Cohere Labs - Daniel Li and Hu Xu, FAIR at Meta

Contrastive Language-Image Pretraining (CLIP) is a popular foundation model, supporting from zero-shot classification, retrieval to encoders for multimodal large language models (MLLMs). Although CLIP is successfully trained on billion-scale image-text pairs from the English world, scaling CLIP's training further to learning from the worldwide web data is still challenging: (1) no curation method is available to handle data points from non-English world; (2) the English performance from existing multilingual CLIP is worse than its English-only counterpart, i.e., "curse of multilinguality" that is common in LLMs. Here, we present Meta CLIP 2, the first recipe training CLIP from scratch on worldwide web-scale image-text pairs. To generalize our findings, we conduct rigorous ablations with minimal changes that are necessary to address the above challenges and present a recipe enabling mutual benefits from English and non-English world data. In zero-shot ImageNet classification, Meta CLIP 2 ViT-H/14 surpasses its English-only counterpart by 0.8% and mSigLIP by 0.7%, and surprisingly sets new state-of-the-art without system-level confounding factors (e.g., translation, bespoke architecture changes) on multilingual benchmarks, such as CVQA with 57.4%, Babel-ImageNet with 50.2% and XM3600 with 64.3% on image-to-text retrieval.

Shang-Wen Daniel Li is a Research Lead and Manager at Fundamental AI Research (FAIR) at Meta. His research focuses on large foundation models, vision and language multimodalities, and self-supervised training. He leads multimodal foundation model pretraining at FAIR, which empowers many research and production use cases across Meta from vision encoding, segmentation to MLLMs and video generation. He serves area chairs for top-tier conferences including NeurIPS, ICML, ICLR, AAAI, ICASSP, and Interspeech. He also worked at Amazon AWS, Amazon Alexa and Apple Siri as Research Scientist before joining FAIR and earned his PhD from MIT CSAIL.

Hu Xu is a Research Scientist at Meta’s Fundamental AI Research (FAIR). His research focuses on foundational data and multi-modal pre-training. He leads many projects such as Meta CLIP, which provides the data and model foundation for Llama, DINO, Perception Encoder, JEPA, Movie Gen, and Meta’s smart glasses etc. Previously, he received his Ph.D. in Computer Science from the University of Illinois at Chicago, advised by Prof. Bing Liu and Prof. Philip S. Yu, and his M.S. from Peking University.

Add event to calendar

Apple Google Office 365 Outlook Outlook.com Yahoo