Verified90% confidenceFactExact time
The CLIP architecture was proposed by OpenAI in 2021 and maps images and text to the same high-dimensional vector space through contrastive learning.
14
Sources
90%
Confidence
Long-term
Relevance
7/5/2026
First Seen
Sources
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80% 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.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77% similarVerifiedThe CLIP model can map images and text to the same vector space, enabling cross-modal retrieval.71% similarUnverified多模态大模型的核心技术突破在于跨模态对齐，通过CLIP等对比学习方法将图像、音频、视频映射到与文本共享的语义空间71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/112542API
curl https://kongchang.com/api/v1/knowledge/claims/112542MCP
get_claim(id=112542)