Unverified50% confidenceFactExact time
LAION团队通过OpenAI的CLIP模型对图像与alt文本的语义相似度进行打分，仅保留相关性较高的图文对
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
UnverifiedMidjourney V3采用CLIP引导（CLIP-Guided）的生成方式，用OpenAI的CLIP模型评估生成图像与文本的语义匹配度76% similarUnverifiedEarly multimodal models like CLIP primarily focused on image-text matching rather than complex UI understanding75% similarVerifiedCLIP（Contrastive Language-Image Pre-training）通过对比学习将图像和文本映射到共享的语义空间74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/536149API
curl https://kongchang.com/api/v1/knowledge/claims/536149MCP
get_claim(id=536149)