Verified90% confidenceFactTime unknown
CLIP was released by OpenAI in 2021, simultaneously training an image encoder and a text encoder through contrastive learning to achieve zero-shot image classification.
7
Sources
90%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Hugging Face Transformers：16万星开源AI框架全面解析
githubhuggingface
Related Entities
Related Claims
VerifiedCLIP（Contrastive Language-Image Pre-training）是OpenAI于2021年发布的多模态模型，通过对比学习在4亿组图文对上训练85% similarVerifiedThe CLIP architecture was proposed by OpenAI in 2021 and maps images and text to the same high-dimensional vector space through contrastive learning.79% similarUnverifiedCLIP由OpenAI与微软联合开发，通过对比学习在4亿张图文对上建立跨模态语义理解73% similarUnverified2022年发布的LayoutLMv3统一了文本和图像的预训练目标70% similarUnverifiedOpenAI CLIP使用4亿图文对预训练，Google ALIGN使用18亿嘈杂图文对训练70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/33176API
curl https://kongchang.com/api/v1/knowledge/claims/33176MCP
get_claim(id=33176)