Transformer

Follow this topic Ask about this topic

Transformers represent inputs as tokens and use self-attention to let each token draw on information from other positions, with feed-forward layers refining the resulting representations. Encoder-only models such as BERT, autoregressive decoder-only language models, and encoder-decoder systems are adapted for tasks ranging from text generation and image recognition to forecasting and scientific simulation. Current approaches improve efficiency and capability through local or sparse attention, retrieval and memory, hierarchical processing, and hybrids with convolutional or state-space layers; evaluation spans language-modeling, long-context, vision, and domain-specific benchmarks. A central tension is whether architectural changes and larger models deliver reliable gains in generalization and reasoning while controlling training and inference cost, and theory and interpretability work continues to probe how Transformers learn and compute.

Latest papers

544 papers70 papers in the last four weeks, up 71% on the four weeks before. 0.4% of all new papers.

Apr 24, 2026

Adaptive Head Budgeting for Efficient Multi-Head Attention

Multi-head attention enables Transformers to capture diverse representations, but all attention heads are typically activated for every input, regardless of task complexity. For coarse-grained tasks such as text classification, where relevant information is often global, this fixed allocation can introduce unnecessary computation. We propose BudgetFormer, a Transformer architecture that dynamically allocates attention heads on a per-input basis. The model learns both a head budget and a relevance distribution to select the most informative heads. To support effective head selection, we introduce a training strategy that balances exploration and exploitation. Experiments on text classification tasks show that BudgetFormer reduces FLOPs and memory usage while matching or surpassing the performance of standard multi-head attention. These results highlight adaptive head allocation as an effective approach to improving Transformer efficiency and performance.

Bilal Faye, Abdoulaye Mbaye, Hanane Azzag +1

Transformer Transformer Attention

Apr 21, 2026

VecHeart: Holistic Four-Chamber Cardiac Anatomy Modeling via Hybrid VecSets

Accurate cardiac anatomy modeling requires the model to be able to handle intricate interrelations among structures. In this paper, we propose VecHeart, a unified framework for holistic reconstruction and generation of four-chamber cardiac structures. To overcome the limitations of current feed-forward implicit methods, specifically their restriction to single-object modeling and their neglect of inter-part correlations, we introduce Hybrid Part Transformer, which leverages part-specific learnable queries and interleaved attention to capture complex inter-chamber dependencies. Furthermore, we propose Anatomical Completion Masking and Modality Alignment strategies, enabling the model to infer complete four-chamber structures from partial, sparse, or noisy observations, even when certain anatomical parts are entirely missing. VecHeart also seamlessly extends to 3D+t dynamic mesh sequence generation, demonstrating exceptional versatility. Experiments show that our method achieves state-of-the-art performance, maintaining high-fidelity reconstruction across diverse challenging scenarios. Code is available at https://github.com/Scalsol/VecHeart.

Yihong Chen, Pascal Fua

CVLAB, EPFL

Transformer 3D Reconstruction Medical Image Analysis 3D Medical Imaging

Apr 21, 2026

Neural posterior estimation of the neutrino direction in IceCube using transformer-encoded normalizing flows on the sphere

IceCube is a cubic-kilometer-scale neutrino detector located at the geographic South Pole. A precise directional reconstruction of IceCube neutrinos is vital for associations with astronomical objects. In this context, we discuss neural posterior estimation of the neutrino direction via a transformer encoder that maps to a normalizing flow on the 2-sphere. It achieves a new state-of-the-art angular resolution for the two main event morphologies in IceCube - tracks and showers - while being significantly faster than traditional B-spline-based likelihood reconstructions. All-sky scans can be performed within seconds rather than hours, and take constant computation time, regardless of whether the posterior extent is arc-minutes or spans the whole sky. We utilize a combination of

C^2

-smooth rational-quadratic splines, scale transformations and rotations to define a novel spherical normalizing-flow distribution whose parameters are predicted as a whole as the output of the transformer encoder. We test several structural choices diverting from the vanilla transformer architecture. In particular, we find dual residual streams, nonlinear QKV projection and a separate class token with its own cross-attention processing to boost test-time performance. The angular resolution for both showers and tracks improves substantially over the whole trained energy range from 100 GeV to 100 PeV. At 100 TeV deposited energy, for example, the median angular resolution improves by a factor of

1.3

for throughgoing tracks, by a factor of

1.7

for showers and by a factor of

2.5

for starting tracks compared to state-of-the art likelihood reconstructions based on B-splines. While previous machine-learning (ML) efforts have managed to obtain competitive shower resolutions, this is the first time an ML-based method outperforms likelihood-based muon reconstructions above 100 GeV.

R. Abbasi, M. Ackermann, J. Adams +411

Normalizing Flows Transformer Direction-of-Arrival Estimation High-Energy Physics

Apr 21, 2026

Nexusformer: Nonlinear Attention Expansion for Stable and Inheritable Transformer Scaling

Scaling Transformers typically necessitates training larger models from scratch, as standard architectures struggle to expand without discarding learned representations. We identify the primary bottleneck in the attention mechanism's linear projections, which strictly confine feature extraction to fixed-dimensional subspaces, limiting both expressivity and incremental capacity. To address this, we introduce Nexusformer, which replaces linear

Q/K/V

projections with a Nexus-Rank layer, a three-stage nonlinear mapping driven by dual activations in progressively higher dimensional spaces. This design overcomes the linearity constraint and enables lossless structured growth: new capacity can be injected along two axes via zero-initialized blocks that preserve pretrained knowledge. Experiments on language modeling and reasoning benchmarks demonstrate that Nexusformer matches Tokenformer's perplexity using up to 41.5% less training compute during progressive scaling (240M to 440M). Furthermore, our analysis of growth dynamics reveals that zero initialization induces a stable convergence trajectory, allowing us to derive a geometric scaling law that accurately predicts performance across expansion scales.

Weijie Zhao, Mingquan Liu, Bolun Wang +4

Luxi Tech, Shenzhen, China · The School of Computer Science, China University of Geosciences, 388 Lumo Road, Wuhan, 430074, Hubei, China.

Transformer Language Model Scaling Laws Transformer Attention Language Model Scaling Attention Mechanisms

Apr 21, 2026

One-Block Transformer (1BT) for EEG-Based Cognitive Workload Assessment

Accurate and continuous estimation of cognitive workload is fundamental to creating adaptive human-machine systems. However, designing architectures that balance representational capacity with computational efficiency has been challenging for practical deployment. This paper introduces 1BT, a One-Block Transformer for compact and efficient EEG-based cognitive workload assessment. The model aggregates multi-channel temporal sequences via a minimal latent bottleneck, using a single cross-attention module followed by lightweight self-attention. A controlled study involving 11 participants performing three cognitively diverse tasks (abstract reasoning, numerical problem-solving, and an interactive video game) was conducted with continuous EEG recordings across two workload levels. Systematic architectural analysis identifies the most compact configuration that preserves high performance, while substantially lowering computational cost. The final model achieves high workload classification performance with under 0.5 million parameters and 0.02 GFLOPs, paving the way for a design direction for real-time cognitive workload monitoring in resource-constrained settings.

Stefanos Gkikas, Christian Arzate Cruz, Thomas Kassiotis +3

Honda Research Institute Japan · Wako City, Japan · Department of Electronic Engineering +6

Efficient Neural Network Inference Transformer EEG Decoding

Apr 20, 2026

Towards a Foundation-Model Paradigm for Aerodynamic Prediction in Three-dimensional Design

Accurate machine-learning models for aerodynamic prediction are essential for accelerating shape optimization, yet remain challenging to develop for complex three-dimensional configurations due to the high cost of generating training data. This work introduces a methodology for efficiently constructing accurate surrogate models for design purposes by first pre-training a large-scale model on diverse geometries and then fine-tuning it with a few more detailed task-specific samples. A Transformer-based architecture, AeroTransformer, is developed and tailored for large-scale training to learn aerodynamics. The methodology is evaluated on transonic wings, where the model is pre-trained on SuperWing, a dataset of nearly 30000 samples with broad geometric diversity, and subsequently fine-tuned to handle specific wing shapes perturbed from the Common Research Model. Results show that, with 450 task-specific samples, the proposed methodology achieves 0.36% error on surface-flow prediction, reducing 84.2% compared to training from scratch. The influence of model configurations and training strategies is also systematically studied to provide guidance on effectively training and deploying such models under limited data and computational budgets. To facilitate reuse, we release the datasets and the pre-trained models at https://github.com/tum-pbs/AeroTransformer. An interactive design tool is also built on the pre-trained model and is available online at https://webwing.pbs.cit.tum.de.

Yunjia Yang, Babak Gholami, Caglar Gurbuz +2

Technical University of Munich, Garching, D-85748, Germany · BMW Group, Bremer Str. 6, Munich, D-80788, Germany

Surrogate Modeling Transformer Computational Fluid Dynamics

Apr 18, 2026

In-Context Learning Under Regime Change

Non-stationary sequences arise naturally in control, forecasting, and decision-making. The data-generating process shifts at unknown times, and models must detect the change, discard or downweight obsolete evidence, and adapt to new dynamics on the fly. Transformer-based foundation models increasingly rely on in-context learning for time series forecasting, tabular prediction, and continuous control. As these models are deployed in non-stationary environments, understanding their ability to detect and adapt to regime shifts is important. We formalize this as an in-context change-point detection problem and formally establish the existence of transformer models that solve this problem. Our construction demonstrates that model complexity, in layers and parameters, depends on the level of information available about the change-point location, from no knowledge to knowing exact timing. We validate our results with experiments on synthetic linear regression and linear dynamical systems, where trained transformers match the performance of optimal baselines across information levels. We also show that encoding and incorporating changepoint knowledge indeed improves the real-world performance of a pretrained foundation models on infectious disease forecasting and on financial volatility forecasting around Federal Open Market Committee (FOMC) announcements without retraining, demonstrating practical applicability to real-world regime changes.

Carson Dudley, Yutong Bi, Xiaofeng Liu +1

Department of Mathematics, University of Michigan · Department of Electrical Engineering and Computer Science, University of Michigan · Michigan Institute for AI and Data in Society, University of Michigan

Transformer Distribution Shift Detection Non-Stationary Time Series Forecasting Change-Point Detection Time-Series Foundation Models

Apr 18, 2026

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.

Ahmed Abdelmuniem Abdalla Mohammed

Independent Researcher

Efficient Transformer Inference Transformer Adaptive Computation

Apr 16, 2026

What Is the Minimum Architecture for Prolepsis? Early Irrevocable Commitment Across Tasks in Small Transformers

When do transformers commit to a decision, and what prevents them from correcting it? We introduce prolepsis: a transformer commits early, task-specific attention heads sustain the commitment, and no layer corrects it. Replicating \citeauthor{lindsey2025biology}'s (\citeyear{lindsey2025biology}) planning-site finding on open models (Gemma~~2 2B, Llama~~3.2 1B), we ask five questions. (Q1)~Planning is invisible to six residual-stream methods; CLTs are necessary. (Q2)~The planning-site spike replicates with identical geometry. (Q3)~Specific attention heads route the decision to the output, filling a gap flagged as invisible to attribution graphs. (Q4)~Search requires

{\leq}16

layers; commitment requires more. (Q5)~Factual recall shows the same motif at a different network depth, with zero overlap between recurring planning heads and the factual top-10. Prolepsis is architectural: the template is shared, the routing substrates differ. All experiments run on a single consumer GPU (16,GB VRAM).

Éric Jacopin

Cosmic AI

Transformer Interpretability Attention Head Analysis Transformer LLM Planning Mechanistic Interpretability Attention Mechanisms+1

Apr 16, 2026

Expressivity of Transformers: A Tropical Geometry Perspective

To quantify the geometric expressivity of transformers, we introduce a tropical geometry framework to characterize their exact spatial partitioning capabilities. By modeling self-attention as a vector-valued tropical rational map, we prove it evaluates exactly to a Power Voronoi Diagram in the zero-temperature limit. Building on this equivalence, we establish a combinatorial rationale for Multi-Head Self-Attention (MHSA): via the Minkowski sum of Newton polytopes, multi-head aggregation expands the polyhedral complexity to

\mathcal{O}(N^H)

, overcoming the

\mathcal{O}(N)

bottleneck of single heads. Extending this to deep architectures, we derive the first tight asymptotic bounds on the number of linear regions in transformers (

Θ(N^{d_{\text{model}}L})

), demonstrating a combinatorial explosion driven intrinsically by sequence length

N

, ambient embedding dimension

d_{\text{model}}

, and network depth

L

. Importantly, we guarantee that this idealized polyhedral skeleton is geometrically stable: finite-temperature soft attention preserves these topological partitions via exponentially tight differential approximation bounds.

Ye Su, Yong Liu

Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen 518055, China · University of Chinese Academy of Sciences, Beijing 101407, China · Gaoling School of Artificial Intelligence, Renmin University of China, Beijing 100872, China

Transformer Expressivity Transformer Self-Attention Attention Mechanisms Multi-Head Attention Neural Network Expressivity

Apr 15, 2026

Hierarchical vs. Flat Iteration in Shared-Weight Transformers

We present an empirical study of whether hierarchically structured, shared-weight recurrence can match the representational quality of independent-layer stacking in a Transformer-based language model. HRM-LM replaces L independent Transformer layers with a two-speed recurrent pair: a Fast module operating at every step for local refinement, and a Slow module operating every T steps for global compression. This recurrent hierarchy is unrolled for M = N x T steps with shared parameters. The central and most robust finding, supported by a parameter-matched Universal Transformer ablation (UniTF, 1.2B) across five independent runs, is a sharp empirical gap between the two approaches.

Sang-Il Han

Korea University of Technology and Education · Biomonster Co.

Transformer Recurrent Transformers Language Modeling Hierarchical Representation Learning

Apr 15, 2026

Three-Phase Transformer

We present Three-Phase Transformer (3PT), a residual-stream structural prior for decoder-only Transformers on a standard SwiGLU + RMSNorm + RoPE + GQA backbone. The hidden vector is partitioned into N equally-sized cyclic channels, each maintained by phase-respecting ops: a per-channel RMSNorm, a 2D Givens rotation between attention and FFN that rotates each channel by theta + i*(2*pi/N), and a head-count constraint aligning GQA heads with the partition. The architecture is a self-stabilizing equilibrium between scrambling and re-imposition, not a bolted-on module. The partition carves out a one-dimensional DC subspace orthogonal to the channels, into which we inject a fixed Gabriel's horn profile r(p) = 1/(p+1) as an absolute-position side-channel composing orthogonally with RoPE's relative-position rotation. The canonical N=3 borrows its metaphor from balanced three-phase AC, where three sinusoids 120 degrees apart sum to zero with no anti-correlated pair. At 123M parameters on WikiText-103, 3PT achieves -7.20% perplexity (-2.62% bits-per-byte) over a matched RoPE-Only baseline at +1,536 parameters (0.00124% of total), with 1.93x step-count convergence speedup (1.64x wall-clock). N behaves as a parameter-sharing knob rather than a unique optimum: at 5.5M an N-sweep over {1,2,3,4,6,8,12} is near-monotone with N=1 winning; at 123M a three-seed sweep finds N=3 and N=1 statistically indistinguishable. The load-bearing mechanism is the channel-partitioned residual stream, per-block rotation, per-phase normalization, and horn DC injection. We characterize (a) self-stabilization of the geometry without explicit enforcement, a novel instance of the conservation-law framework for neural networks; (b) a U-shaped depth profile of rotation-angle drift at 12 layers; (c) orthogonal composition with RoPE, attention, and FFN.

Mohammad R. Abu Ayyash

Brains Build Research, Ramallah, Palestine

Transformer Language Modeling Positional Encoding

Apr 15, 2026

Ordinary Least Squares is a Special Case of Transformer

The statistical essence of the Transformer architecture has long remained elusive: Is it a universal approximator, or a neural network version of known computational algorithms? Through rigorous algebraic proof, we show that the latter better describes Transformer's basic nature: Ordinary Least Squares (OLS) is a special case of the single-layer Linear Transformer. Using the spectral decomposition of the empirical covariance matrix, we construct a specific parameter setting where the attention mechanism's forward pass becomes mathematically equivalent to the OLS closed-form projection. This means attention can solve the problem in one forward pass, not by iterating. Building upon this prototypical case, we further uncover a decoupled slow and fast memory mechanism within Transformers. Finally, the evolution from our established linear prototype to standard Transformers is discussed. This progression facilitates the transition of the Hopfield energy function from linear to exponential memory capacity, thereby establishing a clear continuity between modern deep architectures and classical statistical inference.

Xiaojun Tan, Yuchen Zhao

Center for Peak Performance, Department of Psychology and Behavioral Sciences, Zhejiang University · Hangzhou Higgs Asset Management Co., Ltd. · College of Control Science and Engineering, Zhejiang University

Transformer Transformer Attention

Apr 14, 2026

Transformer Based Machine Fault Detection From Audio Input

In recent years, Sound AI is being increasingly used to predict machine failures. By attaching a microphone to the machine of interest, one can get real time data on machine behavior from the field. Traditionally, Convolutional Neural Net (CNN) architectures have been used to analyze spectrogram images generated from the sounds captured and predict if the machine is functioning as expected. CNN architectures seem to work well empirically even though they have biases like locality and parameter-sharing which may not be completely relevant for spectrogram analysis. With the successful application of transformer-based models in the field of image processing starting with Vision Transformer (ViT) in 2020, there has been significant interest in leveraging these in the field of Sound AI. Since transformer-based architectures have significantly lower inductive biases, they are expected to perform better than CNNs at spectrogram analysis given enough data. This paper demonstrates the effectiveness of transformer-driven architectures in analyzing Sound data and compares the embeddings they generate with CNNs on the specific task of machine fault detection.

Kiran Voderhobli Holla

Fault Detection Transformer Audio Classification

Apr 13, 2026

Subcritical Signal Propagation at Initialization in Normalization-Free Transformers

We study signal propagation at initialization in transformers through the averaged partial Jacobian norm (APJN), a measure of gradient amplification across layers. We extend APJN analysis to transformers with bidirectional attention and permutation-symmetric input token configurations by deriving recurrence relations for activation statistics and APJNs across layers. Our theory predicts how attention modifies the asymptotic behavior of the APJN at large depth and matches APJNs measured in deep vision transformers. The criticality picture known from residual networks carries over to transformers: the pre-LayerNorm architecture exhibits power-law APJN growth, whereas transformers with LayerNorm replaced by elementwise

\tanh

-like nonlinearities have stretched-exponential APJN growth, indicating that the latter are subcritical. Applied to Dynamic Tanh (DyT) and Dynamic erf (Derf) transformers, the theory explains why these architectures can be more sensitive to initialization and optimization choices and require careful tuning for stable training.

Sergey Alekseev

Department of Physics and Astronomy Stony Brook University Stony Brook, NY 11794

Transformer Neural Network Initialization Neural Network Training Dynamics Transformer Attention

Apr 13, 2026

Please Make it Sound like Human: Encoder-Decoder vs. Decoder-Only Transformers for AI-to-Human Text Style Transfer

AI-generated text has become common in academic and professional writing, prompting research into detection methods. Less studied is the reverse: systematically rewriting AI-generated prose to read as genuinely human-authored. We build a parallel corpus of 25,140 paired AI-input and human-reference text chunks, identify 11 measurable stylistic markers separating the two registers, and fine-tune three models: BART-base, BART-large, and Mistral-7B-Instruct with QLoRA. BART-large achieves the highest reference similarity -- BERTScore F1 of 0.924, ROUGE-L of 0.566, and chrF++ of 55.92 -- with 17x fewer parameters than Mistral-7B. We show that Mistral-7B's higher marker shift score reflects overshoot rather than accuracy, and argue that shift accuracy is a meaningful blind spot in current style transfer evaluation.

Utsav Paneru

Department of Computer Science, Kathmandu Engineering College, Kalimati, Kathmandu, Nepal

Text Style Transfer Text Generation Language Model Generation Evaluation Transformer Decoder-Only Language Models

Apr 13, 2026

A Lightweight Transformer for Pain Recognition from Brain Activity

Pain is a multifaceted and widespread phenomenon with substantial clinical and societal burden, making reliable automated assessment a critical objective. This paper presents a lightweight transformer architecture that fuses multiple fNIRS representations through a unified tokenization mechanism, enabling joint modeling of complementary signal views without requiring modality-specific adaptations or increasing architectural complexity. The proposed token-mixing strategy preserves spatial, temporal, and time-frequency characteristics by projecting heterogeneous inputs onto a shared latent representation, using a structured segmentation scheme to control the granularity of local aggregation and global interaction. The model is evaluated on the AI4Pain dataset using stacked raw waveform and power spectral density representations of fNIRS inputs. Experimental results demonstrate competitive pain recognition performance while remaining computationally compact, making the approach suitable for real-time inference on both GPU and CPU hardware.

Stefanos Gkikas, Christian Arzate Cruz, Yu Fang +6

Honda Research Institute Japan, Wako City, Japan · BioSIS (Biosensing & Intelligent Systems) Lab, Centre for Intelligent Computing and Systems, University of Canberra, Canberra, Australia · Department of Electronic Engineering, Hellenic Mediterranean University, Chania, Greece

Apr 13, 2026

Efficient Emotion-Aware Iconic Gesture Prediction for Robot Co-Speech

Co-speech gestures increase engagement and improve speech understanding. Most data-driven robot systems generate rhythmic beat-like motion, yet few integrate semantic emphasis. To address this, we propose a lightweight transformer that derives iconic gesture placement and intensity from text and emotion alone, requiring no audio input at inference time. The model outperforms GPT-4o in both semantic gesture placement classification and intensity regression on the BEAT2 dataset, while remaining computationally compact and suitable for real-time deployment on embodied agents.

Edwin C. Montiel-Vazquez, Christian Arzate Cruz, Stefanos Gkikas +3

School of Engineering and Sciences Tecnologico de Monterrey State of Mexico, Mexico · Honda Research Institute Japan Wako City, Japan · Department of Electronic Engineering Hellenic Mediterranean University Chania, Greece

Co-Speech Gesture Generation Transformer

Apr 12, 2026

INCRT: An Incremental Transformer That Determines Its Own Architecture

Transformer architectures are designed by trial and error: the number of attention heads, the depth, and the head size are fixed before training begins, with no mathematical principle to guide the choice. The result is systematic structural redundancy -- between half and four-fifths of all heads in a trained model can be removed without measurable loss -- because the architecture allocates capacity without reference to the actual requirements of the task.This paper introduces INCRT (Incremental Transformer), an architecture that determines its own structure during training. Starting from a single head, INCRT adds one attention head at a time whenever its current configuration is provably insufficient, and prunes heads that have become redundant. Each growth decision is driven by a single, online-computable geometric quantity derived from the task's directional structure, requiring no separate validation phase and no hand-tuned schedule. Two theorems form the theoretical backbone. The first (homeostatic convergence) establishes that the system always reaches a finite stopping configuration that is simultaneously minimal (no redundant heads) and sufficient (no uncaptured directional energy above the threshold). The second (compressed-sensing analogy) provides a geometric upper bound on the number of heads that this configuration can contain, as a function of the spectral complexity of the task. Experiments on SARS-CoV-2 variant classification and SST-2 sentiment analysis confirm both results: the predicted and observed head counts agree within 12% across all benchmarks, and the final architectures match or exceed BERT-base on distribution-specific tasks while using between three and seven times fewer parameters and no pre-training.

Giansalvo Cirrincione

Laboratoire LTI, Université de Picardie Jules Verne, Amiens, France

Transformer Neural Architecture Search Attention Head Pruning

Apr 11, 2026

Integrating SAINT with Tree-Based Models: A Case Study in Employee Attrition Prediction

Employee attrition presents a major challenge for organizations, increasing costs and reducing productivity. Predicting attrition accurately enables proactive retention strategies, but existing machine learning models often struggle to capture complex feature interactions in tabular HR datasets. While tree-based models such as XGBoost and LightGBM perform well on structured data, traditional encoding techniques like one-hot encoding can introduce sparsity and fail to preserve semantic relationships between categorical features. This study explores a hybrid approach by integrating SAINT (Self-Attention and Intersample Attention Transformer)-generated embeddings with tree-based models to enhance employee attrition prediction. SAINT leverages self-attention mechanisms to model intricate feature interactions. In this study, we explore SAINT both as a standalone classifier and as a feature extractor for tree-based models. We evaluate the performance, generalizability, and interpretability of standalone models (SAINT, XGBoost, LightGBM) and hybrid models that combine SAINT embeddings with tree-based classifiers. Experimental results show that standalone tree-based models outperform both the standalone SAINT model and the hybrid approaches in predictive accuracy and generalization. Contrary to expectations, the hybrid models did not improve performance. One possible explanation is that tree-based models struggle to utilize dense, high-dimensional embeddings effectively. Additionally, the hybrid approach significantly reduced interpretability, making model decisions harder to explain. These findings suggest that transformer-based embeddings, while capturing feature relationships, do not necessarily enhance tree-based classifiers. Future research should explore alternative fusion strategies for integrating deep learning with structured data.

Adil Derrazi, Javad Pourmostafa Roshan Sharami

CSAI Department, Tilburg University, The Netherlands

Tabular Prediction Tabular ML Tabular Representation Learning Transformer Gradient-Boosted Decision Trees

Apr 10, 2026

ELT: Elastic Looped Transformers for Visual Generation

We introduce Elastic Looped Transformers (ELT), a highly parameter-efficient class of visual generative models based on a recurrent transformer architecture. While conventional generative models rely on deep stacks of unique transformer layers, our approach employs iterative, weight-shared transformer blocks to drastically reduce parameter counts while maintaining high synthesis quality. To effectively train these models for image and video generation, we propose the idea of Intra-Loop Self Distillation (ILSD), where student configurations (intermediate loops) are distilled from the teacher configuration (maximum training loops) to ensure consistency across the model's depth in a single training step. Our framework yields a family of elastic models from a single training run, enabling Any-Time inference capability with dynamic trade-offs between computational cost and generation quality, with the same parameter count. ELT significantly shifts the efficiency frontier for visual synthesis. With

4\times

reduction in parameter count under iso-inference-compute settings, ELT achieves a competitive FID of

2.0

on class-conditional ImageNet

256 \times 256

and FVD of

72.8

on class-conditional UCF-101.

Sahil Goyal, Swayam Agrawal, Gautham Govind Anil +3

Image Generation Transformer Video Generation Recurrent Transformers Efficient Transformer Inference

Apr 9, 2026

Improving DNS Exfiltration Detection via Transformer Pretraining

We study whether in-domain pretraining of Bidirectional Encoder Representations from Transformer (BERT) model improves subdomain-level detection of exfiltration at low false positive rates. While previous work mostly examines fine-tuned generic Transformers, it does not aim to isolate the effect of pretraining on the downstream task of classification. To address this gap, we develop a controlled pipeline where we freeze operating points on validation and transfer them to the test set, thus enabling clean ablations across different label and pretraining budgets. Our results show significant improvements in the left tail of the Receiver Operating Characteristic (ROC) curve, especially against randomly initialized baseline. Additionally, within pretrained model variants, increasing the number of pretraining steps helps the most when more labeled data are available for fine-tuning.

Miloš Tomić, Aleksa Cvetanović, Predrag Tadić

School of Electrical Engineering, University of Belgrade, Serbia

Network Intrusion Detection Domain-Adaptive Pretraining Language Model Pretraining Transformer

Apr 9, 2026

Sensitivity-Positional Co-Localization in GQA Transformers

We investigate a fundamental structural question in Grouped Query Attention (GQA) transformers: do the layers most sensitive to task correctness coincide with the layers where positional encoding adaptation has the greatest leverage? We term this the co-localization hypothesis and test it on Llama 3.1 8B, a 32-layer GQA model with a 4:1 query-to-key-value head ratio. We introduce \LSLORA, which restricts LoRA adaptation to layers identified via a novel correctness-differential hidden-state metric, and GARFA (GQA-Aware RoPE Frequency Adaptation), which attaches 8 learnable per-KV-head scalar multipliers to each targeted layer. Contrary to the co-localization hypothesis, we discover strong anti-localization: task-sensitive layers concentrate in the late network (

\ell\in\{23\text{-}31\}

) while RoPE-influential layers dominate the early network (

\ell\in\{0\text{-}9\}

), yielding Spearman

r_s = -0.735

(

p = 1.66\times10^{-6}

). Despite this anti-localization, a 4-way cross-layer ablation shows that applying both interventions to the sensitivity-identified layers outperforms all alternative configurations by 4-16 percentage points across six diverse benchmarks (MMLU, GPQA, HumanEval+, MATH, MGSM, ARC), approaching Claude 3.5 Haiku on HumanEval+ (67.1% vs. 68.3%) at $100 total compute cost.

Manoj Chandrashekar Rao

Independent Researcher

Transformer Rotary Positional Embeddings Grouped-Query Attention Low-Rank Adaptation

Apr 8, 2026

MSGL-Transformer: A Multi-Scale Global-Local Transformer for Rodent Social Behavior Recognition

Recognition of rodent behavior is important for understanding neural and behavioral mechanisms. Traditional manual scoring is time-consuming and prone to human error. We propose MSGL-Transformer, a Multi-Scale Global-Local Transformer for recognizing rodent social behaviors from pose-based temporal sequences. The model employs a lightweight transformer encoder with multi-scale attention to capture motion dynamics across different temporal scales. The architecture integrates parallel short-range, medium-range, and global attention branches to explicitly capture behavior dynamics at multiple temporal scales. We also introduce a Behavior-Aware Modulation (BAM) block, inspired by SE-Networks, which modulates temporal embeddings to emphasize behavior-relevant features prior to attention. We evaluate on two datasets: RatSI (5 behavior classes, 12D pose inputs) and CalMS21 (4 behavior classes, 28D pose inputs). On RatSI, MSGL-Transformer achieves 75.4% mean accuracy and F1-score of 0.745 across nine cross-validation splits, outperforming TCN, LSTM, and Bi-LSTM. On CalMS21, it achieves 87.1% accuracy and F1-score of 0.8745, a +10.7% improvement over HSTWFormer, and outperforms ST-GCN, MS-G3D, CTR-GCN, and STGAT. The same architecture generalizes across both datasets with only input dimensionality and number of classes adjusted.

Muhammad Imran Sharif, Doina Caragea

Department of Computer Science, Kansas State University, Manhattan, KS, 66506, USA.

Animal Behavior Analysis Transformer Skeleton-Based Action Recognition Multi-Scale Attention

Apr 7, 2026

Brain-to-Speech: Prosody Feature Engineering and Transformer-Based Reconstruction

This chapter presents a novel approach to brain-to-speech (BTS) synthesis from intracranial electroencephalography (iEEG) data, emphasizing prosody-aware feature engineering and advanced transformer-based models for high-fidelity speech reconstruction. Driven by the increasing interest in decoding speech directly from brain activity, this work integrates neuroscience, artificial intelligence, and signal processing to generate accurate and natural speech. We introduce a novel pipeline for extracting key prosodic features directly from complex brain iEEG signals, including intonation, pitch, and rhythm. To effectively utilize these crucial features for natural-sounding speech, we employ advanced deep learning models. Furthermore, this chapter introduces a novel transformer encoder architecture specifically designed for brain-to-speech tasks. Unlike conventional models, our architecture integrates the extracted prosodic features to significantly enhance speech reconstruction, resulting in generated speech with improved intelligibility and expressiveness. A detailed evaluation demonstrates superior performance over established baseline methods, such as traditional Griffin-Lim and CNN-based reconstruction, across both quantitative and perceptual metrics. By demonstrating these advancements in feature extraction and transformer-based learning, this chapter contributes to the growing field of AI-driven neuroprosthetics, paving the way for assistive technologies that restore communication for individuals with speech impairments. Finally, we discuss promising future research directions, including the integration of diffusion models and real-time inference systems.

Mohammed Salah Al-Radhi, Géza Németh, Andon Tchechmedjiev +1

Department of Telecommunications and Artificial Intelligence, Budapest University of Technology and Economics, Budapest, Hungary · EuroMov Digital Health in Motion, University of Montpellier, IMT Mines Alès, Montpellier, France

EEG Decoding Neural Decoding Transformer Speech Prosody

Apr 7, 2026

iPhoneme: Brain-to-Text Communication for ALS Using ConformerXL Decoding

Brain-computer interfaces (BCIs) for speech restoration hold transformative potential for the approximately 173,000--232,500 individuals worldwide with ALS-related dysarthria. Despite recent progress, high-performance speech BCIs have been demonstrated in only 22--31 patients globally, largely due to limitations in neural decoding accuracy and practical input interfaces. We present iPhoneme, a brain-to-text communication system that jointly addresses these challenges through integrated modeling and interaction design. The system combines a deep learning phoneme decoder based on a modified Conformer architecture (ConformerXL, 192.9M parameters) with a gaze-assisted phoneme input interface that mitigates the Midas touch problem in eye-tracking systems. The acoustic model incorporates a temporal prenet with multi-scale dilated convolutions and bidirectional GRU for neural jitter correction, temporal subsampling for CTC stability, and Pre-RMSNorm stabilization across 12 encoder blocks, trained with AdamW and cosine scheduling. On the interaction side, iPhoneme introduces a chorded gaze-plus-silent-speech paradigm that replaces dwell-time selection, enabling more efficient input. We evaluate the system on the T15 dataset (45 sessions, 8,071 trials) of 256-channel intracranial EEG from speech motor cortex regions. A 6-gram phoneme language model trained on 3.1M sequences, combined with WFST beam search (beam=128), achieves 92.14% phoneme accuracy (7.86% PER) and 73.39% word accuracy (26.61% WER), approximately 3% above prior state-of-the-art. The system operates on CPU with 180 ms latency, demonstrating real-time, high-accuracy brain-to-text communication for ALS.

Yoonmin Cha, Dawit Chun, Sung Park

School of Data Science and Artificial Intelligence, Taejae University

Conformer Brain-Computer Interfaces Neural Decoding Transformer

Apr 6, 2026

Training Transformers in Cosine Coefficient Space

Linear layers hold most of a transformer's parameters. We replace each linear layer with one that stores

K

out of

mn

two-dimensional DCT coefficients per weight matrix and reconstructs the full matrix through an inverse DCT at every forward pass; the

K

coefficients are the trainable parameters. A 4-layer, 128-dim transformer trained from scratch on character-level Shakespeare reaches validation loss

1.604

at

K = mn/2

, against

1.580

for a standard dense baseline -- a gap of

+0.024

at half the trainable parameter count, within the terminal-epoch variation of the dense run. A rank-48 LoRA factorization at the same trainable parameter count reaches only

1.801

(

+0.221

). The structural advantage of sparse-coefficient over low-rank parameterizations at matched

K

is qualitative. We identify rank flexibility as the mechanism. A random orthonormal basis matches the DCT within noise at

K = mn/2

, and a compression sweep through

K = mn/10

and

K = mn/20

shows that subspaces that can host high-rank matrices keep the loss low, while subspaces that flatten into a low-rank block (zigzag-selection variants) converge onto the observed stable rank and the loss line of the rank-48 LoRA reference in lock-step. Among these orthonormal bases, the DCT is preferred because its separable fast transform admits a fused reconstruction kernel: the materialized weight matrix never leaves on-chip memory, so the parameter saving translates into a bandwidth saving as well.

Mohamed Amine Bergach

Illumina, San Diego, CA, USA

Transformer Neural Network Compression

Apr 4, 2026

Collapse-Free Prototype Readout Layer for Transformer Encoders

DDCL-Attention is a prototype-based readout layer for transformer encoders that replaces simple pooling methods, such as mean pooling or class tokens, with a learned compression mechanism. It uses a small set of global prototype vectors and assigns tokens to them through soft probabilistic matching, producing compact token summaries at linear complexity in sequence length. The method offers three main advantages. First, it avoids prototype collapse through an exact decomposition of the training loss into a reconstruction term and a diversity term, ensuring that prototypes remain distinct. Second, its joint training with the encoder is shown to be stable under a practical timescale condition, using Tikhonov's singular perturbation theory and explicit learning-rate constraints. Third, the same framework supports three uses: a final readout layer, a differentiable codebook extending VQ-VAE, and a hierarchical document compressor. Experiments on four datasets confirm the theoretical predictions: the loss decomposition holds exactly, prototype separation grows as expected when the stability condition is met, and the codebook reaches full utilization, outperforming standard hard vector quantization. An additional study on orbital debris classification shows that the method also applies beyond standard NLP and vision tasks, including scientific tabular data.

Giansalvo Cirrincione, Rahul Ranjeev Kumar

Laboratory LTI, Université de Picardie Jules Verne, Amiens, France · Institute of Energy and Resources, Charles Darwin University, Darwin, NT, Australia

Token Compression Prototype Learning Transformer Vector-Quantized Variational Autoencoder

Apr 4, 2026

ART: Adaptive Relational Transformer for Pedestrian Trajectory Prediction with Temporal-Aware Relations

Accurate prediction of real-world pedestrian trajectories is crucial for a wide range of robot-related applications. Recent approaches typically adopt graph-based or transformer-based frameworks to model interactions. Despite their effectiveness, these methods either introduce unnecessary computational overhead or struggle to represent the diverse and time-varying characteristics of human interactions. In this work, we present an Adaptive Relational Transformer (ART), which introduces a Temporal-Aware Relation Graph (TARG) to explicitly capture the evolution of pairwise interactions and an Adaptive Interaction Pruning (AIP) mechanism to reduce redundant computations efficiently. Extensive evaluations on ETH/UCY and NBA benchmarks show that ART delivers state-of-the-art accuracy with high computational efficiency.

Ruochen Li, Ziyi Chang, Junyan Hu +3

Department of Computer Science, Durham University, UK · the School of Computing and Information Systems, Singapore Management University, Singapore

Pedestrian Trajectory Prediction Pedestrian Behavior Modeling Multi-Agent Trajectory Prediction Transformer Spatiotemporal GNNs

Apr 3, 2026

The Spectral Lifecycle of Transformer Training: Transient Compression Waves, Persistent Spectral Gradients, and the Q/K--V Asymmetry

We present the first systematic study of weight matrix singular value spectra during transformer pretraining, tracking full SVD decompositions of every weight matrix at 25-step intervals across three model scales (30M--285M parameters). We discover three phenomena: (1)~Transient Compression Waves: stable rank compression propagates as a traveling wave from early to late layers, creating a dramatic gradient that peaks early then reverses -- late layers eventually over-compress past early layers. (2)~Persistent Spectral Gradients: the power-law exponent~

α

develops a permanent depth gradient forming a non-monotonic inverted-U in deeper models, with peaks shifting toward earlier layers as depth increases. (3)~Q/K--V Functional Asymmetry: value/output projections compress uniformly while query/key projections carry the full depth-dependent dynamics. The dissociation between transient compression and persistent spectral shape reveals that rank and spectral shape encode fundamentally different information about training. We formalize this as a two-timescale dynamical model and derive scaling laws (

Δα\propto L^{0.26}

,

R^2{=}0.99

). We validate on nine models across three families (custom, GPT-2, Pythia; 30M--1B parameters; 8--36 layers), demonstrate that

α

predicts layer importance (

ρ{=}0.69

--

0.84

,

p{<}0.02

), and show that spectral-guided pruning outperforms Last-N heuristics by

1.1{\times}

--

3.6{\times}

across seven models in two families (GPT-2 124M--774M, Pythia 160M--1B), with worst-vs-best gaps up to

23.7{\times}

confirming the causal role of spectral structure.

Yi Liu

Transformer Training Dynamics Transformer Neural Network Training Dynamics Spectral Methods