Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Proceedings of the AAAI Conference on Artificial Intelligence

UQ-ViT: Harmonizing Extreme Activations with Hardware-Friendly Uniform Quantization in Vision Transformers

Tao Jiang, Yucheng Jiang, Xiwen Yao, Gong Cheng, Junwei Han

Published Mar 14, 2026• 1 citation
Publisher

Abstract

Post-Training Quantization enables efficient Vision Transformer (ViTs) deployment with a small calibration data, and its prevalent use of uniform quantization harnesses AI accelerator matrix cores for high-speed inference. However, the application of uniform quantization is fundamentally challenged by the extreme non-uniformity of activation distributions.Specifically, the power-law nature of post-Softmax attention scores and the significant inter-channel variance in post-GELU activations create a dilemma for conventional quantization, as it struggles to preserve critical high-magnitude values without sacrificing overall precision. To resolve this core conflict, we introduce UQ-ViT (Uniform Quantization for Vision Transformers), a novel uniform quantization framework designed to reconcile high precision with hardware efficiency. Central to UQ-ViT are two operators: Dynamic Elimination of Maximum (DeMax) and Normalization Quantization (NormQuant). DeMax is a quantization operator for post-Softmax attention scores that utilizes uniform quantization. It dynamically eliminates and preserves dominant values, effectively mitigating quantization loss from the extreme values in the power-law distribution. NormQuant utilizes a per-channel quantization strategy during quantization and reverts to a per-tensor format for dequantization, achieving both high accuracy and computational efficiency. Crucially, it is applicable to any linear layer, enabling effective quantization of post-GELU activations in ViTs. Through extensive experiments on various ViTs and vision tasks, including image classification, object detection, and instance segmentation, we demonstrate that our proposed approach outperforms existing methods, achieving superior accuracy while ensuring hardware friendliness.

DOI
10.1609/aaai.v40i27.39393

Related papers

GO-ViT: 이상치 채널 그룹화를 통한 비전 트랜스포머 완전 양자화

기준하

In this study, we focus on two important branches in quantization. The first is to enable lower bit operations to take advantage of lower memory usage, increased computation speed, and reduced model size. And the second is to fully quantize to enable integer-only operations and utilize On-device utilities that enables integer-accelerated operations. In particular, the layer norm input activation required for full quantization is large in value and the presence of outliers is the main cause of performance degradation in low bit settings. In particular, since the distribution of outliers in actviations is consistent in specific channels, more efficient quantization is possible when this information is utilized in the quantization process. The GO-ViT method uses a groupwise scaling method that distinguishes inliers and outliers and determines scaling factors for each corresponding channel to improve performance even in low bits. To verify the versatile applicability of the GO-ViT methodology, which can be used in both Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT), we conducted experiments on various datasets such as ImageNet, CIFAR-100, and Something-Something v2 and several transformer architectures such as Swin, ViT, DeiT, and Video Swin. The experimental results show that GO-ViT can improve the Top-1 accuracy at low bit compared to the PTQ baseline model. In addition, it shows higher performance than the QAT baseline model tested on the Cifar100 dataset by preserving high values of activations while starting with high accuracy from the beginning.

•Jan 1, 2025•0 citations

A novel noise-resilient quantum-classical vision transformer with quantum state tomography for high-precision lung cancer classification

Sumaya Mustofa, Yousuf Rayhan Emon, Md․Zahid Hasan +1 more

Lung cancer remains the leading cause of cancer-related mortality worldwide, with accurate subtype classification being critical for improving patient outcomes. Vision Transformers (ViTs) have recently advanced medical image analysis, but their integration with quantum computing remains underexplored, leaving a gap in leveraging quantum advantages for clinical applications. This study proposes a quantum-enhanced Vision Transformer (QViT) that embeds variational quantum circuits (VQCs) within attention and feed-forward layers to enrich feature learning. To assess robustness, two circuit families, Basic VQC and Quantum Approximate Optimization Algorithm (QAOA) based VQC, were evaluated across two depths, yielding four QViT configurations. The models were analyzed through cancer classification tasks, systematic quantum state tomography (QST) to examine quantum state properties, and noise resilience testing under realistic Noisy Intermediate-Scale Quantum (NISQ) conditions, reflecting current hardware limitations. On 3150 CT (Computed Tomography) images, the QViT-QAOA-D1 configuration achieved 98.52% accuracy with a training time of 674 s, outperforming other configurations. Quantum analyses further showed that Basic VQC provides stronger entanglement but requires deeper, less NISQ-friendly circuits, whereas QAOA maintains high-purity states at shallower depths. These results suggest that QViT provides a practical and scalable framework, where quantum-enhanced representations can improve diagnostic performance while remaining compatible with current hardware constraints. Future work will focus on validation using patient-level datasets and deployment on real quantum hardware to further assess clinical applicability and scalability.

Computers & Electrical Engineering•Jun 8, 2026•0 citations

Quantum-Enhanced Vision Transformer for Flood Detection Using Remote Sensing Imagery

Soumyajit Maity, Behzad Ghanbarian

Reliable flood detection is critical for disaster management, yet classical deep learning models often struggle with the high-dimensional, nonlinear complexities inherent in remote sensing data. To mitigate these limitations, we introduced a novel Quantum-Enhanced Vision Transformer (ViT) that synergizes the global context-awareness of transformers with the expressive feature extraction capabilities of quantum computing. Using remote sensing imagery, we developed a hybrid architecture that processes inputs through parallel pathways, a ViT backbone and a quantum branch utilizing a 4-qubit parameterized quantum circuit for localized feature mapping. These distinct representations were fused to optimize binary classification. Results showed that the proposed hybrid model significantly outperformed a classical ViT baseline, increased overall accuracy from 84.48 % to 94.47 % and the F1-score from 0.841 to 0.944. Notably, the quantum integration substantially improved discriminative power in complex terrains for both class. These findings validate the potential of quantum-classical hybrid models to enhance precision in hydrological monitoring and earth observation applications.

arXiv:2603.13689•IEEE Dallas Circuits and Systems Conference (DCAS)•Apr 10, 2026•1 citation

Quantum Doubly Stochastic Transformers

Jannis Born, Filip Skogh, Kahn Rhrissorrakrai +3 more

At the core of the Transformer, the softmax normalizes the attention matrix to be right stochastic. Previous research has shown that this often de-stabilizes training and that enforcing the attention matrix to be doubly stochastic (through Sinkhorn's algorithm) consistently improves performance across different tasks, domains and Transformer flavors. However, Sinkhorn's algorithm is iterative, approximative, non-parametric and thus inflexible w.r.t. the obtained doubly stochastic matrix (DSM). Recently, it has been proven that DSMs can be obtained with a parametric quantum circuit, yielding a novel quantum inductive bias for DSMs with no known classical analogue. Motivated by this, we demonstrate the feasibility of a hybrid classical-quantum doubly stochastic Transformer (QDSFormer) that replaces the softmax in the self-attention layer with a variational quantum circuit. We study the expressive power of the circuit and find that it yields more diverse DSMs that better preserve information than classical operators. Across multiple small-scale object recognition tasks, we find that our QDSFormer consistently surpasses both a standard ViT and other doubly stochastic Transformers. Beyond the Sinkformer, this comparison includes a novel quantum-inspired doubly stochastic Transformer (based on QR decomposition) that can be of independent interest. Our QDSFormer also shows improved training stability and lower performance variation suggesting that it may mitigate the notoriously unstable training of ViTs on small-scale data.

arXiv:2504.16275•neural information processing systems•Jan 1, 2025•2 citations

Hybrid Vision Transformer and Quantum Convolutional Neural Network for Image Classification

Wang, Mingzhu, Shang, Yun

Quantum machine learning (QML) holds promise for computational advantage, yet progress on real-world tasks is hindered by classical preprocessing and noisy devices. We introduce ViT-QCNN-FT, a hybrid framework that integrates a fine-tuned Vision Transformer with a quantum convolutional neural network (QCNN) to compress high-dimensional images into features suited for noisy intermediate-scale quantum (NISQ) devices. By systematically probing entanglement, we show that ansatzes with uniformly distributed entanglement entropy consistently deliver superior non-local feature fusion and state-of-the-art accuracy (99.77% on CIFAR-10). Surprisingly, quantum noise emerges as a double-edged factor: in some cases, it enhances accuracy (+2.71% under amplitude damping). Strikingly, substituting the QCNN with classical counterparts of equal parameter count leads to a dramatic 29.36% drop, providing unambiguous evidence of quantum advantage. Our study establishes a principled pathway for co-designing classical and quantum architectures, pointing toward practical QML capable of tackling complex, high-dimensional learning tasks.

arXiv:2510.12291•Oct 14, 2025•1 citation

Rethinking Practical and Efficient Quantization Calibration for Vision-Language Models

Zhenhao Shang, Haizhao Jing, Guoting Wei +4 more

Post-training quantization (PTQ) is a primary approach for deploying large language models without fine-tuning, and the quantized performance is often strongly affected by the calibration in PTQ. By contrast, in vision-language models (VLMs), substantial differences between visual and text tokens in their activation distributions and sensitivities to quantization error pose significant challenges for effective calibration during PTQ. In this work, we rethink what PTQ calibration should align with in VLMs and propose the Token-level Importance-aware Layer-wise Quantization framework (TLQ). Guided by gradient information, we design a token-level importance integration mechanism for quantization error, and use it to construct a token-level calibration set, enabling a more fine-grained calibration strategy. Furthermore, TLQ introduces a multi-GPU, quantization-exposed layer-wise calibration scheme. This scheme keeps the layer-wise calibration procedure consistent with the true quantized inference path and distributes the complex layer-wise calibration workload across multiple RTX3090 GPUs, thereby reducing reliance on the large memory of A100 GPUs. TLQ is evaluated across two models, three model scales, and two quantization settings, consistently achieving performance improvements across all settings, indicating its strong quantization stability. The code will be released publicly.

arXiv:2602.07899•Feb 8, 2026•2 citations
Post-Training Quantization enables efficient Vision Transformer (ViTs) deployment with a small calibration data, and its prevalent use of uniform quantization harnesses AI accelerator matrix cores for high-speed inference. However, the application of uniform quantization is fundamentally challenged by the extreme non-uniformity of activation distributions.Specifically, the power-law nature of post-Softmax attention scores and the significant inter-channel variance in post-GELU activations create a dilemma for conventional quantization, as it struggles to preserve critical high-magnitude values without sacrificing overall precision. To resolve this core conflict, we introduce UQ-ViT (Uniform Quantization for Vision Transformers), a novel uniform quantization framework designed to reconcile high precision with hardware efficiency. Central to UQ-ViT are two operators: Dynamic Elimination of Maximum (DeMax) and Normalization Quantization (NormQuant). DeMax is a quantization operator for post-Softmax attention scores that utilizes uniform quantization. It dynamically eliminates and preserves dominant values, effectively mitigating quantization loss from the extreme values in the power-law distribution. NormQuant utilizes a per-channel quantization strategy during quantization and reverts to a per-tensor format for dequantization, achieving both high accuracy and computational efficiency. Crucially, it is applicable to any linear layer, enabling effective quantization of post-GELU activations in ViTs. Through extensive experiments on various ViTs and vision tasks, including image classification, object detection, and instance segmentation, we demonstrate that our proposed approach outperforms existing methods, achieving superior accuracy while ensuring hardware friendliness.
In this study, we focus on two important branches in quantization. The first is to enable lower bit operations to take advantage of lower memory usage, increased computation speed, and reduced model size. And the second is to fully quantize to enable integer-only operations and utilize On-device utilities that enables integer-accelerated operations. In particular, the layer norm input activation required for full quantization is large in value and the presence of outliers is the main cause of performance degradation in low bit settings. In particular, since the distribution of outliers in actviations is consistent in specific channels, more efficient quantization is possible when this information is utilized in the quantization process. The GO-ViT method uses a groupwise scaling method that distinguishes inliers and outliers and determines scaling factors for each corresponding channel to improve performance even in low bits. To verify the versatile applicability of the GO-ViT methodology, which can be used in both Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT), we conducted experiments on various datasets such as ImageNet, CIFAR-100, and Something-Something v2 and several transformer architectures such as Swin, ViT, DeiT, and Video Swin. The experimental results show that GO-ViT can improve the Top-1 accuracy at low bit compared to the PTQ baseline model. In addition, it shows higher performance than the QAT baseline model tested on the Cifar100 dataset by preserving high values of activations while starting with high accuracy from the beginning.
Lung cancer remains the leading cause of cancer-related mortality worldwide, with accurate subtype classification being critical for improving patient outcomes. Vision Transformers (ViTs) have recently advanced medical image analysis, but their integration with quantum computing remains underexplored, leaving a gap in leveraging quantum advantages for clinical applications. This study proposes a quantum-enhanced Vision Transformer (QViT) that embeds variational quantum circuits (VQCs) within attention and feed-forward layers to enrich feature learning. To assess robustness, two circuit families, Basic VQC and Quantum Approximate Optimization Algorithm (QAOA) based VQC, were evaluated across two depths, yielding four QViT configurations. The models were analyzed through cancer classification tasks, systematic quantum state tomography (QST) to examine quantum state properties, and noise resilience testing under realistic Noisy Intermediate-Scale Quantum (NISQ) conditions, reflecting current hardware limitations. On 3150 CT (Computed Tomography) images, the QViT-QAOA-D1 configuration achieved 98.52% accuracy with a training time of 674 s, outperforming other configurations. Quantum analyses further showed that Basic VQC provides stronger entanglement but requires deeper, less NISQ-friendly circuits, whereas QAOA maintains high-purity states at shallower depths. These results suggest that QViT provides a practical and scalable framework, where quantum-enhanced representations can improve diagnostic performance while remaining compatible with current hardware constraints. Future work will focus on validation using patient-level datasets and deployment on real quantum hardware to further assess clinical applicability and scalability.
Reliable flood detection is critical for disaster management, yet classical deep learning models often struggle with the high-dimensional, nonlinear complexities inherent in remote sensing data. To mitigate these limitations, we introduced a novel Quantum-Enhanced Vision Transformer (ViT) that synergizes the global context-awareness of transformers with the expressive feature extraction capabilities of quantum computing. Using remote sensing imagery, we developed a hybrid architecture that processes inputs through parallel pathways, a ViT backbone and a quantum branch utilizing a 4-qubit parameterized quantum circuit for localized feature mapping. These distinct representations were fused to optimize binary classification. Results showed that the proposed hybrid model significantly outperformed a classical ViT baseline, increased overall accuracy from 84.48 % to 94.47 % and the F1-score from 0.841 to 0.944. Notably, the quantum integration substantially improved discriminative power in complex terrains for both class. These findings validate the potential of quantum-classical hybrid models to enhance precision in hydrological monitoring and earth observation applications.
At the core of the Transformer, the softmax normalizes the attention matrix to be right stochastic. Previous research has shown that this often de-stabilizes training and that enforcing the attention matrix to be doubly stochastic (through Sinkhorn's algorithm) consistently improves performance across different tasks, domains and Transformer flavors. However, Sinkhorn's algorithm is iterative, approximative, non-parametric and thus inflexible w.r.t. the obtained doubly stochastic matrix (DSM). Recently, it has been proven that DSMs can be obtained with a parametric quantum circuit, yielding a novel quantum inductive bias for DSMs with no known classical analogue. Motivated by this, we demonstrate the feasibility of a hybrid classical-quantum doubly stochastic Transformer (QDSFormer) that replaces the softmax in the self-attention layer with a variational quantum circuit. We study the expressive power of the circuit and find that it yields more diverse DSMs that better preserve information than classical operators. Across multiple small-scale object recognition tasks, we find that our QDSFormer consistently surpasses both a standard ViT and other doubly stochastic Transformers. Beyond the Sinkformer, this comparison includes a novel quantum-inspired doubly stochastic Transformer (based on QR decomposition) that can be of independent interest. Our QDSFormer also shows improved training stability and lower performance variation suggesting that it may mitigate the notoriously unstable training of ViTs on small-scale data.
Quantum machine learning (QML) holds promise for computational advantage, yet progress on real-world tasks is hindered by classical preprocessing and noisy devices. We introduce ViT-QCNN-FT, a hybrid framework that integrates a fine-tuned Vision Transformer with a quantum convolutional neural network (QCNN) to compress high-dimensional images into features suited for noisy intermediate-scale quantum (NISQ) devices. By systematically probing entanglement, we show that ansatzes with uniformly distributed entanglement entropy consistently deliver superior non-local feature fusion and state-of-the-art accuracy (99.77% on CIFAR-10). Surprisingly, quantum noise emerges as a double-edged factor: in some cases, it enhances accuracy (+2.71% under amplitude damping). Strikingly, substituting the QCNN with classical counterparts of equal parameter count leads to a dramatic 29.36% drop, providing unambiguous evidence of quantum advantage. Our study establishes a principled pathway for co-designing classical and quantum architectures, pointing toward practical QML capable of tackling complex, high-dimensional learning tasks.
Post-training quantization (PTQ) is a primary approach for deploying large language models without fine-tuning, and the quantized performance is often strongly affected by the calibration in PTQ. By contrast, in vision-language models (VLMs), substantial differences between visual and text tokens in their activation distributions and sensitivities to quantization error pose significant challenges for effective calibration during PTQ. In this work, we rethink what PTQ calibration should align with in VLMs and propose the Token-level Importance-aware Layer-wise Quantization framework (TLQ). Guided by gradient information, we design a token-level importance integration mechanism for quantization error, and use it to construct a token-level calibration set, enabling a more fine-grained calibration strategy. Furthermore, TLQ introduces a multi-GPU, quantization-exposed layer-wise calibration scheme. This scheme keeps the layer-wise calibration procedure consistent with the true quantized inference path and distributes the complex layer-wise calibration workload across multiple RTX3090 GPUs, thereby reducing reliance on the large memory of A100 GPUs. TLQ is evaluated across two models, three model scales, and two quantization settings, consistently achieving performance improvements across all settings, indicating its strong quantization stability. The code will be released publicly.