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 721 records · Page 40Linked to original sources

arXiv2023

The Efficacy of Transformer-based Adversarial Attacks in Security Domains

Today, the security of many domains rely on the use of Machine Learning to detect threats, identify vulnerabilities, and safeguard systems from attacks. Recently, transformer architectures have improved the state-of-the-art performance on a wide range of tasks such as malware detection and network intrusion detection. But, before abandoning current approaches to transformers, it is crucial to understand their properties and implications on cybersecurity applications. In this paper, we evaluate the robustness of transformers to adversarial samples for system defenders (i.e., resiliency to adversarial perturbations generated on different types of architectures) and their adversarial strength for system attackers (i.e., transferability of adversarial samples generated by transformers to other target models). To that effect, we first fine-tune a set of pre-trained transformer, Convolutional Neural Network (CNN), and hybrid (an ensemble of transformer and CNN) models to solve different downstream image-based tasks. Then, we use an attack algorithm to craft 19,367 adversarial examples on each model for each task. The transferability of these adversarial examples is measured by evaluating each set on other models to determine which models offer more adversarial strength, and consequently, more robustness against these attacks. We find that the adversarial examples crafted on transformers offer the highest transferability rate (i.e., 25.7% higher than the average) onto other models. Similarly, adversarial examples crafted on other models have the lowest rate of transferability (i.e., 56.7% lower than the average) onto transformers. Our work emphasizes the importance of studying transformer architectures for attacking and defending models in security domains, and suggests using them as the primary architecture in transfer attack settings.

cs.CR↗

arXiv2022

Rethinking Attention with Performers

We introduce Performers, Transformer architectures which can estimate regular (softmax) full-rank-attention Transformers with provable accuracy, but using only linear (as opposed to quadratic) space and time complexity, without relying on any priors such as sparsity or low-rankness. To approximate softmax attention-kernels, Performers use a novel Fast Attention Via positive Orthogonal Random features approach (FAVOR+), which may be of independent interest for scalable kernel methods. FAVOR+ can be also used to efficiently model kernelizable attention mechanisms beyond softmax. This representational power is crucial to accurately compare softmax with other kernels for the first time on large-scale tasks, beyond the reach of regular Transformers, and investigate optimal attention-kernels. Performers are linear architectures fully compatible with regular Transformers and with strong theoretical guarantees: unbiased or nearly-unbiased estimation of the attention matrix, uniform convergence and low estimation variance. We tested Performers on a rich set of tasks stretching from pixel-prediction through text models to protein sequence modeling. We demonstrate competitive results with other examined efficient sparse and dense attention methods, showcasing effectiveness of the novel attention-learning paradigm leveraged by Performers.

cs.LG↗

arXiv2025

UniLegs: Universal Multi-Legged Robot Control through Morphology-Agnostic Policy Distillation

Developing controllers that generalize across diverse robot morphologies remains a significant challenge in legged locomotion. Traditional approaches either create specialized controllers for each morphology or compromise performance for generality. This paper introduces a two-stage teacher-student framework that bridges this gap through policy distillation. First, we train specialized teacher policies optimized for individual morphologies, capturing the unique optimal control strategies for each robot design. Then, we distill this specialized expertise into a single Transformer-based student policy capable of controlling robots with varying leg configurations. Our experiments across five distinct legged morphologies demonstrate that our approach preserves morphology-specific optimal behaviors, with the Transformer architecture achieving 94.47% of teacher performance on training morphologies and 72.64% on unseen robot designs. Comparative analysis reveals that Transformer-based architectures consistently outperform MLP baselines by leveraging attention mechanisms to effectively model joint relationships across different kinematic structures. We validate our approach through successful deployment on a physical quadruped robot, demonstrating the practical viability of our morphology-agnostic control framework. This work presents a scalable solution for developing universal legged robot controllers that maintain near-optimal performance while generalizing across diverse morphologies.

cs.RO↗

arXiv2023

Domain Generalisation with Bidirectional Encoder Representations from Vision Transformers

Domain generalisation involves pooling knowledge from source domain(s) into a single model that can generalise to unseen target domain(s). Recent research in domain generalisation has faced challenges when using deep learning models as they interact with data distributions which differ from those they are trained on. Here we perform domain generalisation on out-of-distribution (OOD) vision benchmarks using vision transformers. Initially we examine four vision transformer architectures namely ViT, LeViT, DeiT, and BEIT on out-of-distribution data. As the bidirectional encoder representation from image transformers (BEIT) architecture performs best, we use it in further experiments on three benchmarks PACS, Home-Office and DomainNet. Our results show significant improvements in validation and test accuracy and our implementation significantly overcomes gaps between within-distribution and OOD data.

cs.CV↗

arXiv2025

What Does It Mean to Be a Transformer? Insights from a Theoretical Hessian Analysis

The Transformer architecture has inarguably revolutionized deep learning, overtaking classical architectures like multi-layer perceptrons (MLPs) and convolutional neural networks (CNNs). At its core, the attention block differs in form and functionality from most other architectural components in deep learning--to the extent that, in comparison to MLPs/CNNs, Transformers are more often accompanied by adaptive optimizers, layer normalization, learning rate warmup, etc. The root causes behind these outward manifestations and the precise mechanisms that govern them remain poorly understood. In this work, we bridge this gap by providing a fundamental understanding of what distinguishes the Transformer from the other architectures--grounded in a theoretical comparison of the (loss) Hessian. Concretely, for a single self-attention layer, (a) we first entirely derive the Transformer's Hessian and express it in matrix derivatives; (b) we then characterize it in terms of data, weight, and attention moment dependencies; and (c) while doing so further highlight the important structural differences to the Hessian of classical networks. Our results suggest that various common architectural and optimization choices in Transformers can be traced back to their highly non-linear dependencies on the data and weight matrices, which vary heterogeneously across parameters. Ultimately, our findings provide a deeper understanding of the Transformer's unique optimization landscape and the challenges it poses.

cs.LG↗

arXiv2025

Deformable Mamba for Wide Field of View Segmentation

Recent advancements in the Mamba architecture, with its linear computational complexity, being a promising alternative to transformer architectures suffering from quadratic complexity. While existing works primarily focus on adapting Mamba as vision encoders, the critical role of task-specific Mamba decoders remains under-explored, particularly for distortion-prone dense prediction tasks. This paper addresses two interconnected challenges: (1) The design of a Mamba-based decoder that seamlessly adapts to various architectures (e.g., CNN-, Transformer-, and Mamba-based backbones), and (2) The performance degradation in decoders lacking distortion-aware capability when processing wide-FoV images (e.g., 180° fisheye and 360° panoramic settings). We propose the Deformable Mamba Decoder, an efficient distortion-aware decoder that integrates Mamba's computational efficiency with adaptive distortion awareness. Comprehensive experiments on five wide-FoV segmentation benchmarks validate its effectiveness. Notably, our decoder achieves a +2.5% performance improvement on the 360° Stanford2D3D segmentation benchmark while reducing 72% parameters and 97% FLOPs, as compared to the widely-used decoder heads.

cs.CV↗

arXiv2024

Perceiving Longer Sequences With Bi-Directional Cross-Attention Transformers

We present a novel bi-directional Transformer architecture (BiXT) which scales linearly with input size in terms of computational cost and memory consumption, but does not suffer the drop in performance or limitation to only one input modality seen with other efficient Transformer-based approaches. BiXT is inspired by the Perceiver architectures but replaces iterative attention with an efficient bi-directional cross-attention module in which input tokens and latent variables attend to each other simultaneously, leveraging a naturally emerging attention-symmetry between the two. This approach unlocks a key bottleneck experienced by Perceiver-like architectures and enables the processing and interpretation of both semantics ('what') and location ('where') to develop alongside each other over multiple layers -- allowing its direct application to dense and instance-based tasks alike. By combining efficiency with the generality and performance of a full Transformer architecture, BiXT can process longer sequences like point clouds, text or images at higher feature resolutions and achieves competitive performance across a range of tasks like point cloud part segmentation, semantic image segmentation, image classification, hierarchical sequence modeling and document retrieval. Our experiments demonstrate that BiXT models outperform larger competitors by leveraging longer sequences more efficiently on vision tasks like classification and segmentation, and perform on par with full Transformer variants on sequence modeling and document retrieval -- but require $28\%$ fewer FLOPs and are up to $8.4\times$ faster.

cs.CV↗

arXiv2024

G-Transformer: Counterfactual Outcome Prediction under Dynamic and Time-varying Treatment Regimes

In the context of medical decision making, counterfactual prediction enables clinicians to predict treatment outcomes of interest under alternative courses of therapeutic actions given observed patient history. In this work, we present G-Transformer for counterfactual outcome prediction under dynamic and time-varying treatment strategies. Our approach leverages a Transformer architecture to capture complex, long-range dependencies in time-varying covariates while enabling g-computation, a causal inference method for estimating the effects of dynamic treatment regimes. Specifically, we use a Transformer-based encoder architecture to estimate the conditional distribution of relevant covariates given covariate and treatment history at each time point, then produces Monte Carlo estimates of counterfactual outcomes by simulating forward patient trajectories under treatment strategies of interest. We evaluate G-Transformer extensively using two simulated longitudinal datasets from mechanistic models, and a real-world sepsis ICU dataset from MIMIC-IV. G-Transformer outperforms both classical and state-of-the-art counterfactual prediction models in these settings. To the best of our knowledge, this is the first Transformer-based architecture that supports g-computation for counterfactual outcome prediction under dynamic and time-varying treatment strategies.

cs.LG↗

arXiv2025

Graph-of-Causal Evolution: Challenging Chain-of-Model for Reasoning

In view of the problem that each subchain in the chain-of-model (CoM) relies only on the information of the previous subchain and may lose long-range dependencies due to the causal mask blocking the global context flow between multi-level subchains, this work proposes a graph of causal evolution (GoCE). Its core principle is to map the implicit token representation into a differentiable and sparse causal adjacency matrix, then permeate causal constraints through each layer of calculation using causal-masked attention and causal-MoE. By combining intervention consistency loss test and self-evolution gate, the dynamic balance between causal structure learning and adaptive updating of transformer architecture is realized. The researcher built experimental environments in sandboxes built with Claude Sonnet 4, o4-mini-high, and DeepSeek R1 respectively with the transformer variant architecture introduced in GoCE. It is evaluated on publicly available datasets including CLUTRR, CLADDER, EX-FEVER, and CausalQA and compared with the baseline LLMs. The finding proves that GoCE strengthens the transformer's ability to capture long-range causal dependencies, while the ability to self-evolve is improved. It not only surpasses the design of CoM in terms of design principles, but also provides experience for future research on causal learning and continuous adaptive improvement.

cs.LG↗

arXiv2024

U-DiTs: Downsample Tokens in U-Shaped Diffusion Transformers

Diffusion Transformers (DiTs) introduce the transformer architecture to diffusion tasks for latent-space image generation. With an isotropic architecture that chains a series of transformer blocks, DiTs demonstrate competitive performance and good scalability; but meanwhile, the abandonment of U-Net by DiTs and their following improvements is worth rethinking. To this end, we conduct a simple toy experiment by comparing a U-Net architectured DiT with an isotropic one. It turns out that the U-Net architecture only gain a slight advantage amid the U-Net inductive bias, indicating potential redundancies within the U-Net-style DiT. Inspired by the discovery that U-Net backbone features are low-frequency-dominated, we perform token downsampling on the query-key-value tuple for self-attention that bring further improvements despite a considerable amount of reduction in computation. Based on self-attention with downsampled tokens, we propose a series of U-shaped DiTs (U-DiTs) in the paper and conduct extensive experiments to demonstrate the extraordinary performance of U-DiT models. The proposed U-DiT could outperform DiT-XL/2 with only 1/6 of its computation cost. Codes are available at https://github.com/YuchuanTian/U-DiT.

cs.CV↗

arXiv2023

Short-Term Electricity Load Forecasting Using the Temporal Fusion Transformer: Effect of Grid Hierarchies and Data Sources

Recent developments related to the energy transition pose particular challenges for distribution grids. Hence, precise load forecasts become more and more important for effective grid management. Novel modeling approaches such as the Transformer architecture, in particular the Temporal Fusion Transformer (TFT), have emerged as promising methods for time series forecasting. To date, just a handful of studies apply TFTs to electricity load forecasting problems, mostly considering only single datasets and a few covariates. Therefore, we examine the potential of the TFT architecture for hourly short-term load forecasting across different time horizons (day-ahead and week-ahead) and network levels (grid and substation level). We find that the TFT architecture does not offer higher predictive performance than a state-of-the-art LSTM model for day-ahead forecasting on the entire grid. However, the results display significant improvements for the TFT when applied at the substation level with a subsequent aggregation to the upper grid-level, resulting in a prediction error of 2.43% (MAPE) for the best-performing scenario. In addition, the TFT appears to offer remarkable improvements over the LSTM approach for week-ahead forecasting (yielding a predictive error of 2.52% (MAPE) at the lowest). We outline avenues for future research using the TFT approach for load forecasting, including the exploration of various grid levels (e.g., grid, substation, and household level).

cs.LG↗

arXiv2025

VORTEX: Challenging CNNs at Texture Recognition by using Vision Transformers with Orderless and Randomized Token Encodings

Texture recognition has recently been dominated by ImageNet-pre-trained deep Convolutional Neural Networks (CNNs), with specialized modifications and feature engineering required to achieve state-of-the-art (SOTA) performance. However, although Vision Transformers (ViTs) were introduced a few years ago, little is known about their texture recognition ability. Therefore, in this work, we introduce VORTEX (ViTs with Orderless and Randomized Token Encodings for Texture Recognition), a novel method that enables the effective use of ViTs for texture analysis. VORTEX extracts multi-depth token embeddings from pre-trained ViT backbones and employs a lightweight module to aggregate hierarchical features and perform orderless encoding, obtaining a better image representation for texture recognition tasks. This approach allows seamless integration with any ViT with the common transformer architecture. Moreover, no fine-tuning of the backbone is performed, since they are used only as frozen feature extractors, and the features are fed to a linear SVM. We evaluate VORTEX on nine diverse texture datasets, demonstrating its ability to achieve or surpass SOTA performance in a variety of texture analysis scenarios. By bridging the gap between texture recognition with CNNs and transformer-based architectures, VORTEX paves the way for adopting emerging transformer foundation models. Furthermore, VORTEX demonstrates robust computational efficiency when coupled with ViT backbones compared to CNNs with similar costs. The method implementation and experimental scripts are publicly available in our online repository.

cs.CV↗

arXiv2026

From Next Token Prediction to (STRIPS) World Models

We study whether next-token prediction can yield world models that truly support planning, in a controlled symbolic setting where propositional STRIPS action models are learned from action traces alone and correctness can be evaluated exactly. We introduce two architectures. The first is the STRIPS Transformer, a symbolically aligned model grounded in theoretical results linking transformers and the formal language structure of STRIPS domains. The second is a standard transformer architecture without explicit symbolic structure built in, for which we study different positional encoding schemes and attention aggregation mechanisms. We evaluate both architectures on five classical planning domains, measuring training accuracy, generalization, and planning performance across domains and problem sizes. Interestingly, both approaches can be used to produce models that support planning with off-the-shelf STRIPS planners over exponentially many unseen initial states and goals. Although the STRIPS Transformer incorporates a strong symbolic inductive bias, it is harder to optimize and requires larger datasets to generalize reliably. In contrast, a standard transformer with stick-breaking attention achieves near-perfect training accuracy and strong generalization. Finally, standard transformers without stick-breaking attention do not generalize to long traces, whereas a symbolic STRIPS model extracted from a transformer trained on shorter traces does.

cs.AI↗

arXiv2026

LADY: Linear Attention for Autonomous Driving Efficiency without Transformers

End-to-end autonomous driving has emerged as a promising paradigm. However, state-of-the-art methods rely heavily on Transformer architectures. The inherent quadratic complexity of Transformers restricts their ability to model long-range spatial and temporal dependencies, particularly on resource-constrained edge platforms. Given the inherent demand for efficient temporal modeling in autonomous driving, this computational bottleneck severely constrains real-time deployment. While linear attention mechanisms offer a computationally efficient alternative, existing architectures are predominantly limited to self-attention, lacking the cross-modal capabilities essential for autonomous driving. In this work, we propose LADY, the first fully linear attention-based generative model for end-to-end autonomous driving. LADY incorporates a novel, lightweight linear cross-attention (LICA) mechanism to enable effective cross-modal interaction while preserving linearity. A key advantage of our framework is its ability to fuse long-range temporal contexts during inference with constant computational and memory costs ($O(1)$), regardless of the historical sequence length. Experiments on the NAVSIM and Bench2Drive benchmarks demonstrate that LADY achieves performance comparable to state-of-the-art methods, delivering competitive planning accuracy with significantly reduced latency. Furthermore, efficiency benchmarking on edge devices validates the model's feasibility for resource-limited scenarios.

cs.AI↗

arXiv2026

FixationFormer: Direct Utilization of Expert Gaze Trajectories for Chest X-Ray Classification

Expert eye movements provide a rich, passive source of domain knowledge in radiology, offering a powerful cue for integrating diagnostic reasoning into computer-aided analysis. However, direct integration into CNN-based systems, which historically have dominated the medical image analysis domain, is challenging: gaze recordings are sequential, temporally dense yet spatially sparse, noisy, and variable across experts. As a consequence, most existing image-based models utilize reduced representations such as heatmaps. In contrast, gaze naturally aligns with transformer architectures, as both are sequential in nature and rely on attention to highlight relevant input regions. In this work, we introduce FixationFormer, a transformer-based architecture that represents expert gaze trajectories as sequences of tokens, thereby preserving their temporal and spatial structure. By modeling gaze sequences jointly with image features, our approach addresses sparsity and variability in gaze data while enabling a more direct and fine-grained integration of expert diagnostic cues through explicit cross-attention between the image and gaze token sequences. We evaluate our method on three publicly available benchmark chest X-ray datasets and demonstrate that it achieves state-of-the-art classification performance, highlighting the value of representing gaze as a sequence in transformer-based medical image analysis.

cs.CV↗

arXiv2020

End-to-end Contextual Perception and Prediction with Interaction Transformer

In this paper, we tackle the problem of detecting objects in 3D and forecasting their future motion in the context of self-driving. Towards this goal, we design a novel approach that explicitly takes into account the interactions between actors. To capture their spatial-temporal dependencies, we propose a recurrent neural network with a novel Transformer architecture, which we call the Interaction Transformer. Importantly, our model can be trained end-to-end, and runs in real-time. We validate our approach on two challenging real-world datasets: ATG4D and nuScenes. We show that our approach can outperform the state-of-the-art on both datasets. In particular, we significantly improve the social compliance between the estimated future trajectories, resulting in far fewer collisions between the predicted actors.

cs.CV↗

arXiv2022

You are AllSet: A Multiset Function Framework for Hypergraph Neural Networks

Hypergraphs are used to model higher-order interactions amongst agents and there exist many practically relevant instances of hypergraph datasets. To enable efficient processing of hypergraph-structured data, several hypergraph neural network platforms have been proposed for learning hypergraph properties and structure, with a special focus on node classification. However, almost all existing methods use heuristic propagation rules and offer suboptimal performance on many datasets. We propose AllSet, a new hypergraph neural network paradigm that represents a highly general framework for (hyper)graph neural networks and for the first time implements hypergraph neural network layers as compositions of two multiset functions that can be efficiently learned for each task and each dataset. Furthermore, AllSet draws on new connections between hypergraph neural networks and recent advances in deep learning of multiset functions. In particular, the proposed architecture utilizes Deep Sets and Set Transformer architectures that allow for significant modeling flexibility and offer high expressive power. To evaluate the performance of AllSet, we conduct the most extensive experiments to date involving ten known benchmarking datasets and three newly curated datasets that represent significant challenges for hypergraph node classification. The results demonstrate that AllSet has the unique ability to consistently either match or outperform all other hypergraph neural networks across the tested datasets.

cs.LG↗

arXiv2021

Multi-Stream Transformers

Transformer-based encoder-decoder models produce a fused token-wise representation after every encoder layer. We investigate the effects of allowing the encoder to preserve and explore alternative hypotheses, combined at the end of the encoding process. To that end, we design and examine a $\textit{Multi-stream Transformer}$ architecture and find that splitting the Transformer encoder into multiple encoder streams and allowing the model to merge multiple representational hypotheses improves performance, with further improvement obtained by adding a skip connection between the first and the final encoder layer.

cs.CL↗