide a rigorous mathematical derivation demonstrating that the Feed-Forward Network (FFN) in a standard Transformer is a specialized case (a closure) of this generalized cross-attention, revealing its role in implicit knowledge retrieval and validating our design. This theoretical framework provides a new lens for understanding FFNs and lays the foundation for future research exploring enhanced interpretability, adaptability, and scalability, enabling richer interplay with external knowledge bases and other systems.

cs.LG↗

arXiv2024

Token Statistics Transformer: Linear-Time Attention via Variational Rate Reduction

The attention operator is arguably the key distinguishing factor of transformer architectures, which have demonstrated state-of-the-art performance on a variety of tasks. However, transformer attention operators often impose a significant computational burden, with the computational complexity scaling quadratically with the number of tokens. In this work, we propose a novel transformer attention operator whose computational complexity scales linearly with the number of tokens. We derive our network architecture by extending prior work which has shown that a transformer style architecture naturally arises by "white-box" architecture design, where each layer of the network is designed to implement an incremental optimization step of a maximal coding rate reduction objective (MCR$^2$). Specifically, we derive a novel variational form of the MCR$^2$ objective and show that the architecture that results from unrolled gradient descent of this variational objective leads to a new attention module called Token Statistics Self-Attention (TSSA). TSSA has linear computational and memory complexity and radically departs from the typical attention architecture that computes pairwise similarities between tokens. Experiments on vision, language, and long sequence tasks show that simply swapping TSSA for standard self-attention, which we refer to as the Token Statistics Transformer (ToST), achieves competitive performance with conventional transformers while being significantly more computationally efficient and interpretable. Our results also somewhat call into question the conventional wisdom that pairwise similarity style attention mechanisms are critical to the success of transformer architectures. Code will be available at https://github.com/RobinWu218/ToST.

cs.LG↗

arXiv2024

The Need for Speed: Pruning Transformers with One Recipe

We introduce the $\textbf{O}$ne-shot $\textbf{P}$runing $\textbf{T}$echnique for $\textbf{I}$nterchangeable $\textbf{N}$etworks ($\textbf{OPTIN}$) framework as a tool to increase the efficiency of pre-trained transformer architectures $\textit{without requiring re-training}$. Recent works have explored improving transformer efficiency, however often incur computationally expensive re-training procedures or depend on architecture-specific characteristics, thus impeding practical wide-scale adoption. To address these shortcomings, the OPTIN framework leverages intermediate feature distillation, capturing the long-range dependencies of model parameters (coined $\textit{trajectory}$), to produce state-of-the-art results on natural language, image classification, transfer learning, and semantic segmentation tasks $\textit{without re-training}$. Given a FLOP constraint, the OPTIN framework will compress the network while maintaining competitive accuracy performance and improved throughput. Particularly, we show a $\leq 2$% accuracy degradation from NLP baselines and a $0.5$% improvement from state-of-the-art methods on image classification at competitive FLOPs reductions. We further demonstrate the generalization of tasks and architecture with comparative performance using Mask2Former for semantic segmentation and cnn-style networks. OPTIN presents one of the first one-shot efficient frameworks for compressing transformer architectures that generalizes well across different class domains, in particular: natural language and image-related tasks, without $\textit{re-training}$.

cs.LG↗

arXiv2025

A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks

In recent years, there has been a trend in the field of Reinforcement Learning (RL) towards large action models trained offline on large-scale datasets via sequence modeling. Existing models are primarily based on the Transformer architecture, which result in powerful agents. However, due to slow inference times, Transformer-based approaches are impractical for real-time applications, such as robotics. Recently, modern recurrent architectures, such as xLSTM and Mamba, have been proposed that exhibit parallelization benefits during training similar to the Transformer architecture while offering fast inference. In this work, we study the aptitude of these modern recurrent architectures for large action models. Consequently, we propose a Large Recurrent Action Model (LRAM) with an xLSTM at its core that comes with linear-time inference complexity and natural sequence length extrapolation abilities. Experiments on 432 tasks from 6 domains show that LRAM compares favorably to Transformers in terms of performance and speed.

cs.LG↗

arXiv2026

Architectural Bias in Face Presentation Attack Detection: A Comparative Study of Vision Transformers and Convolutional Neural Networks

Face Presentation Attack Detection (PAD) systems constitute a critical security layer in biometric authentication; however, existing approaches exhibit systematic performance disparities across demographic groups, disproportionately affecting individuals with darker skin tones. This paper presents a comparative empirical investigation of whether Vision Transformer architectures reduce demographic bias in face PAD systems relative to convolutional baselines. Experiments are conducted on the CASIA-SURF Cross-Ethnicity Face Anti-Spoofing (CeFA) dataset. Three architectures are evaluated: a Multimodal ViT-Tiny trained from scratch, a ResNet18 CNN baseline, and a pretrained DeiT-S fine-tuned on CeFA across African, East Asian, and zero-shot Central Asian demographic groups. DeiT-S achieves the highest overall accuracy of 97.27% and the lowest EER of 0.86%, outperforming ResNet18 at 90.15% accuracy. In terms of fairness, DeiT-S reduces the inter-ethnic ACER gap between African and East Asian subjects to 0.13%, compared to 0.75% reported in an LBP-based work [6], representing an 83% reduction. Most notably, while ResNet18 records a BPCER of 10.44% on zero-shot Central Asian subjects, DeiT-S maintains 2.89% on the same unseen group, demonstrating a 3.6x generalization advantage. These results suggest that pretrained Vision Transformers achieve superior PAD accuracy, produce smaller demographic performance gaps, and generalize more equitably across unseen demographic groups, indicating that cross-demographic fairness in PAD may partly be influenced by architectural design.

cs.CV↗

arXiv2020

Stabilizing Transformer-Based Action Sequence Generation For Q-Learning

Since the publication of the original Transformer architecture (Vaswani et al. 2017), Transformers revolutionized the field of Natural Language Processing. This, mainly due to their ability to understand timely dependencies better than competing RNN-based architectures. Surprisingly, this architecture change does not affect the field of Reinforcement Learning (RL), even though RNNs are quite popular in RL, and time dependencies are very common in RL. Recently, Parisotto et al. 2019) conducted the first promising research of Transformers in RL. To support the findings of this work, this paper seeks to provide an additional example of a Transformer-based RL method. Specifically, the goal is a simple Transformer-based Deep Q-Learning method that is stable over several environments. Due to the unstable nature of Transformers and RL, an extensive method search was conducted to arrive at a final method that leverages developments around Transformers as well as Q-learning. The proposed method can match the performance of classic Q-learning on control environments while showing potential on some selected Atari benchmarks. Furthermore, it was critically evaluated to give additional insights into the relation between Transformers and RL.

cs.LG↗

arXiv2022

When Adversarial Training Meets Vision Transformers: Recipes from Training to Architecture

Vision Transformers (ViTs) have recently achieved competitive performance in broad vision tasks. Unfortunately, on popular threat models, naturally trained ViTs are shown to provide no more adversarial robustness than convolutional neural networks (CNNs). Adversarial training is still required for ViTs to defend against such adversarial attacks. In this paper, we provide the first and comprehensive study on the adversarial training recipe of ViTs via extensive evaluation of various training techniques across benchmark datasets. We find that pre-training and SGD optimizer are necessary for ViTs' adversarial training. Further considering ViT as a new type of model architecture, we investigate its adversarial robustness from the perspective of its unique architectural components. We find, when randomly masking gradients from some attention blocks or masking perturbations on some patches during adversarial training, the adversarial robustness of ViTs can be remarkably improved, which may potentially open up a line of work to explore the architectural information inside the newly designed models like ViTs. Our code is available at https://github.com/mo666666/When-Adversarial-Training-Meets-Vision-Transformers.

cs.CV↗

arXiv2023

Full Point Encoding for Local Feature Aggregation in 3D Point Clouds

Point cloud processing methods exploit local point features and global context through aggregation which does not explicity model the internal correlations between local and global features. To address this problem, we propose full point encoding which is applicable to convolution and transformer architectures. Specifically, we propose Full Point Convolution (FPConv) and Full Point Transformer (FPTransformer) architectures. The key idea is to adaptively learn the weights from local and global geometric connections, where the connections are established through local and global correlation functions respectively. FPConv and FPTransformer simultaneously model the local and global geometric relationships as well as their internal correlations, demonstrating strong generalization ability and high performance. FPConv is incorporated in classical hierarchical network architectures to achieve local and global shape-aware learning. In FPTransformer, we introduce full point position encoding in self-attention, that hierarchically encodes each point position in the global and local receptive field. We also propose a shape aware downsampling block which takes into account the local shape and the global context. Experimental comparison to existing methods on benchmark datasets show the efficacy of FPConv and FPTransformer for semantic segmentation, object detection, classification, and normal estimation tasks. In particular, we achieve state-of-the-art semantic segmentation results of 76% mIoU on S3DIS 6-fold and 72.2% on S3DIS Area5.

cs.CV↗

arXiv2024

Advances in Transformers for Robotic Applications: A Review

The introduction of Transformers architecture has brought about significant breakthroughs in Deep Learning (DL), particularly within Natural Language Processing (NLP). Since their inception, Transformers have outperformed many traditional neural network architectures due to their "self-attention" mechanism and their scalability across various applications. In this paper, we cover the use of Transformers in Robotics. We go through recent advances and trends in Transformer architectures and examine their integration into robotic perception, planning, and control for autonomous systems. Furthermore, we review past work and recent research on use of Transformers in Robotics as pre-trained foundation models and integration of Transformers with Deep Reinforcement Learning (DRL) for autonomous systems. We discuss how different Transformer variants are being adapted in robotics for reliable planning and perception, increasing human-robot interaction, long-horizon decision-making, and generalization. Finally, we address limitations and challenges, offering insight and suggestions for future research directions.

cs.RO↗

arXiv2025

SegResMamba: An Efficient Architecture for 3D Medical Image Segmentation

The Transformer architecture has opened a new paradigm in the domain of deep learning with its ability to model long-range dependencies and capture global context and has outpaced the traditional Convolution Neural Networks (CNNs) in many aspects. However, applying Transformer models to 3D medical image datasets presents significant challenges due to their high training time, and memory requirements, which not only hinder scalability but also contribute to elevated CO$_2$ footprint. This has led to an exploration of alternative models that can maintain or even improve performance while being more efficient and environmentally sustainable. Recent advancements in Structured State Space Models (SSMs) effectively address some of the inherent limitations of Transformers, particularly their high memory and computational demands. Inspired by these advancements, we propose an efficient 3D segmentation model for medical imaging called SegResMamba, designed to reduce computation complexity, memory usage, training time, and environmental impact while maintaining high performance. Our model uses less than half the memory during training compared to other state-of-the-art (SOTA) architectures, achieving comparable performance with significantly reduced resource demands.

cs.CV↗

arXiv2025

JoFormer (Journey-based Transformer): Theory and Empirical Analysis on the Tiny Shakespeare Dataset

Transformers have demonstrated remarkable success in sequence modeling, yet effectively incorporating positional information remains a challenging and active area of research. In this paper, we introduce JoFormer, a journey-based Transformer architecture grounded in a recently proposed non-commutative algebra for composing transformations across positions. JoFormer represents relative positions through learnable directional transforms that are sequentially composed along the input, thereby extending and generalizing existing approaches based on relative position representations. We derive the JoFormer attention mechanism from first principles and show that it subsumes standard methods such as rotary transformations as special cases. To evaluate its effectiveness, we compare JoFormer to the RoFormer baseline on the Tiny Shakespeare character-level language modeling task. Our results demonstrate that JoFormer consistently achieves lower perplexity and faster convergence, highlighting the advantages of its more expressive, journey-based treatment of position. Notably, the per-token JoFormer is still a primitive, conceptual variant with layer-independent angles, yet it already demonstrates strong performance-underscoring its promise as a proof of concept for more expressive architectures. We conclude by discussing how JoFormer offers a principled approach to integrating positional structure into Transformer architectures. The code used in this work is available at https://github.com/mahesh-godavarti/joformer.

cs.LG↗

arXiv2025

Leveraging Decoder Architectures for Learned Sparse Retrieval

Learned Sparse Retrieval (LSR) has traditionally focused on small-scale encoder-only transformer architectures. With the advent of large-scale pre-trained language models, their capability to generate sparse representations for retrieval tasks across different transformer-based architectures, including encoder-only, decoder-only, and encoder-decoder models, remains largely unexplored. This study investigates the effectiveness of LSR across these architectures, exploring various sparse representation heads and model scales. Our results highlight the limitations of using large language models to create effective sparse representations in zero-shot settings, identifying challenges such as inappropriate term expansions and reduced performance due to the lack of expansion. We find that the encoder-decoder architecture with multi-tokens decoding approach achieves the best performance among the three backbones. While the decoder-only model performs worse than the encoder-only model, it demonstrates the potential to outperform when scaled to a high number of parameters.

cs.IR↗