Verified80% confidenceFactTime unknown
CLIP（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
7
Sources
80%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedCLIP was released by OpenAI in 2021, simultaneously training an image encoder and a text encoder through contrastive learning to achieve zero-shot image classification.85% similarPartially VerifiedThe CLIP architecture was proposed by OpenAI in 2021 and maps images and text to the same high-dimensional vector space through contrastive learning.80% 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79% similarUnverifiedCLIP (Contrastive Language-Image Pre-training) was exposed to billions of image-text pairs during training.79% 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78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/28980API
curl https://kongchang.com/api/v1/knowledge/claims/28980MCP
get_claim(id=28980)