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 1,351 records · Page 75Linked to original sources

arXiv2026

Adaptive Computation Depth via Learned Token Routing in Transformers

Standard transformer architectures apply the same number of layers to every token regardless of contextual difficulty. We present Token-Selective Attention (TSA), a learned per-token gate on residual updates between consecutive transformer blocks. Each gate is a lightweight two-layer multi-layer perceptron (MLP) that produces a continuous halting probability, making the mechanism end-to-end differentiable with 1.7% parameter overhead and no changes to the base architecture. Notably, TSA learns difficulty-proportional routing without any explicit depth pressure: even at $λ=0$ (no depth regularisation), the task-loss gradient alone drives the router to skip 20% of token-layer operations. On character-level language modeling, TSA saved 14-23% of token-layer operations (TLOps) across Tiny-Shakespeare and enwik8 at <0.5% quality loss. At matched efficiency, TSA achieved 0.7% lower validation loss than early exit, and the learned routing transfers directly to inference-time sparse execution for real wall-clock speedup.

cs.LG↗

arXiv2026

Revisiting Padded Transformer Expressivity: Which Architectural Choices Matter and Which Don't

Recent work describes what transformers can and cannot compute through connections to boolean circuits, but existing results lack exact characterizations and are sensitive to modeling choices. Padded transformers -- to whose input filler symbols such as ``...'' are appended -- emerge as a useful gadget for establishing equivalences to circuit classes by providing polynomial space for adaptive parallel computation. However, only a limited set of padded transformer idealizations has been studied, leaving open how robustly these equivalences hold under changes to attention type, model width, and uniformity. We find that, under practical assumptions, padded transformers are surprisingly robust to all of these, and identify numeric precision and model depth as the main factors affecting expressivity. Concretely, we prove that polynomially padded $\text{L-uniform}$ constant-precision transformers are equivalent to $\text{L-uniform AC}^0$, while growing-precision ones achieve $\text{L-uniform TC}^0$ regardless of width. Furthermore, looping enables sequential processing analogous to circuits: $\log^d N$-looped constant-precision transformers reach $\text{FO-uniform AC}^d$, and growing-precision ones reach $\text{FO-uniform TC}^d$. Interestingly, growing width or precision beyond logarithmic does not increase expressivity, and all our results hold for both softmax and average hard attention transformers.

cs.LG↗

arXiv2023

The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformers

This paper studies the curious phenomenon for machine learning models with Transformer architectures that their activation maps are sparse. By activation map we refer to the intermediate output of the multi-layer perceptrons (MLPs) after a ReLU activation function, and by sparse we mean that on average very few entries (e.g., 3.0% for T5-Base and 6.3% for ViT-B16) are nonzero for each input to MLP. Moreover, larger Transformers with more layers and wider MLP hidden dimensions are sparser as measured by the percentage of nonzero entries. Through extensive experiments we demonstrate that the emergence of sparsity is a prevalent phenomenon that occurs for both natural language processing and vision tasks, on both training and evaluation data, for Transformers of various configurations, at layers of all depth levels, as well as for other architectures including MLP-mixers and 2-layer MLPs. We show that sparsity also emerges using training datasets with random labels, or with random inputs, or with infinite amount of data, demonstrating that sparsity is not a result of a specific family of datasets. We discuss how sparsity immediately implies a way to significantly reduce the FLOP count and improve efficiency for Transformers. Moreover, we demonstrate perhaps surprisingly that enforcing an even sparser activation via Top-k thresholding with a small value of k brings a collection of desired but missing properties for Transformers, namely less sensitivity to noisy training data, more robustness to input corruptions, and better calibration for their prediction confidence.

cs.LG↗

arXiv2026

Selective Synchronization Attention

The Transformer architecture has become the foundation of modern deep learning, yet its core self-attention mechanism suffers from quadratic computational complexity and lacks grounding in biological neural computation. We propose Selective Synchronization Attention (SSA), a novel attention mechanism that replaces the standard dot-product self-attention with a closed-form operator derived from the steady-state solution of the Kuramoto model of coupled oscillators. In SSA, each token is represented as an oscillator characterized by a learnable natural frequency and phase; the synchronization strength between token pairs, determined by a frequency-dependent coupling and phase-locking condition, serves as the attention weight. This formulation provides three key advantages: (i) natural sparsity arising from the phase-locking threshold, whereby tokens with incompatible frequencies automatically receive zero attention weight without explicit masking; (ii) unified positional-semantic encoding through the natural frequency spectrum, eliminating the need for separate positional encodings; and (iii) a single-pass, closed-form computation that avoids iterative ODE integration, with all components (coupling, order parameter, synchronization) derived from the oscillatory framework. We instantiate SSA within the Oscillatory Synchronization Network (OSN), a drop-in replacement for the Transformer block. Analysis of the synchronization matrices reveals non-uniform, head-diverse coupling patterns even at initialization, demonstrating a stronger architectural inductive bias than the approximately uniform attention produced by randomly initialized Transformers.

cs.LG↗

arXiv2026

Continuous-Depth Transformers with Learned Control Dynamics

We present a hybrid transformer architecture that replaces discrete middle layers with a continuous-depth Neural Ordinary Differential Equation (ODE) block, enabling inference-time control over generation attributes via a learned steering signal. Unlike standard transformers that process representations through fixed discrete layers, our approach treats depth as a continuous variable governed by a learned vector field $F_θ(H, τ, u)$, where $u$ is a low-dimensional control signal injected via explicit concatenation. We validate the architecture through four experiments: (1) gradient flow stability with zero exploding/vanishing gradient events, (2) semantic steering achieving 98\%/88\% accuracy for positive/negative sentiment control, (3) continuous interpolation validated by a negligible 0.068\% trajectory divergence between fixed and adaptive solvers, and (4) efficiency benchmarking demonstrating latency parity with standard discrete baselines. Additionally, we show that adaptive ODE solvers reveal geometric structure in the learned dynamics: the control signal partitions the vector field into distinct dynamical regimes with different curvature characteristics. The adjoint method enables $O(1)$ memory training regardless of integration depth. Our results demonstrate that continuous-depth dynamics with learned control signals provide a viable, efficient mechanism for steerable language generation.

cs.LG↗

arXiv2022

GTrans: Spatiotemporal Autoregressive Transformer with Graph Embeddings for Nowcasting Extreme Events

Spatiotemporal time series nowcasting should preserve temporal and spatial dynamics in the sense that generated new sequences from models respect the covariance relationship from history. Conventional feature extractors are built with deep convolutional neural networks (CNN). However, CNN models have limits to image-like applications where data can be formed with high-dimensional arrays. In contrast, applications in social networks, road traffic, physics, and chemical property prediction where data features can be organized with nodes and edges of graphs. Transformer architecture is an emerging method for predictive models, bringing high accuracy and efficiency due to attention mechanism design. This paper proposes a spatiotemporal model, namely GTrans, that transforms data features into graph embeddings and predicts temporal dynamics with a transformer model. According to our experiments, we demonstrate that GTrans can model spatial and temporal dynamics and nowcasts extreme events for datasets. Furthermore, in all the experiments, GTrans can achieve the highest F1 and F2 scores in binary-class prediction tests than the baseline models.

cs.LG↗

arXiv2022

Transformers in 3D Point Clouds: A Survey

Transformers have been at the heart of the Natural Language Processing (NLP) and Computer Vision (CV) revolutions. The significant success in NLP and CV inspired exploring the use of Transformers in point cloud processing. However, how do Transformers cope with the irregularity and unordered nature of point clouds? How suitable are Transformers for different 3D representations (e.g., point- or voxel-based)? How competent are Transformers for various 3D processing tasks? As of now, there is still no systematic survey of the research on these issues. For the first time, we provided a comprehensive overview of increasingly popular Transformers for 3D point cloud analysis. We start by introducing the theory of the Transformer architecture and reviewing its applications in 2D/3D fields. Then, we present three different taxonomies (i.e., implementation-, data representation-, and task-based), which can classify current Transformer-based methods from multiple perspectives. Furthermore, we present the results of an investigation of the variants and improvements of the self-attention mechanism in 3D. To demonstrate the superiority of Transformers in point cloud analysis, we present comprehensive comparisons of various Transformer-based methods for classification, segmentation, and object detection. Finally, we suggest three potential research directions, providing benefit references for the development of 3D Transformers.

cs.CV↗

arXiv2024

A Survey on Graph Neural Networks and Graph Transformers in Computer Vision: A Task-Oriented Perspective

Graph Neural Networks (GNNs) have gained momentum in graph representation learning and boosted the state of the art in a variety of areas, such as data mining (\emph{e.g.,} social network analysis and recommender systems), computer vision (\emph{e.g.,} object detection and point cloud learning), and natural language processing (\emph{e.g.,} relation extraction and sequence learning), to name a few. With the emergence of Transformers in natural language processing and computer vision, graph Transformers embed a graph structure into the Transformer architecture to overcome the limitations of local neighborhood aggregation while avoiding strict structural inductive biases. In this paper, we present a comprehensive review of GNNs and graph Transformers in computer vision from a task-oriented perspective. Specifically, we divide their applications in computer vision into five categories according to the modality of input data, \emph{i.e.,} 2D natural images, videos, 3D data, vision + language, and medical images. In each category, we further divide the applications according to a set of vision tasks. Such a task-oriented taxonomy allows us to examine how each task is tackled by different GNN-based approaches and how well these approaches perform. Based on the necessary preliminaries, we provide the definitions and challenges of the tasks, in-depth coverage of the representative approaches, as well as discussions regarding insights, limitations, and future directions.

cs.CV↗

arXiv2023

Modeling the Rhythm from Lyrics for Melody Generation of Pop Song

Creating a pop song melody according to pre-written lyrics is a typical practice for composers. A computational model of how lyrics are set as melodies is important for automatic composition systems, but an end-to-end lyric-to-melody model would require enormous amounts of paired training data. To mitigate the data constraints, we adopt a two-stage approach, dividing the task into lyric-to-rhythm and rhythm-to-melody modules. However, the lyric-to-rhythm task is still challenging due to its multimodality. In this paper, we propose a novel lyric-to-rhythm framework that includes part-of-speech tags to achieve better text setting, and a Transformer architecture designed to model long-term syllable-to-note associations. For the rhythm-to-melody task, we adapt a proven chord-conditioned melody Transformer, which has achieved state-of-the-art results. Experiments for Chinese lyric-to-melody generation show that the proposed framework is able to model key characteristics of rhythm and pitch distributions in the dataset, and in a subjective evaluation, the melodies generated by our system were rated as similar to or better than those of a state-of-the-art alternative.

eess.AS↗

arXiv2024

xTrimoGene: An Efficient and Scalable Representation Learner for Single-Cell RNA-Seq Data

Advances in high-throughput sequencing technology have led to significant progress in measuring gene expressions at the single-cell level. The amount of publicly available single-cell RNA-seq (scRNA-seq) data is already surpassing 50M records for humans with each record measuring 20,000 genes. This highlights the need for unsupervised representation learning to fully ingest these data, yet classical transformer architectures are prohibitive to train on such data in terms of both computation and memory. To address this challenge, we propose a novel asymmetric encoder-decoder transformer for scRNA-seq data, called xTrimoGene$^α$ (or xTrimoGene for short), which leverages the sparse characteristic of the data to scale up the pre-training. This scalable design of xTrimoGene reduces FLOPs by one to two orders of magnitude compared to classical transformers while maintaining high accuracy, enabling us to train the largest transformer models over the largest scRNA-seq dataset today. Our experiments also show that the performance of xTrimoGene improves as we scale up the model sizes, and it also leads to SOTA performance over various downstream tasks, such as cell type annotation, perturb-seq effect prediction, and drug combination prediction. xTrimoGene model is now available for use as a service via the following link: https://api.biomap.com/xTrimoGene/apply.

cs.LG↗

arXiv2024

Importance-Aware Image Segmentation-based Semantic Communication for Autonomous Driving

This article studies the problem of image segmentation-based semantic communication in autonomous driving. In real traffic scenes, detecting the key objects (e.g., vehicles, pedestrians and obstacles) is more crucial than that of other objects to guarantee driving safety. Therefore, we propose a vehicular image segmentation-oriented semantic communication system, termed VIS-SemCom, where image segmentation features of important objects are transmitted to reduce transmission redundancy. First, to accurately extract image semantics, we develop a semantic codec based on Swin Transformer architecture, which expands the perceptual field thus improving the segmentation accuracy. Next, we propose a multi-scale semantic extraction scheme via assigning the number of Swin Transformer blocks for diverse resolution features, thus highlighting the important objects' accuracy. Furthermore, the importance-aware loss is invoked to emphasize the important objects, and an online hard sample mining (OHEM) strategy is proposed to handle small sample issues in the dataset. Experimental results demonstrate that the proposed VIS-SemCom can achieve a coding gain of nearly 6 dB with a 60% mean intersection over union (mIoU), reduce the transmitted data amount by up to 70% with a 60% mIoU, and improve the segmentation intersection over union (IoU) of important objects by 4%, compared to traditional transmission scheme.

cs.NI↗

arXiv2025

Quantum Transformer: Accelerating model inference via quantum linear algebra

Powerful generative artificial intelligence from large language models (LLMs) harnesses extensive computational resources for inference. In this work, we investigate the transformer architecture, a key component of these models, under the lens of fault-tolerant quantum computing. We develop quantum subroutines to construct the building blocks in the transformer, including the self-attention, residual connection with layer normalization, and feed-forward network. As an important subroutine, we show how to efficiently implement the Hadamard product and element-wise functions of matrices on quantum computers. Our algorithm prepares an amplitude encoding of the transformer output, which can be measured for prediction or use in the next layer. We find that the matrix norm of the input sequence plays a dominant role in the quantum complexity. With numerical experiments on open-source LLMs, including for bio-informatics applications, we demonstrate the potential of a quantum speedup for transformer inference in practical regimes.

quant-ph↗

arXiv2024

TransformerFAM: Feedback attention is working memory

While Transformers have revolutionized deep learning, their quadratic attention complexity hinders their ability to process infinitely long inputs. We propose Feedback Attention Memory (FAM), a novel Transformer architecture that leverages a feedback loop to enable the network to attend to its own latent representations. This design fosters the emergence of working memory within the Transformer, allowing it to process indefinitely long sequences. TransformerFAM requires no additional weights, enabling seamless integration with pre-trained models. Our experiments show that TransformerFAM significantly improves Transformer performance on long-context tasks across various model sizes (1B, 8B, and 24B). These results showcase the potential to empower Large Language Models (LLMs) to process sequences of unlimited length.

cs.LG↗

arXiv2024

Kraken: Inherently Parallel Transformers For Efficient Multi-Device Inference

Large Transformer networks are increasingly used in settings where low inference latency can improve the end-user experience and enable new applications. However, autoregressive inference is resource intensive and requires parallelism for efficiency. Parallelism introduces collective communication that is both expensive and represents a phase when hardware resources are underutilized. Towards mitigating this, Kraken is an evolution of the standard Transformer architecture that is designed to complement existing tensor parallelism schemes for efficient inference on multi-device systems. By introducing a fixed degree of intra-layer model parallelism, the architecture allows collective operations to be overlapped with compute, decreasing latency and increasing hardware utilization. When trained on OpenWebText, Kraken models reach a similar perplexity as standard Transformers while also preserving their language modeling capabilities when evaluated on the SuperGLUE benchmark. Importantly, when tested on multi-GPU systems using TensorRT-LLM engines, Kraken speeds up Time To First Token by a mean of 35.6% across a range of model sizes, context lengths, and degrees of tensor parallelism.

cs.LG↗

arXiv2026

RoLID-11K: A Dashcam Dataset for Small-Object Roadside Litter Detection

Roadside litter poses environmental, safety and economic challenges, yet current monitoring relies on labour-intensive surveys and public reporting, providing limited spatial coverage. Existing vision datasets for litter detection focus on street-level still images, aerial scenes or aquatic environments, and do not reflect the unique characteristics of dashcam footage, where litter appears extremely small, sparse and embedded in cluttered road-verge backgrounds. We introduce RoLID-11K, the first large-scale dataset for roadside litter detection from dashcams, comprising over 11k annotated images spanning diverse UK driving conditions and exhibiting pronounced long-tail and small-object distributions. We benchmark a broad spectrum of modern detectors, from accuracy-oriented transformer architectures to real-time YOLO models, and analyse their strengths and limitations on this challenging task. Our results show that while CO-DETR and related transformers achieve the best localisation accuracy, real-time models remain constrained by coarse feature hierarchies. RoLID-11K establishes a challenging benchmark for extreme small-object detection in dynamic driving scenes and aims to support the development of scalable, low-cost systems for roadside-litter monitoring. The dataset is available at https://github.com/xq141839/RoLID-11K.

cs.CV↗

arXiv2026

Toward a foundational thermal model for residential buildings

The building energy community lacks a foundational thermal model, i.e., a single pretrained model capable of generalizing across diverse buildings, climates, and control strategies without building-specific calibration. Achieving this vision requires architectural principles that capture universal thermal dynamics rather than memorizing building-specific patterns. We take a step toward this goal by presenting a physics-informed transformer architecture that embeds domain knowledge, e.g., derivative enrichment and Euler-based numerical integration, into a decoder-only framework. We incorporate static building features extracted from simulation models and employ Rotary Position Embedding attention to capture temporal dependencies. Evaluated on the CityLearn dataset spanning 247 residential buildings across three climate zones, our model achieves one-step prediction accuracy (RMSE of 0.30°C in Texas, 0.29°C in Vermont) while outperforming both traditional baselines and fine-tuned Time-Series Foundation Models. We also demonstrate zero-shot transferability: models trained on as few as two buildings generalize to unseen buildings and climate zones without fine-tuning. Despite the limitation of simulated residential buildings, our results establish physics-informed architectural principles as a promising foundation for universal building thermal models.

cs.LG↗

arXiv2026

CDiT: Conditional Diffusion Transformer for Geometry-Aware Terahertz Cross Far- and Near-Field Channel Generation

Accurate channel modeling is fundamental to design and evaluation of Terahertz (THz) ultra-massive multiple-input multiple-output (UM-MIMO) systems. However, existing model-based approaches typically rely on simplified assumptions, such as sparsity or predefined parametric structures, which are insufficient to capture the complex spatial variations and cross far-/near-field propagation characteristics of practical THz channels. In this paper, a conditional diffusion transformer (CDiT) framework is proposed for high-fidelity THz channel generation. By leveraging the state-of-the-art hybrid planar-spherical wave model (HPSM), THz channel modeling is formulated as a geometry-aware conditional generative learning problem in the sparse beamspace domain. Position information is incorporated as a conditioning signal within a diffusion-transformer architecture, enabling effective learning of the spatially dependent channel distribution. By combining the strong distribution modeling capability of diffusion models with the global dependency modeling strength of transformers, the proposed framework achieves controllable and high-fidelity THz channel synthesis. Extensive experiments on realistic THz channel datasets demonstrate that the proposed framework converges stably and significantly outperforms representative benchmark methods. The proposed framework provides a promising data-driven paradigm for THz channel modeling in next-generation wireless systems.

eess.SP↗

arXiv2021

GT U-Net: A U-Net Like Group Transformer Network for Tooth Root Segmentation

To achieve an accurate assessment of root canal therapy, a fundamental step is to perform tooth root segmentation on oral X-ray images, in that the position of tooth root boundary is significant anatomy information in root canal therapy evaluation. However, the fuzzy boundary makes the tooth root segmentation very challenging. In this paper, we propose a novel end-to-end U-Net like Group Transformer Network (GT U-Net) for the tooth root segmentation. The proposed network retains the essential structure of U-Net but each of the encoders and decoders is replaced by a group Transformer, which significantly reduces the computational cost of traditional Transformer architectures by using the grouping structure and the bottleneck structure. In addition, the proposed GT U-Net is composed of a hybrid structure of convolution and Transformer, which makes it independent of pre-training weights. For optimization, we also propose a shape-sensitive Fourier Descriptor (FD) loss function to make use of shape prior knowledge. Experimental results show that our proposed network achieves the state-of-the-art performance on our collected tooth root segmentation dataset and the public retina dataset DRIVE. Code has been released at https://github.com/Kent0n-Li/GT-U-Net.

cs.CV↗