The DOI matches its Crossref record; 106 of the 111 references it cites were found in a registry or an index.
Multiscale Vision Transformers
Open accessAbstract
We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10× more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers. Code is available at: https://github.com/facebookresearch/SlowFast.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 2×
“As in original ViT [25], we use residual connections [45] and Layer Normalization (LN) [2] in the pre-normalization configuration that applies LN at the beginning of the residual function, and our MLPs consist of two linear layers with GELU activation [48], where the first layer expands the dimension from D to 4D, and…”
MethodResultVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 6×
“In contrast to original Multi Head Attention (MHA) operators [98], where the channel dimension and the spatiotemporal resolution remains fixed, MHPA pools the sequence of latent tensors to reduce the sequence length (resolution) of the attended input.”
BackgroundMethodInfluentialVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“Incorporating downsampling, shift invariance, and shared weights, ConvNets are de-facto standard backbones for computer vision tasks for image [65, 62, 88, 90, 46, 12, 15, 34, 93, 81, 41] and video [87, 31, 11, 79, 69, 106, 96, 30, 105, 35, 29, 117, 57].”
BackgroundImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Incorporating downsampling, shift invariance, and shared weights, ConvNets are de-facto standard backbones for computer vision tasks for image [70, 67, 94, 96, 51, 15, 18, 39, 99, 87, 46] and video [93, 36, 14, 85, 74, 112, 102, 34, 111, 40, 33, 123, 62].”
BackgroundFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1506.01497 · Cited by 70,812
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021) · 2021 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00986 · Cited by 35,534 · cited 1×
“Finally, compared to the best model of concurrent Swin [79] (which was designed for image recognition), MViT has +0.”
ResultRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Research works printed but not linked: 11
- gao, 2019, Res2net: A new multi-scale backbone architecture, IEEE PAMI
- fan, 2020, Pyslowfast
- feichtenhofer, 2019, SlowFast networks for video recognition in ActivityNet challenge 2019
- 10.1109/CVPR.2009.5206848 · DOI · Paper page
- 10.1109/CVPR.2019.00033 · DOI
- hang, 2020, ResNeSt Split-Attention Networks
- 10.1109/ICCV.2019.00718 · DOI
- 10.1109/CVPR.2015.7298594 · DOI · Paper page
- carion, 2020, Endto-end object detection with transformers, Proc ECCV, 213
- loshchilov, 2018, Fixing weight decay regularization in adam
- qiu, 2017, Learning spatiotemporal representation with pseudo-3d residual networks, Proc ICCV
Cited by
Sorted by citations
Items 101–200 of 1,003
SimVPv2: Towards Simple Yet Powerful Spatiotemporal Predictive Learning
Tan, C., Gao, Z., Li, S., & Li, S. Z. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3543051 · Cited by 143
“MViT[66]introducesamultiscalepyra-midfeatureextractor,capturingbothfine-grainedandhigh-level temporalpatternsacrossvaryingresolutions.”
Flow-Guided Transformer for Video Inpainting
Zhang, K., Fu, J., & Liu, D. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19797-0_5 · Cited by 88
Hydra Attention: Efficient Attention with Many Heads
Bolya, D., Fu, C.-Y., Dai, X., Zhang, P., & Hoffman, J. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-25082-8_3 · Cited by 96
Wang, R., Chen, D., Wu, Z., Chen, Y., Dai, X., Liu, M., Yuan, L., & Jiang, Y. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00611 · Cited by 91
“Some works [4, 18, 42, 69] also adapt vision transformers to the video domain and achieve superior performance compared to previous CNN-based architectures.”
BackgroundResultDisCoVQA: Temporal Distortion-Content Transformers for Video Quality Assessment
Wu, H., Chen, C., Liao, L., Hou, J., Sun, W., Yan, Q., & Lin, W. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3249741 · Cited by 95
“Many transformer-based video backbones, e.g. ViViT [1], MViT [8] and Video Swin Transformers [25], have achieved even better performance on video recognition with pure videos than conventional CNN-based methods combined with optical flows and much better than pure 3D-CNN meth-ods.”
BackgroundResultEgoVLPv2: Egocentric Video-Language Pre-training with Fusion in the Backbone
Pramanick, S., Song, Y., Nag, S., Lin, K. Q., Shah, H., Shou, M. Z., Chellappa, R., & Zhang, P. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00487 · Cited by 106
“Notably, such fusion in the backbone strategy is not only limited to TimeSformer and RoBERTa; but can also be applied to any transformer-based video [60, 22, 2] and text [16, 79, 109] encoders.”
BackgroundReal-Time Online Video Detection with Temporal Smoothing Transformers
Zhao, Y., & Krähenbühl, P. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19830-4_28 · Cited by 96
DirecFormer: A Directed Attention in Transformer Approach to Robust Action Recognition
Truong, T.-D., Bui, Q.-H., Duong, C. N., Seo, H., Phung, S. L., Li, X., & Luu, K. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01940 · Cited by 83
Temporally Efficient Vision Transformer for Video Instance Segmentation
Yang, S., Wang, X., Li, Y., Fang, Y., Fang, J., Liu, W., Zhao, X., & Shan, Y. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00290 · Cited by 83
“…[51], has shown strong long-range context modeling ability and obtained great successes on im* age classification [14, 16, 22, 32, 47, 50, 54], object detection [7, 34, 44, 68], semantic segmentation [12, 43, 59], instance segmentation [13, 18, 66], and video recognition [1, 4,15,33,37,65,67].”
BackgroundMethodSelf-citationJoint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers
Wei, S., Ye, T., Zhang, S., Tang, Y., & Liang, J. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00208 · Cited by 95
“Since the transformer [30] was proved efficacious in NLP tasks, numerous studies have explored methods to acclimate the transformer architecture to computer vision tasks [1,3,5–7, Vanilla ViTs.”
BackgroundMethodInfluentialLin, F., Crawford, S., Guillot, K., Zhang, Y., Chen, Y., Yuan, X., Chen, L., Williams, S., Minvielle, R., Xiao, X., et al. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00531 · Cited by 85
“Several subsequent methods have been developed, including DeiT [46] for data-efficient ViT through knowledge distillation, Swin [33] for computation-efficient ViT using shifted windows, PVT [51] for dense prediction tasks, and MAE [18] for self-supervised learning, among others [7, 21, 14, 55, 3, 52, 31, 45].”
MethodViTMatte: Boosting image matting with pre-trained plain vision transformers
Yao, J., Wang, X., Yang, S., & Wang, B. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.102091 · Cited by 96
UniFormerV2: Unlocking the Potential of Image ViTs for Video Understanding
Li, K., Wang, Y., He, Y., Li, Y., Wang, Y., Wang, L., & Qiao, Y. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00157 · Cited by 91
“However, like other specialized video backbones [19, 37, 42], UniFormer is difficult to scale up due to the necessity of costly image pretraining.”
BackgroundMethodSelf-citationObject-Region Video Transformers
Herzig, R., Avraham, E. B., Mangalam, K., Bar, A., Chechik, G., Rohrbach, A., Darrell, T., & Globerson, A. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00315 · Cited by 84
Self-citationStand-Alone Inter-Frame Attention in Video Models
Long, F., Qiu, Z., Pan, Y., Yao, T., Luo, J., & Mei, T. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00319 · Cited by 78
“Particularly, we follow the one-clip and 3-crop settings [2,8,34] for testing on Something-Something.”
MethodOmniMAE: Single Model Masked Pretraining on Images and Videos
Girdhar, R., El-Nouby, A., Singh, M., Alwala, K. V., Joulin, A., & Misra, I. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01003 · Cited by 101
Transformer-Based Multi-Scale Feature Integration Network for Video Saliency Prediction
Zhou, X., Wu, S., Shi, R., Zheng, B., Wang, S., Yin, H., Zhang, J., & Yan, C. C. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3278410 · Cited by 76
LoViT: Long Video Transformer for surgical phase recognition
Liu, Y., Boels, M., García-Peraza-Herrera, L. C., Vercauteren, T. K., Prokar Dasgupta, A., Granados, A., & Sebastien Ourselin, A. (2024) · Medical Image Analysis · doi:10.1016/j.media.2024.103366 · Cited by 99
Efficient Video Transformers with Spatial-Temporal Token Selection
Wang, J., Yang, X., Li, H., Liu, L., Wu, Z., & Jiang, Y. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19833-5_5 · Cited by 83
Rethinking Transformers Pre-training for Multi-Spectral Satellite Imagery
Noman, M., Naseer, M., Cholakkal, H., Anwer, R. M., Khan, S., & Khan, F. S. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.02627 · Cited by 94
“Convolutional neural networks and transformers [1, 7, 9, 10, 14, 15, 19, 20, 23] have employed the multi-scale information and show promising results on classification, detection, and segmentation tasks.”
BackgroundCoarse-to-Fine Feature Mining for Video Semantic Segmentation
Sun, G., Liu, Y., Ding, H., Probst, T., & Gool, L. V. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00313 · Cited by 72
“Last but not least, although self-attention can technically model global relationships, a too long sequence length would limit its performance, as demonstrated in [12, 21, 29, 56, 79, 84] where downsampling features into small scales leads to better performance than the original long sequence length.”
BackgroundRethinking Video ViTs: Sparse Video Tubes for Joint Image and Video Learning
Piergiovanni, A., Kuo, W., & Angelova, A. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00220 · Cited by 72
“However, adaptations to video are both challenging and computationally intensive, so video versions have been been specially designed to handle the larger number of frames, for example, ViViT [3], MultiView [62], TimeSFormer [6] and others [13].”
BackgroundEfficient Emotional Adaptation for Audio-Driven Talking-Head Generation
Gan, Y., Yang, Z., Yue, X., Sun, L., & Yang, Y. (2023) · 2023 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.02069 · Cited by 72
“Following the groundbreaking work of transformers [45] in NLP [12, 39], recent works have achieved remarkable progress in video-related tasks, including action recognition [2, 15, 30], video segmentation [50, 51, 58, 34], cross-model understanding [26, 6], etc.”
BackgroundMultiscale Deep Learning for Detection and Recognition: A Comprehensive Survey
Jiao, L., Wang, M., Liu, X., Li, L., Liu, F., Feng, Z., Yang, S., & Hou, B. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3389454 · Cited by 83
“Combining the idea of multiscale feature extraction in CNN structure with Transformer, Fan et al. [160] proposed multiscale ViTs (MViT) for video and image recognition.”
Token Fusion: Bridging the Gap between Token Pruning and Token Merging
Kim, M., Gao, S., Hsu, Y.-C., Shen, Y., & Jin, H. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00141 · Cited by 103
PCAT-UNet: UNet-like network fused convolution and transformer for retinal vessel segmentation
Chen, D., Yang, W., Wang, L., Tan, S., Lin, J., & Bu, W. (2022) · PLoS ONE · doi:10.1371/journal.pone.0262689 · Cited by 70
Student behavior recognition for interaction detection in the classroom environment
Li, Y., Qi, X., Saudagar, A. K. J., Badshah, A. M., Muhammad, K., & Liu, S. (2023) · Image and Vision Computing · doi:10.1016/j.imavis.2023.104726 · Cited by 79
Self-citationFedDiff: Diffusion Model Driven Federated Learning for Multi-Modal and Multi-Clients
Li, D., Xie, W., Wang, Z., Lü, Y., Li, Y., & Fang, L. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3407131 · Cited by 84
“The specific procedure is as follows: Firstly, we construct the transformation matrix ( W Q , W K , W V ) to decompose the input intermediate feature (cid:101) x ds into three distinct tensors: Q , K , and V [43].”
MethodSelf-citationLearning Multi-Granularity Temporal Characteristics for Face Anti-Spoofing
Wang, Z., Wang, Q., Deng, W., & Guo, G. (2022) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2022.3158062 · Cited by 73
Adaptive Mixture of Experts Learning for Generalizable Face Anti-Spoofing
Zhou, Q., Zhang, K.-Y., Yao, T., Yi, R., Ding, S., & Ma, L. (2022) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3503161.3547769 · Cited by 69
“Compared to the attention mechanism [12, 14, 39] in Transformers that needs to interact token embeddings with each other, DEA instead queries the common features and only interacts with the feature of the domain experts, which largely simplifies”
BackgroundEfficient Hierarchical Entropy Model for Learned Point Cloud Compression
Song, R. Y., Fu, C., Liu, S., & Li, G. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01381 · Cited by 70
“It is inspired by vision transformers designed for 2D images [8, 22, 23, 45].”
BackgroundVision-Language Models Can Identify Distracted Driver Behavior From Naturalistic Videos
Hasan, M. Z., Chen, J., Wang, J., Rahman, M. S., Joshi, A., Velipasalar, S., Hegde, C., Sharma, A., & Sarkar, S. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3381175 · Cited by 71
“It is proficient in most visual understanding tasks, since it was trained on a large variety (around 400 million) of image-text pairs from the web.”
BackgroundPolyp-Mixer: An Efficient Context-Aware MLP-Based Paradigm for Polyp Segmentation
Shi, J.-H., Zhang, Q., Tang, Y., & Zhang, Z.-Q. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3197643 · Cited by 67
“Given an input image with the size of H × W × 3, we first split it into patches by overlapping patch embedding [38].”
MethodMasked Image Modeling with Local Multi-Scale Reconstruction
Wang, H., Tang, Y., Wang, Y., Guo, J., Deng, Z., & Han, K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00211 · Cited by 73
“To this end, many advanced vision transformers [17, 33, 43, 45, 51] hard-code the multi-scale property to the architectures and boost the performance.”
BackgroundVision Transformers with Hierarchical Attention
Liu, Y., Wu, Y.-H., Sun, G., Zhang, L., Chhatkuli, A., & Gool, L. V. (2024) · Machine Intelligence Research · doi:10.1007/s11633-024-1393-8 · Cited by 100
“PVT [18] and MViT [20] downsample the feature to compute attention in a reduced length at the cost of losing contextual details.”
BackgroundMethodInfluentialFuture Transformer for Long-term Action Anticipation
Gong, D., Lee, J., Kim, M., Ha, S. J., & Cho, M. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00306 · Cited by 67
GateHUB: Gated History Unit with Background Suppression for Online Action Detection
Chen, J., Mittal, G. K., Yu, Y., Kong, Y., & Chen, M. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01930 · Cited by 72
“Most of the previous transformer-based methods [1, 3, 17, 34] focus on action recognition in trimmed videos [7] (videos spanning few seconds) due to the quadratic complexity w.r.t. video length.”
BackgroundPale Transformer: A General Vision Transformer Backbone with Pale-Shaped Attention
Wu, S., Wu, T., Tan, H., & Guo, G. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i3.20176 · Cited by 73
Rethinking attention mechanism in time series classification
Zhao, B., Xing, H., Wang, X., Song, F., & Xiao, Z. (2023) · Information Sciences · doi:10.1016/j.ins.2023.01.093 · Cited by 65
Transformer technology in molecular science
Jian, J., Ke, L., Chen, L., Dou, B., Zhu, Y., Liu, J., Zhang, B., Zhou, T., & Wei, G. (2024) · Wiley Interdisciplinary Reviews Computational Molecular Science · doi:10.1002/wcms.1725 · Cited by 73
“This article is categorized under: Data Science > Chemoinformatics Data Science > Artificial Intelligence/Machine Learning biology, chemistry, machine learning, molecular science, transformer technology”
Yan, S., Yang, B., Chen, A., Zhao, X., & Zhang, S. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103019 · Cited by 65
Token Merging: Your ViT But Faster
Bolya, D., Fu, C.-Y., Dai, X., Zhang, P., Feichtenhofer, C., & Hoffman, J. (2022) · arXiv · doi:10.48550/arxiv.2210.09461 · Cited by 140
“5, we show the results of applying our method off-the-shelf and during MAE fine-tuning using ViT-L from Spatiotemporal MAE compared to the relevant state-of-the-art on Kinetics-400 classification: Swin (Liu et al., 2022c) pretrained on ImageNet-21k, MViTv2 (Li et al., 2022) pretrained with MaskFeats (Wei et al., 2022)…”
MethodInfluentialSelf-citationDTCM: Joint Optimization of Dark Enhancement and Action Recognition in Videos
Tu, Z., Liu, Y., Zhang, Y., Mu, Q., & Yuan, J. (2023) · IEEE Transactions on Image Processing · doi:10.1109/tip.2023.3286254 · Cited by 64
“92% Top-1 accuracy, and MViT-B [77] gets 40.”
MethodInfluentialTransformer Meets Remote Sensing Video Detection and Tracking: A Comprehensive Survey
Jiao, L., Zhang, X., Liu, X., Liu, F., Yang, S., Ma, W., Li, L., Chen, P., Feng, Z., Guo, Y., et al. (2023) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2023.3289293 · Cited by 68
“In addition, the development of advanced transformers has injected some inspiration into the RS field [23], [24], [25], [26], [27], [28], [29].”
BackgroundXie, Y., Niu, J., Zhang, Y., & Ren, F. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3186707 · Cited by 65
“…region-based graph can be regarded as an image or heat map, where each regularly arranged node is a pixel, so the well-studied CNN structures, such as weight sharing, patching [11], feature pyramid [12], and Encoder-Decoder U-Net [17], can capture spatial features efficiently and effectively.”
BackgroundMemory-and-Anticipation Transformer for Online Action Understanding
Wang, J., Chen, G., Huang, Y., Wang, L., & Lu, T. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01271 · Cited by 62
“MViT [13] builds a feature hierarchy by progressively expanding channel capacity and reducing video spatial-temporal resolution.”
BackgroundEAN: Event Adaptive Network for Enhanced Action Recognition
Tian, Y., Yan, Y., Zhai, G., Guo, G., & Gao, Z. (2022) · International Journal of Computer Vision · doi:10.1007/s11263-022-01661-1 · Cited by 64
“To model the global information, dense interaction models [24][25][26] calculate the paired correlations at all positions, which inevitably introduce the background noise signals.”
MethodTariq, A., Iqbal, M. M., Iqbal, M. J., & Ahmad, I. (2025) · IEEE Access · doi:10.1109/access.2025.3555638 · Cited by 61
“Vision transformers allow for effective feature extraction, facilitating the identification of crucial object features within these complex images [4].”
Two-stream spatio-temporal GCN-transformer networks for skeleton-based action recognition
Chen, D., Chen, M., Wu, P., Wu, M., Zhang, T., & Li, C. (2025) · Scientific Reports · doi:10.1038/s41598-025-87752-8 · Cited by 72
CNNs, RNNs and Transformers in human action recognition: a survey and a hybrid model
Alomar, K., Aysel, H. I., & Cai, X. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11388-3 · Cited by 104
“The Hybrid CNN-ViT model showcased the highest and continuously improving accuracy rates across all frame lengths: 94.12% for 12 frames, 94.56% for 18 frames, and an impressive 95.78% for 24 frames.”
InfluentialHuman‐centered attention‐aware networks for action recognition
Liu, S., Li, Y., & Fu, W. (2022) · International Journal of Intelligent Systems · doi:10.1002/int.23029 · Cited by 59
Self-citationYou, H., Xiong, Y., Dai, X., Wu, B., Zhang, P., Fan, H., Vajda, P., & Lin, Y. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01387 · Cited by 67
“To further improve ViTs’ achievable accuracy-efficiency trade-offs, CrossViT [9], PiT [26], PVT [59], MViT [21] and Swin-Transformer Swin-Transformer [ 38 ] Efficient ViT Variants.”
Self-citationViTs as backbones: Leveraging vision transformers for feature extraction
Elharrouss, O., Himeur, Y., Mahmood, Y., Alrabaee, S., Ouamane, A., Bensaali, F., Bechqito, Y., & Chouchane, A. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.102951 · Cited by 69
Towards Holistic Surgical Scene Understanding
Valderrama, N., Puentes, P. R., Hernández, I., Ayobi, N., Verlyck, M., Santander, J., Caicedo, J. C., Fernández, N., & Arbeláez, P. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-16449-1_42 · Cited by 58
“MViT [9] is a transformer-based architecture for video and image recognition that leverages multiscale feature hierarchies.”
BackgroundMethodInfluentialReversible Vision Transformers
Mangalam, K., Fan, H., Li, Y., Wu, C.-Y., Xiong, B., Feichtenhofer, C., & Malik, J. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01056 · Cited by 62
“The stage-transition block closely follows the design of the resolution upsampling blocks in MViT [18] with the following crucial modifications:”
BackgroundMethodInfluentialSelf-citationPaçal, İ., & Işık, G. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10769-z · Cited by 58
“The Multi-scale Vision Transformers are introduced as a transformer structure for visual data modeling, including images and videos [45].”
S2WAT: Image Style Transfer via Hierarchical Vision Transformer Using Strips Window Attention
Zhang, C., Xu, X., Wang, L., Dai, Z., & Yang, J. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i7.28529 · Cited by 57
Machine Learning for Human Activity Recognition: State-of-the-Art Techniques and Emerging Trends
Hossen, M. A., & Abas, P. E. (2025) · Journal of Imaging · doi:10.3390/jimaging11030091 · Cited by 70
MAR: M asked Autoencoders for Efficient A ction R ecognition
Qing, Z., Zhang, S., Huang, Z., Wang, X., Wang, Y., Lv, Y., Gao, C., & Sang, N. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3263288 · Cited by 76
“MViT [5], [47] studies hierarchical Trans-formers with several channel-resolution scale stages, and proposes pooling attention to reduce computation.”
BackgroundDual-Path Adaptation from Image to Video Transformers
Park, J., Lee, J., & Sohn, K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00219 · Cited by 64
End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames
Liu, S., Zhang, C.-L., Zhao, C., & Ghanem, B. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01759 · Cited by 71
“Analogously, in the realms of image and video understanding, architectures such as ViT [14] and MViT [16], have also witnessed the effectiveness of this scaling strategy.”
BackgroundThe effectiveness of MAE pre-pretraining for billion-scale pretraining
Singh, M., Duval, Q., Alwala, K. V., Fan, H., Aggarwal, V., Adcock, A., Joulin, A., Dollár, P., Feichtenhofer, C., Girshick, R., et al. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00505 · Cited by 70
“For all video finetuning, we use relative position embeddings [24], which improves our performance by 0.”
BackgroundMethodSelf-citationSKiT: a Fast Key Information Video Transformer for Online Surgical Phase Recognition
Liu, Y., Huo, J., Peng, J., Sparks, R. E., Prokar Dasgupta, A., Granados, A., & Sebastien Ourselin, A. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01927 · Cited by 62
Unleashing Vanilla Vision Transformer with Masked Image Modeling for Object Detection
Fang, Y. K., Yang, S., Wang, S., Ge, Y., Shan, Y., & Wang, X. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00574 · Cited by 64
“The multi-scale inputs naturally exist in hierarchical ConvNet [27, 47, 23, 60, 48] / ViT [35, 54, 17, 30, 62] backbones.”
BackgroundGroup Contextualization for Video Recognition
Hao, Y., Zhang, H., Ngo, C., & He, X. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00100 · Cited by 54
Towards Scalable Neural Representation for Diverse Videos
He, B., Yang, X., Wang, H., Wu, Z., Chen, H., Huang, S., Ren, Y., Lim, S.-N., & Shrivastava, A. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00594 · Cited by 54
Diffusion Models as Masked Autoencoders
Wei, C., Mangalam, K., Huang, P.-Y., Li, Y., Fan, H., Hu, X., Wang, H., Xie, C., Yuille, A., & Feichtenhofer, C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01492 · Cited by 80
Self-citationHow Much Temporal Long-Term Context is Needed for Action Segmentation?
Bahrami, E., Francesca, G., & Gall, J. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00950 · Cited by 57
“MViT [13] uses a pooling operation to reduce the space-time resolution before attention and MViTv2 [31] improves MViT by adding the relative position along (x,y,t) and residual connections.”
MethodAncilia: Scalable Intelligent Video Surveillance for the Artificial Intelligence of Things
Pazho, A. D., Neff, C., Noghre, G. A., Ardabili, B. R., Yao, S., Baharani, M., & Tabkhi, H. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3263725 · Cited by 56
“There are many works that use pixel-based methods for these tasks that achieve superior accuracies than SotA posed-based methods, such as I3D [54], MVIT [55], and Stargazer [56].”
MethodFan, H., Murrell, T., Wang, H., Alwala, K. V., Li, Y., Li, Y., Xiong, B., Ravi, N., Li, M., Yang, H., et al. (2021) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3474085.3478329 · Cited by 50
“Top-1 Top-5 FLOPs (G) Param (M) I3D R50, 8 × 8 [25] 73.3 90.7 37.5 x 3 x 10 28.0 X3D-L, 16 × 5 [8] 77.4 93.3 26.6 x 3 x 10 6.2 SlowFast R101, 16 × 8 [9] 78.7 93.6 215.6 x 3 x 10 53.8 MViT-B, 16 × 4 [7] 78.8 93.9 70.5 × 3 × 10 36.6 Table 1: Video classification results on Kinetics-400 [18].”
InfluentialSelf-citationAudio-Visual Efficient Conformer for Robust Speech Recognition
Burchi, M., & Timofte, R. (2023) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00229 · Cited by 54
“Similar to the pooling attention proposed by the Multiscale Vision Transformer (MViT) [13] for video and image recognition, the patch attention proceed to an average”
MethodQiao, X., Roy, S. K., & Huang, W. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3314550 · Cited by 51
“Fan et al. [43] introduced multiscale vision transformers (MViTs) for image recognition.”
BackgroundDCS-TransUperNet: Road Segmentation Network Based on CSwin Transformer with Dual Resolution
Zhang, Z., Miao, C., Liu, C., & Tian, Q. (2022) · Applied Sciences · doi:10.3390/app12073511 · Cited by 44
“Moreover, by connecting the multi-scale feature with the Transformer model, Haoqi Fan et al. proposed a multi-scale visual Transformer (MViT) [20] for video and image recognition.”
MethodLearning Implicit Class Knowledge for RGB-D Co-Salient Object Detection With Transformers
Zhang, N., Han, J., & Liu, N. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3185550 · Cited by 51
Alfasly, S., Chui, C. K., Jiang, Q., Lu, J. J., & Chen, X. (2022) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2022.3190367 · Cited by 51
On the Benefits of 3D Pose and Tracking for Human Action Recognition
Rajasegaran, J., Pavlakos, G., Kanazawa, A., Feichtenhofer, C., & Malik, J. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00069 · Cited by 57
“We run MViT [14] (pretrained on MaskFeat [64]) at 1Hz on every track in Kinetics-400 to generate pseudo groundtruth annotations.”
BackgroundMethodInfluentialSelf-citationMultiscale Facial Expression Recognition Based on Dynamic Global and Static Local Attention
Xu, J., Li, Y., Yang, G., He, L., & Luo, K. (2024) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2024.3458464 · Cited by 52
“The multiscale approach has been noted to impact the representation of low-frequency and high-frequency information in deep features [26].”
InfluentialSelf-citationVideo Mamba Suite: State Space Model as a Versatile Alternative for Video Understanding
Chen, G., Huang, Y., Xu, J., Pei, B., Wang, J., Chen, Z., Li, Z., Lu, T., & Wang, L. (2025) · International Journal of Computer Vision · doi:10.1007/s11263-025-02597-y · Cited by 154
“…model for extracting temporal features in time-dependent tasks; 2) Video temporal module as a complementary component to image encoders; 3) Multimodal interaction network for vision-text information fusion; and 4) Spatiotempo-ral model for direct spatiotemporal feature modeling from RGB frames.”
CSiT: A Multiscale Vision Transformer for Hyperspectral Image Classification
He, W., Huang, W., Liao, S., Xu, Z., & Yan, J. (2022) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2022.3216335 · Cited by 50
“Fan et al. [34] collected multiscale level features and vision transformer architecture, and it achieved potential results on video and image recognition tasks.”
BackgroundKhan, S. A., & Dang-Nguyen, D.-T. (2024) · IEEE Access · doi:10.1109/access.2023.3348450 · Cited by 53
“…models is rooted in their recent notable achievements across a spectrum of computer vision tasks such as image classification [14], [15], [16], object detection [17], [18], image segmentation [19], video classification [20], multi-modal learning [21], [22], 3D analysis [23], [24] and…”
BackgroundInfluentialLeo, M., Bernava, G. M., Carcagnì, P., & Distante, C. (2022) · Sensors · doi:10.3390/s22030866 · Cited by 52
“It connects the seminal idea of multiscale feature hierarchies with transformer models [98].”
BackgroundVARS: Video Assistant Referee System for Automated Soccer Decision Making from Multiple Views
Held, J., Cioppa, A., Giancola, S., Hamdi, A., Ghanem, B., & Droogenbroeck, M. V. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00537 · Cited by 57
“Recently, the Multiscale Vision Transformer (MViT) [15, 37] came as a way to combine the strengths of both convolutional neural networks (CNNs) and transformers for video classification, capturing both spatial and temporal attentions.”
BackgroundMethodInfluentialLiu, Z., Wang, S., Gu, X., Wang, D., Dong, Q., & Cui, B. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3403144 · Cited by 48
“Its core gradually expands the channel dimension while reducing the spatial dimension (the sequence length) [54].”
BackgroundUllah, W., Javed, K., Khan, M. A., Alghayadh, F. Y., Bhatt, M. W., Naimi, I. S. A., & Ofori, I. (2024) · Discover Sustainability · doi:10.1007/s43621-024-00307-1 · Cited by 48
Spatiotemporal Self-attention Modeling with Temporal Patch Shift for Action Recognition
Xiang, W., Li, C., Wang, B., Wei, X., Hua, X., & Lei Zhang, A. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20062-5_36 · Cited by 47
Boosting Few-shot Action Recognition with Graph-guided Hybrid Matching
Xing, J., Wang, M., Ruan, Y., Chen, B., Guo, Y., Mu, B., Dai, G., Wang, J., & Liu, Y. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00167 · Cited by 47
“Inspired by some temporal modeling methods based on attention mechanism [1, 43, 9, 37, 2], we design a learnable dense temporal modeling module, which consists of a temporal patch relation modeling block and a temporal channel relation modeling block, as shown in Fig.”
MethodRTUNet: Residual transformer UNet specifically for pancreas segmentation
Qiu, C., Liu, Z., Song, Y., Yin, J., Han, K., Zhu, Y., Liu, Y., & Sheng, V. S. (2022) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2022.104173 · Cited by 46
Spotting Temporally Precise, Fine-Grained Events in Video
Hong, J., Zhang, H., Gharbi, M., Fisher, M., & Fatahalian, K. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19833-5_3 · Cited by 51
AdaMAE: Adaptive Masking for Efficient Spatiotemporal Learning with Masked Autoencoders
Bandara, W. G. C., Patel, N., Gholami, A., Nikkhah, M., Agrawal, M., & Patel, V. M. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01394 · Cited by 53
“6 MViTv1 [9] MViTv1-B N/A 64 455×1×3 37 67.”
InfluentialLong, S., Zhao, Z. W., Pi, J., Wang, S., & Wang, J. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00996 · Cited by 53
“Though ViT has achieved competitive accuracy in vision tasks [1, 5, 9, 12, 19, 31, 34, 41], it needs huge memory and computational resources.”
BackgroundAttenGait: Gait recognition with attention and rich modalities
Castro, F. M., Delgado-Escaño, R., Hernández-García, R., Marín-Jiménez, M. J., & Guil, N. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110171 · Cited by 46
Video-FocalNets: Spatio-Temporal Focal Modulation for Video Action Recognition
Wasim, S. T., Khattak, M. U., Naseer, M., Khan, S. H., Shah, M. A., & Khan, F. S. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01267 · Cited by 47
Gao, Z. Y., Yang, J., Jiang, F., Jiao, X., Dashtipour, K. K., Gogate, M., & Hussain, A. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.111977 · Cited by 50
V2X-ViTv2: Improved Vision Transformers for Vehicle-to-Everything Cooperative Perception
Xu, R., Chen, C.-J., Tu, Z., & Yang, M. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3479222 · Cited by 45
“Ourmodel is motivated by a recent work [27], [28] in which attention is computed on high-resolution features is addressed by initially using spatial pooling on the feature map via strid-den convolution, scaled down by a factor N .”
Video prediction by efficient transformers
Ye, X., & Bilodeau, G.-A. (2022) · Image and Vision Computing · doi:10.1016/j.imavis.2022.104612 · Cited by 51
“Pooling can be an alternative to sequence length reduction [43].”
BackgroundWang, J., & Torresani, L. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01366 · Cited by 50
“While most video transformers maintain a constant space-time resolution S × T and fixed feature dimension D across all layers, some works have proposed patchpooling [10] and patch-merging [29], which progressively reduce the space-time sequence length.”
BackgroundMethodResultInfluentialChen, S., Ren, S., Wang, G., Huang, M., & Xue, C. (2023) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2023.3247949 · Cited by 49
“Fan et al. [46] proposed a multiscale ViT structure with multiscale feature pyramids.”
BackgroundDriver Distraction Behavior Recognition for Autonomous Driving: Approaches, Datasets and Challenges
Tan, D., Tian, W., Wang, C., Chen, L., & Xiong, L. (2024) · IEEE Transactions on Intelligent Vehicles · doi:10.1109/tiv.2024.3405990 · Cited by 48
“Inordertoextract features of different scales, Liang et al. [69] presented an action detection system that incorporates enhanced multi-scale vision transformers (MViT v2 [70], [71]) for the purpose of extracting hierarchical features, referred to as Stargazer.”
Ghosh, S., Dhar, S., Yoddha, R., Kumar, S., Thakur, A. K., & Jana, N. D. (2024) · Procedia Computer Science · doi:10.1016/j.procs.2024.04.284 · Cited by 44
WiMANS: A Benchmark Dataset for WiFi-Based Multi-user Activity Sensing
Huang, S., Li, K., You, D., Chen, Y., Lin, A., Liu, S., Li, X., & McCann, J. A. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72946-1_5 · Cited by 44
Metadata
- Authors
- Fan, H., Xiong, B., Mangalam, K., Li, Y., Yan, Z., Malik, J., & Feichtenhofer, C.
- Affiliations
- Meta (Israel) (IL)
- Meta (United States) (US)
- Venue
- 2021 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2021-10
- Pages
- 6804-6815
- OpenAlex
- W3202698164
- Cited by
- 1,407
- References
- 111
- Citations by year
- 2026: 147 · 2025: 345 · 2024: 380 · 2023: 349 · 2022: 176 · 2021: 10
- FWCI
- 74.81
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University)
- TIB Data Manager
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Human Pose and Action Recognition
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- video recognition, image classification, feature pyramid, spatial resolution reduction, computational efficiency
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
