SEARCH · Search arXiv

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 649 records · Page 36Linked to original sources

arXiv2022

An Exploratory Study on Code Attention in BERT

Many recent models in software engineering introduced deep neural models based on the Transformer architecture or use transformer-based Pre-trained Language Models (PLM) trained on code. Although these models achieve the state of the arts results in many downstream tasks such as code summarization and bug detection, they are based on Transformer and PLM, which are mainly studied in the Natural Language Processing (NLP) field. The current studies rely on the reasoning and practices from NLP for these models in code, despite the differences between natural languages and programming languages. There is also limited literature on explaining how code is modeled. Here, we investigate the attention behavior of PLM on code and compare it with natural language. We pre-trained BERT, a Transformer based PLM, on code and explored what kind of information it learns, both semantic and syntactic. We run several experiments to analyze the attention values of code constructs on each other and what BERT learns in each layer. Our analyses show that BERT pays more attention to syntactic entities, specifically identifiers and separators, in contrast to the most attended token [CLS] in NLP. This observation motivated us to leverage identifiers to represent the code sequence instead of the [CLS] token when used for code clone detection. Our results show that employing embeddings from identifiers increases the performance of BERT by 605% and 4% F1-score in its lower layers and the upper layers, respectively. When identifiers' embeddings are used in CodeBERT, a code-based PLM, the performance is improved by 21-24% in the F1-score of clone detection. The findings can benefit the research community by using code-specific representations instead of applying the common embeddings used in NLP, and open new directions for developing smaller models with similar performance.

cs.SE↗

arXiv2022

COFS: Controllable Furniture layout Synthesis

Scalable generation of furniture layouts is essential for many applications in virtual reality, augmented reality, game development and synthetic data generation. Many existing methods tackle this problem as a sequence generation problem which imposes a specific ordering on the elements of the layout making such methods impractical for interactive editing or scene completion. Additionally, most methods focus on generating layouts unconditionally and offer minimal control over the generated layouts. We propose COFS, an architecture based on standard transformer architecture blocks from language modeling. The proposed model is invariant to object order by design, removing the unnatural requirement of specifying an object generation order. Furthermore, the model allows for user interaction at multiple levels enabling fine grained control over the generation process. Our model consistently outperforms other methods which we verify by performing quantitative evaluations. Our method is also faster to train and sample from, compared to existing methods.

cs.CV↗

arXiv2023

Quality evaluation of point clouds: a novel no-reference approach using transformer-based architecture

With the increased interest in immersive experiences, point cloud came to birth and was widely adopted as the first choice to represent 3D media. Besides several distortions that could affect the 3D content spanning from acquisition to rendering, efficient transmission of such volumetric content over traditional communication systems stands at the expense of the delivered perceptual quality. To estimate the magnitude of such degradation, employing quality metrics became an inevitable solution. In this work, we propose a novel deep-based no-reference quality metric that operates directly on the whole point cloud without requiring extensive pre-processing, enabling real-time evaluation over both transmission and rendering levels. To do so, we use a novel model design consisting primarily of cross and self-attention layers, in order to learn the best set of local semantic affinities while keeping the best combination of geometry and color information in multiple levels from basic features extraction to deep representation modeling.

cs.CV↗

arXiv2023

AGFormer: Efficient Graph Representation with Anchor-Graph Transformer

To alleviate the local receptive issue of GCN, Transformers have been exploited to capture the long range dependences of nodes for graph data representation and learning. However, existing graph Transformers generally employ regular self-attention module for all node-to-node message passing which needs to learn the affinities/relationships between all node's pairs, leading to high computational cost issue. Also, they are usually sensitive to graph noises. To overcome this issue, we propose a novel graph Transformer architecture, termed Anchor Graph Transformer (AGFormer), by leveraging an anchor graph model. To be specific, AGFormer first obtains some representative anchors and then converts node-to-node message passing into anchor-to-anchor and anchor-to-node message passing process. Thus, AGFormer performs much more efficiently and also robustly than regular node-to-node Transformers. Extensive experiments on several benchmark datasets demonstrate the effectiveness and benefits of proposed AGFormer.

cs.LG↗

arXiv2024

Revolutionizing Cyber Threat Detection with Large Language Models: A privacy-preserving BERT-based Lightweight Model for IoT/IIoT Devices

The field of Natural Language Processing (NLP) is currently undergoing a revolutionary transformation driven by the power of pre-trained Large Language Models (LLMs) based on groundbreaking Transformer architectures. As the frequency and diversity of cybersecurity attacks continue to rise, the importance of incident detection has significantly increased. IoT devices are expanding rapidly, resulting in a growing need for efficient techniques to autonomously identify network-based attacks in IoT networks with both high precision and minimal computational requirements. This paper presents SecurityBERT, a novel architecture that leverages the Bidirectional Encoder Representations from Transformers (BERT) model for cyber threat detection in IoT networks. During the training of SecurityBERT, we incorporated a novel privacy-preserving encoding technique called Privacy-Preserving Fixed-Length Encoding (PPFLE). We effectively represented network traffic data in a structured format by combining PPFLE with the Byte-level Byte-Pair Encoder (BBPE) Tokenizer. Our research demonstrates that SecurityBERT outperforms traditional Machine Learning (ML) and Deep Learning (DL) methods, such as Convolutional Neural Networks (CNNs) or Recurrent Neural Networks (RNNs), in cyber threat detection. Employing the Edge-IIoTset cybersecurity dataset, our experimental analysis shows that SecurityBERT achieved an impressive 98.2% overall accuracy in identifying fourteen distinct attack types, surpassing previous records set by hybrid solutions such as GAN-Transformer-based architectures and CNN-LSTM models. With an inference time of less than 0.15 seconds on an average CPU and a compact model size of just 16.7MB, SecurityBERT is ideally suited for real-life traffic analysis and a suitable choice for deployment on resource-constrained IoT devices.

cs.CR↗

arXiv2024

GaitPT: Skeletons Are All You Need For Gait Recognition

The analysis of patterns of walking is an important area of research that has numerous applications in security, healthcare, sports and human-computer interaction. Lately, walking patterns have been regarded as a unique fingerprinting method for automatic person identification at a distance. In this work, we propose a novel gait recognition architecture called Gait Pyramid Transformer (GaitPT) that leverages pose estimation skeletons to capture unique walking patterns, without relying on appearance information. GaitPT adopts a hierarchical transformer architecture that effectively extracts both spatial and temporal features of movement in an anatomically consistent manner, guided by the structure of the human skeleton. Our results show that GaitPT achieves state-of-the-art performance compared to other skeleton-based gait recognition works, in both controlled and in-the-wild scenarios. GaitPT obtains 82.6% average accuracy on CASIA-B, surpassing other works by a margin of 6%. Moreover, it obtains 52.16% Rank-1 accuracy on GREW, outperforming both skeleton-based and appearance-based approaches.

cs.CV↗

arXiv2023

PolyMaX: General Dense Prediction with Mask Transformer

Dense prediction tasks, such as semantic segmentation, depth estimation, and surface normal prediction, can be easily formulated as per-pixel classification (discrete outputs) or regression (continuous outputs). This per-pixel prediction paradigm has remained popular due to the prevalence of fully convolutional networks. However, on the recent frontier of segmentation task, the community has been witnessing a shift of paradigm from per-pixel prediction to cluster-prediction with the emergence of transformer architectures, particularly the mask transformers, which directly predicts a label for a mask instead of a pixel. Despite this shift, methods based on the per-pixel prediction paradigm still dominate the benchmarks on the other dense prediction tasks that require continuous outputs, such as depth estimation and surface normal prediction. Motivated by the success of DORN and AdaBins in depth estimation, achieved by discretizing the continuous output space, we propose to generalize the cluster-prediction based method to general dense prediction tasks. This allows us to unify dense prediction tasks with the mask transformer framework. Remarkably, the resulting model PolyMaX demonstrates state-of-the-art performance on three benchmarks of NYUD-v2 dataset. We hope our simple yet effective design can inspire more research on exploiting mask transformers for more dense prediction tasks. Code and model will be made available.

cs.CV↗

arXiv2024

PIXART-δ: Fast and Controllable Image Generation with Latent Consistency Models

This technical report introduces PIXART-δ, a text-to-image synthesis framework that integrates the Latent Consistency Model (LCM) and ControlNet into the advanced PIXART-α model. PIXART-α is recognized for its ability to generate high-quality images of 1024px resolution through a remarkably efficient training process. The integration of LCM in PIXART-δ significantly accelerates the inference speed, enabling the production of high-quality images in just 2-4 steps. Notably, PIXART-δ achieves a breakthrough 0.5 seconds for generating 1024x1024 pixel images, marking a 7x improvement over the PIXART-α. Additionally, PIXART-δ is designed to be efficiently trainable on 32GB V100 GPUs within a single day. With its 8-bit inference capability (von Platen et al., 2023), PIXART-δ can synthesize 1024px images within 8GB GPU memory constraints, greatly enhancing its usability and accessibility. Furthermore, incorporating a ControlNet-like module enables fine-grained control over text-to-image diffusion models. We introduce a novel ControlNet-Transformer architecture, specifically tailored for Transformers, achieving explicit controllability alongside high-quality image generation. As a state-of-the-art, open-source image generation model, PIXART-δ offers a promising alternative to the Stable Diffusion family of models, contributing significantly to text-to-image synthesis.

cs.CV↗

arXiv2025

Generic controllability of equivariant systems and applications to particle systems and neural networks

There exist many examples of systems which have some symmetries, and which one may monitor with symmetry preserving controls. Since symmetries are preserved along the evolution, full controllability is not possible, and controllability has to be considered inside sets of states with same symmetries. We prove that generic systems with symmetries are controllable in this sense. This result has several applications, for instance: (i) generic controllability of particle systems when the kernel of interaction between particles plays the role of a mean-field control; (ii) generic controllability for families of vector fields on manifolds with boundary; (iii) universal interpolation for neural networks architectures with "generic" self attention-type layers - a type of layers ubiquitous in recent neural networks architectures, e.g., in the Transformers architecture. The tools we develop could help address various other questions of control of equivariant systems.

math.DS↗

arXiv2024

Reducing catastrophic forgetting of incremental learning in the absence of rehearsal memory with task-specific token

Deep learning models generally display catastrophic forgetting when learning new data continuously. Many incremental learning approaches address this problem by reusing data from previous tasks while learning new tasks. However, the direct access to past data generates privacy and security concerns. To address these issues, we present a novel method that preserves previous knowledge without storing previous data. This method is inspired by the architecture of a vision transformer and employs a unique token capable of encapsulating the compressed knowledge of each task. This approach generates task-specific embeddings by directing attention differently based on the task associated with the data, thereby effectively mimicking the impact of having multiple models through tokens. Our method incorporates a distillation process that ensures efficient interactions even after multiple additional learning steps, thereby optimizing the model against forgetting. We measured the performance of our model in terms of accuracy and backward transfer using a benchmark dataset for different task-incremental learning scenarios. Our results demonstrate the superiority of our approach, which achieved the highest accuracy and lowest backward transfer among the compared methods. In addition to presenting a new model, our approach lays the foundation for various extensions within the spectrum of vision-transformer architectures.

cs.LG↗

arXiv2025

PCA-Featured Transformer for Jamming Detection in 5G UAV Networks

Unmanned Aerial Vehicles (UAVs) face significant security risks from jamming attacks, which can compromise network functionality. Traditional detection methods often fall short when confronting AI-powered jamming that dynamically modifies its behavior, while contemporary machine learning approaches frequently demand substantial feature engineering and struggle with temporal patterns in attack signatures. The vulnerability extends to 5G networks employing Time Division Duplex (TDD) or Frequency Division Duplex (FDD), where service quality may deteriorate due to deliberate interference. We introduce a novel U-shaped transformer architecture that leverages Principal Component Analysis (PCA) to refine feature representations for improved wireless security. The training process is regularized by incorporating the output entropy uncertainty into the loss function, a mechanism inspired by the Soft Actor-Critic (SAC) algorithm in Reinforcement Learning (RL) to enable robust jamming detection techniques. The architecture features a modified transformer encoder specially designed to process critical wireless signal features, including Received Signal Strength Indicator (RSSI) and Signal-to- Interference-plus-Noise Ratio (SINR) measurements. We complement this with a custom positional encoding mechanism that specifically accounts for the inherent periodicity of wireless signals,enabling a more accurate representation of temporal signal patterns. In addition, we propose a batch size scheduler and implement chunking techniques to optimize convergence for time series data. These advancements contribute to up to a ten times improvement in training speed within the advanced U-shaped encoder-decoder transformer model introduced in this study. Experimental evaluations demonstrate the effectiveness of our entropy-based approach, achieving detection rates of 85.06% in NLoS scenarios.

cs.LG↗

arXiv2025

Vision Transformer Neural Architecture Search for Out-of-Distribution Generalization: Benchmark and Insights

While ViTs have achieved across machine learning tasks, deploying them in real-world scenarios faces a critical challenge: generalizing under OoD shifts. A crucial research gap exists in understanding how to design ViT architectures, both manually and automatically, for better OoD generalization. To this end, we introduce OoD-ViT-NAS, the first systematic benchmark for ViTs NAS focused on OoD generalization. This benchmark includes 3000 ViT architectures of varying computational budgets evaluated on 8 common OoD datasets. Using this benchmark, we analyze factors contributing to OoD generalization. Our findings reveal key insights. First, ViT architecture designs significantly affect OoD generalization. Second, ID accuracy is often a poor indicator of OoD accuracy, highlighting the risk of optimizing ViT architectures solely for ID performance. Third, we perform the first study of NAS for ViTs OoD robustness, analyzing 9 Training-free NAS methods. We find that existing Training-free NAS methods are largely ineffective in predicting OoD accuracy despite excelling at ID accuracy. Simple proxies like Param or Flop surprisingly outperform complex Training-free NAS methods in predicting OoD accuracy. Finally, we study how ViT architectural attributes impact OoD generalization and discover that increasing embedding dimensions generally enhances performance. Our benchmark shows that ViT architectures exhibit a wide range of OoD accuracy, with up to 11.85% improvement for some OoD shifts. This underscores the importance of studying ViT architecture design for OoD. We believe OoD-ViT-NAS can catalyze further research into how ViT designs influence OoD generalization.

cs.LG↗

arXiv2025

Do Protein Transformers Have Biological Intelligence?

Deep neural networks, particularly Transformers, have been widely adopted for predicting the functional properties of proteins. In this work, we focus on exploring whether Protein Transformers can capture biological intelligence among protein sequences. To achieve our goal, we first introduce a protein function dataset, namely Protein-FN, providing over 9000 protein data with meaningful labels. Second, we devise a new Transformer architecture, namely Sequence Protein Transformers (SPT), for computationally efficient protein function predictions. Third, we develop a novel Explainable Artificial Intelligence (XAI) technique called Sequence Score, which can efficiently interpret the decision-making processes of protein models, thereby overcoming the difficulty of deciphering biological intelligence bided in Protein Transformers. Remarkably, even our smallest SPT-Tiny model, which contains only 5.4M parameters, demonstrates impressive predictive accuracy, achieving 94.3% on the Antibiotic Resistance (AR) dataset and 99.6% on the Protein-FN dataset, all accomplished by training from scratch. Besides, our Sequence Score technique helps reveal that our SPT models can discover several meaningful patterns underlying the sequence structures of protein data, with these patterns aligning closely with the domain knowledge in the biology community. We have officially released our Protein-FN dataset on Hugging Face Datasets https://huggingface.co/datasets/Protein-FN/Protein-FN. Our code is available at https://github.com/fudong03/BioIntelligence.

cs.LG↗

arXiv2025

Pose Matters: Evaluating Vision Transformers and CNNs for Human Action Recognition on Small COCO Subsets

This study explores human action recognition using a three-class subset of the COCO image corpus, benchmarking models from simple fully connected networks to transformer architectures. The binary Vision Transformer (ViT) achieved 90% mean test accuracy, significantly exceeding multiclass classifiers such as convolutional networks (approximately 35%) and CLIP-based models (approximately 62-64%). A one-way ANOVA (F = 61.37, p < 0.001) confirmed these differences are statistically significant. Qualitative analysis with SHAP explainer and LeGrad heatmaps indicated that the ViT localizes pose-specific regions (e.g., lower limbs for walking or running), while simpler feed-forward models often focus on background textures, explaining their errors. These findings emphasize the data efficiency of transformer representations and the importance of explainability techniques in diagnosing class-specific failures.

cs.CV↗

arXiv2026

ThinkingViT: Matryoshka Thinking Vision Transformer for Elastic Inference

ViTs deliver SOTA performance, yet their fixed computational budget prevents scalable deployment across heterogeneous hardware. Recent Matryoshka-style Transformer architectures mitigate this by embedding nested subnetworks within a single model to enable scalable inference. However, these models allocate the same amount of compute to all inputs, regardless of their complexity, which leads to inefficiencies. To address this, we introduce ThinkingViT, a nested ViT architecture that employs progressive thinking stages to dynamically adjust inference computation based on input difficulty. ThinkingViT first activates a small subset of the most important attention heads to produce an initial prediction. If the prediction confidence exceeds a predefined threshold, inference terminates early. Otherwise, within the same backbone, it activates a larger subset of attention heads and conducts a new forward pass. This process continues iteratively until the model reaches the predefined confidence level or exhausts its maximum capacity. To boost the performance of subsequent rounds, we introduce a Token Recycling approach that fuses the input embeddings with the embeddings from the previous stage. Experiments show that ThinkingViT surpasses nested baselines by up to 2.0 percentage points (p.p.) in accuracy at the same throughput and by up to 2.9 p.p. at equal GMACs on ImageNet-1K. We show that the backbone-preserving design of ThinkingViT allows it to serve as a plug-in upgrade for ViTs in downstream tasks such as semantic segmentation. We also demonstrate that ThinkingViT transfers effectively to other architectures such as Swin Transformers. The source code is available at https://github.com/ds-kiel/ThinkingViT.

cs.CV↗

arXiv2025

SINAI at eRisk@CLEF 2023: Approaching Early Detection of Gambling with Natural Language Processing

This paper describes the participation of the SINAI team in the eRisk@CLEF lab. Specifically, one of the proposed tasks has been addressed: Task 2 on the early detection of signs of pathological gambling. The approach presented in Task 2 is based on pre-trained models from Transformers architecture with comprehensive preprocessing data and data balancing techniques. Moreover, we integrate Long-short Term Memory (LSTM) architecture with automodels from Transformers. In this Task, our team has been ranked in seventh position, with an F1 score of 0.126, out of 49 participant submissions and achieves the highest values in recall metrics and metrics related to early detection.

cs.CL↗

arXiv2026

A Constrained Optimization Perspective of Unrolled Transformers

We introduce a constrained optimization framework for training transformers that behave like optimization descent algorithms. Specifically, we enforce layerwise descent constraints on the objective function and replace standard empirical risk minimization (ERM) with a primal-dual training scheme. This approach yields models whose intermediate representations decrease the loss monotonically in expectation across layers. We apply our method to both unrolled transformer architectures and conventional pretrained transformers on tasks of video denoising and text classification. Across these settings, we observe constrained transformers achieve stronger robustness to perturbations and maintain higher out-of-distribution generalization, while preserving in-distribution performance.

cs.LG↗

arXiv2026

Improving Parameter Utilization by Sharing Neural Experts Across Layers in Transformers

Transformer-based large language models often suffer from inter-layer parameter redundancy, where functional transformations are redundantly learned across network depths. We propose CS-MoE, a novel Transformer architecture featuring cross-layer expert sharing to address this inefficiency. Deviating from the widely used Mixture-of-Experts (MoE) architecture that terminates each Transformer block with layer-isolated experts, CS-MoE combines layer-independent experts with concurrent access to a centralized, globally shared expert pool. This \textit{Global Experts Sharing} mechanism enables elastic control over token-level parameter activation and computational consumption (FLOPs). Experiments demonstrate that CS-MoE achieves lower perplexity than equal-scale dense Transformers while activating only 55\% of parameters. Furthermore, its performance scales monotonically with an increased number of activated experts and approaches MoE counterparts that consume more FLOPs by expanding the shared pool with a fixed FLOPs budget. CS-MoE also establishes a flexible Pareto frontier between computational cost and model capacity, offering an efficient alternative for computation-constrained environments.

cs.LG↗