Sigmoid Loss for Language Image Pre-Training
Open accessAbstract
We propose a simple pairwise sigmoid loss for imagetext pre-training. Unlike standard contrastive learning with softmax normalization, the sigmoid loss operates solely on image-text pairs and does not require a global view of the pairwise similarities for normalization. The sigmoid loss simultaneously allows further scaling up the batch size, while also performing better at smaller batch sizes. With only four TPUv4 chips, we can train a Base CLIP model at 4k batch size and a Large LiT model at 20k batch size, the latter achieves 84.5% ImageNet zero-shot accuracy in two days. This disentanglement of the batch size from the loss further allows us to study the impact of examples vs pairs and negative to positive ratio. Finally, we push the batch size to the extreme, up to one million, and find that the benefits of growing batch size quickly diminish, with a more reasonable batch size of 32k being sufficient. We hope our research motivates further explorations in improving the quality and efficiency of language-image pre-training.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Existing foundation models for vision and language such as CLIP, SigLIP, and Llama 2 are capable of generalization stemming from Internet-scale pretraining.
“existing foundation models for vision and language such as CLIP [8], SigLIP [9], and Llama 2 [10] are capable of these types of generalization and more, stemming from the priors captured by their Internet-scale pretraining datasets.”
From OpenVLA: An Open-Source Vision-Language-Action Model (2024) · cites this paper as [9]
The OpenVLA model architecture consists of a fused vision backbone that combines the SigLIP and DinoV2 encoders.
“The OpenVLA model architecture consists of a fused vision backbone that combines the SigLIP [9] and DinoV2 [25] encoders.”
From OpenVLA: An Open-Source Vision-Language-Action Model (2024) · cites this paper as [9]
Cites
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929 · see citing works
Masked Autoencoders Are Scalable Vision Learners
He, K., Chen, X., Xie, S., Li, Y., Dollar, P., & Girshick, R. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.01553 · Cited by 8,489 · see citing works
Representation Learning with Contrastive Predictive Coding
Oord, A. V. D., Li, Y., & Vinyals, O. (2018) · arXiv · doi:10.48550/arxiv.1807.03748 · Cited by 6,242 · see citing works
Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision
Jia, C., Yang, Y., Xia, Y., Chen, Y.-T., Parekh, Z., Pham, H., Le, Q. V., Sung, Y., Li, Z., & Duerig, T. (2021) · arXiv · doi:10.48550/arxiv.2102.05918 · Cited by 1,977 · see citing works
Symbolic Discovery of Optimization Algorithms
Chen, X., Liang, C., Huang, D., Real, E., Wang, K., Liu, Y., Pham, H., Dong, X., Luong, T., Hsieh, C.-J., et al. (2023) · arXiv · doi:10.48550/arxiv.2302.06675 · Cited by 303 · see citing works
Flagged first, then by citations
Cited by
Sorted by citations
RemoteCLIP: A Vision Language Foundation Model for Remote Sensing
Liu, F., Chen, D., Guan, Z., Zhou, X., Zhu, J., Ye, Q., Fu, L., & Zhou, J. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3390838 · Cited by 507 · see citing works
Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs
Tong, S., Liu, Z., Zhai, Y., Ma, Y., LeCun, Y., & Xie, S. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.00914 · Cited by 189 · see citing works
Sellergren, A., Kazemzadeh, S., Jaroensri, T., Kiraly, A., Traverse, M., Kohlberger, T., Xu, S., Jamil, F., Hughes, C., Lau, C., et al. (2025) · arXiv · doi:10.48550/arxiv.2507.05201 · Cited by 49 · see citing works
OpenVLA: An Open-Source Vision-Language-Action Model
Kim, M. J., Pertsch, K., Karamcheti, S., Xiao, T., Balakrishna, A., Nair, S., Rafailov, R., Foster, E., Lam, G., Sanketi, P., et al. (2024) · arXiv · doi:10.48550/arxiv.2406.09246 · Cited by 62 · see citing works
“Yet beyond robotics, existing foundation models for vision and language such as CLIP [8], SigLIP [9], and Llama 2 [10] are capable of these types of generalization and more, stemming from the priors captured by their Internet-scale pretraining datasets.”
BackgroundMethodGenerating dermatopathology reports from gigapixel whole slide images with HistoGPT
Tran, M., Schmidle, P., Guo, R. R., Wagner, S. J., Koch, V., Lupperger, V., Novotny, B., Murphree, D. H., Hardway, H. D., D’Amato, M., et al. (2025) · Nature Communications · doi:10.1038/s41467-025-60014-x · Cited by 36 · see citing works
HM3D-OVON: A Dataset and Benchmark for Open-Vocabulary Object Goal Navigation
Yokoyama, N., Ramrakhya, R., Das, A., Batra, D., & Ha, S. (2024) · 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros58592.2024.10802709 · Cited by 22 · see citing works
Cosmos World Foundation Model Platform for Physical AI
NVIDIA, :, Agarwal, N., Ali, A., Bala, M., Balaji, Y., Barker, E., Cai, T., Chattopadhyay, P., Chen, Y., et al. (2025) · arXiv · doi:10.48550/arxiv.2501.03575 · Cited by 13 · see citing works
Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models
Karamcheti, S., Nair, S., Balakrishna, A., Liang, P., Kollar, T., & Sadigh, D. (2024) · arXiv · doi:10.48550/arxiv.2402.07865 · Cited by 7
SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics
Shukor, M., Aubakirova, D., Capuano, F., Kooijmans, P., Palma, S., Zouitine, A., Aractingi, M., Pascal, C., Russi, M., Marafioti, A., et al. (2025) · arXiv · doi:10.48550/arxiv.2506.01844 · Cited by 3 · see citing works
“SmolVLM-2 relies on SigLIP (Zhai et al., 2023) to encode visual features for SmolLM2 language decoder (Allal et al., 2025).”
What Matters When Building Vision Language Models for Product Image Analysis?
Trabelsi, A., Zontak, M., Qian, Y., Jackson, B., Khan, S., & Batur, U. (2025) · 2025 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw65960.2025.00151 · Cited by 3 · see citing works
STORM: Token-Efficient Long Video Understanding for Multimodal LLMs
Jiang, J., Li, X., Liu, Z., Li, M., Chen, G., Li, Z., Huang, D.-A., Liu, G., Yu, Z., Keutzer, K., et al. (2025) · 2025 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · doi:10.1109/iccvw69036.2025.00614 · Cited by 2 · see citing works
WorldVLA: Towards Autoregressive Action World Model
Cen, J., Yu, C., Yuan, H., Jiang, Y., Huang, S., Guo, J., Li, X., Song, Y., Luo, H., Wang, F., et al. (2025) · arXiv · doi:10.48550/arxiv.2506.21539 · Cited by 1 · see citing works
“Another approach integrates diffusion processes into LLMs for image generation while relying on additional visual encoders, such as CLIP (Radford et al., 2021; Zhai et al., 2023), for image understanding (Chen et al., 2025; Tong et al., 2024).”
InfiMM-WebMath-40B: Advancing Multimodal Pre-Training for Enhanced Mathematical Reasoning
Han, X., Jian, Y., Hu, X., Liu, H., Wang, Y., Fan, Q., Ai, Y., Huang, H., He, R., Yang, Z., et al. (2024) · arXiv · doi:10.48550/arxiv.2409.12568 · Cited by 1 · see citing works
Platonic Representation Hypothesis on World Models
Li, W., MA, C., Xiong, H., Chen, Y.-C., & Zhang, L. (2026) · arXiv · doi:10.48550/arxiv.2608.23720
“He et al., 2016, Zhai et al., 2023, He et al., 2022], thereby providing the models with diverse visual feature priors.”
Enhancing Vision Foundation Models via Multimodal Continual Pre-Training
Chen, Y., Meng, L., Peng, W., Tao, J., Xu, C., Wu, Z., & Jiang, Y.-G. (2025) · arXiv · doi:10.48550/arxiv.2503.18931
- Authors
- Zhai, X., Mustafa, B., Kolesnikov, A., & Beyer, L.
- Affiliations
- Google (Switzerland) (CH)
- Venue
- 2023 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2023-10-01
- Pages
- 11941-11952
- OpenAlex
- W4390873312
- Cited by
- 1,127
- References
- 5
- Citations by year
- 2026: 575 · 2025: 471 · 2024: 79 · 2023: 2
- FWCI
- 120.53
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- DeepMind
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- contrastive learning, pairwise loss, batch size scaling, CLIP, liver tumor segmentation, image-text pairs, zero-shot image classification
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
