Unverified50% confidenceFactExact time
CLIP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
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedOpenAI CLIP使用4亿图文对预训练，Google ALIGN使用18亿嘈杂图文对训练79% similarVerifiedCLIP（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79% 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.77% similarVerifiedCLIP was released by OpenAI in 2021, simultaneously training an image encoder and a text encoder through contrastive learning to achieve zero-shot image classification.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/553552API
curl https://kongchang.com/api/v1/knowledge/claims/553552MCP
get_claim(id=553552)