ThinkingViT: Matryoshka Thinking Vision Transformer for Elastic Inference

• URL: http://arxiv.org/abs/2507.10800v1
• Date: Mon, 14 Jul 2025 20:54:41 GMT
• Title: ThinkingViT: Matryoshka Thinking Vision Transformer for Elastic Inference
• Authors: Ali Hojjat, Janek Haberer, Soren Pirk, Olaf Landsiedel,
• Abstract summary: Vision Transformers deliver state-of-the-art performance, yet their fixed budget prevents scalable deployment across heterogeneous hardware.<n>We introduce ThinkingViT, a nested ViT architecture that employs progressive thinking stages to dynamically adjust inference based on input difficulty.<n>ThinkingViT surpasses nested baselines by up to 2.0 percentage points (p.p.p.) in accuracy at the same throughput and by up to 2.9 p.p. at equal GMACs on ImageNet-1K.
• Score: 0.41942958779358674
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision Transformers deliver state-of-the-art performance, yet their fixed computational budget prevents scalable deployment across heterogeneous hardware. Recent nested Transformer architectures mitigate this by embedding nested subnetworks within a single model to enable scalable inference. However, these models allocate the same amount of compute to all inputs, regardless of their complexity, which leads to inefficiencies. To address this, we introduce ThinkingViT, a nested ViT architecture that employs progressive thinking stages to dynamically adjust inference computation based on input difficulty. ThinkingViT initiates inference by activating a small subset of the most important attention heads and terminates early if predictions reach sufficient certainty. Otherwise, it activates additional attention heads and re-evaluates the input. At the core of ThinkingViT is our Token Recycling mechanism, which conditions each subsequent inference stage on the embeddings from the previous stage, enabling progressive improvement. Due to its backbone-preserving design, ThinkingViT also serves as a plugin upgrade for vanilla ViT. Experiments show that ThinkingViT surpasses nested baselines by up to 2.0 percentage points (p.p.) in accuracy at the same throughput and by up to 2.9 p.p. at equal GMACs on ImageNet-1K. The source code is available at https://github.com/ds-kiel/ThinkingViT.

Related papers

• Denoising Vision Transformers [43.03068202384091]
  We propose a two-stage denoising approach, termed Denoising Vision Transformers (DVT) In the first stage, we separate the clean features from those contaminated by positional artifacts by enforcing cross-view feature consistency with neural fields on a per-image basis. In the second stage, we train a lightweight transformer block to predict clean features from raw ViT outputs, leveraging the derived estimates of the clean features as supervision.
  arXiv  Detail & Related papers  (2024-01-05T18:59:52Z)
• PriViT: Vision Transformers for Fast Private Inference [55.36478271911595]
  Vision Transformer (ViT) architecture has emerged as the backbone of choice for state-of-the-art deep models for computer vision applications. ViTs are ill-suited for private inference using secure multi-party protocols, due to the large number of non-polynomial operations. We propose PriViT, an algorithm to selectively " Taylorize" nonlinearities in ViTs while maintaining their prediction accuracy.
  arXiv  Detail & Related papers  (2023-10-06T21:45:05Z)
• A Close Look at Spatial Modeling: From Attention to Convolution [70.5571582194057]
  Vision Transformers have shown great promise recently for many vision tasks due to the insightful architecture design and attention mechanism. We generalize self-attention formulation to abstract a queryirrelevant global context directly and integrate the global context into convolutions. With less than 14M parameters, our FCViT-S12 outperforms related work ResT-Lite by 3.7% top1 accuracy on ImageNet-1K.
  arXiv  Detail & Related papers  (2022-12-23T19:13:43Z)
• Rethinking Hierarchicies in Pre-trained Plain Vision Transformer [76.35955924137986]
  Self-supervised pre-training vision transformer (ViT) via masked image modeling (MIM) has been proven very effective. customized algorithms should be carefully designed for the hierarchical ViTs, e.g., GreenMIM, instead of using the vanilla and simple MAE for the plain ViT. This paper proposes a novel idea of disentangling the hierarchical architecture design from the self-supervised pre-training.
  arXiv  Detail & Related papers  (2022-11-03T13:19:23Z)
• Self-Distilled Vision Transformer for Domain Generalization [58.76055100157651]
  Vision transformers (ViTs) are challenging the supremacy of CNNs on standard benchmarks. We propose a simple DG approach for ViTs, coined as self-distillation for ViTs. We empirically demonstrate notable performance gains with different DG baselines and various ViT backbones in five challenging datasets.
  arXiv  Detail & Related papers  (2022-07-25T17:57:05Z)
• I-ViT: Integer-only Quantization for Efficient Vision Transformer Inference [3.067607520161916]
  Vision Transformers (ViTs) have achieved state-of-the-art performance on various computer vision applications. These models have considerable storage and computational overheads, making their deployment and efficient inference on edge devices challenging. We propose I-ViT, an integer-only quantization scheme for ViTs, to enable ViTs to perform the entire computational graph of inference with integer arithmetic and bit-shifting.
  arXiv  Detail & Related papers  (2022-07-04T13:37:38Z)
• Auto-scaling Vision Transformers without Training [84.34662535276898]
  We propose As-ViT, an auto-scaling framework for Vision Transformers (ViTs) without training. As-ViT automatically discovers and scales up ViTs in an efficient and principled manner. As a unified framework, As-ViT achieves strong performance on classification and detection.
  arXiv  Detail & Related papers  (2022-02-24T06:30:55Z)
• So-ViT: Mind Visual Tokens for Vision Transformer [27.243241133304785]
  We propose a new classification paradigm, where the second-order, cross-covariance pooling of visual tokens is combined with class token for final classification. We develop a light-weight, hierarchical module based on off-the-shelf convolutions for visual token embedding. The results show our models, when trained from scratch, outperform the competing ViT variants, while being on par with or better than state-of-the-art CNN models.
  arXiv  Detail & Related papers  (2021-04-22T09:05:09Z)
• DeepViT: Towards Deeper Vision Transformer [92.04063170357426]
  Vision transformers (ViTs) have been successfully applied in image classification tasks recently. We show that, unlike convolution neural networks (CNNs)that can be improved by stacking more convolutional layers, the performance of ViTs saturate fast when scaled to be deeper. We propose a simple yet effective method, named Re-attention, to re-generate the attention maps to increase their diversity.
  arXiv  Detail & Related papers  (2021-03-22T14:32:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.