Event ·Multi-source

实测双曲嵌入对图像搜索排序的影响:Hyper3-CLIP V1 对比 OpenAI CLIP

Milvus 团队用 5 万张 ImageNet 验证图测试 Hyper3-CLIP V1:Lorentz 距离比 cosine 将分类准确率从 50.14% 提到 51.06%,但在 132 个大类搜索上落后 cosine 2.2 平均精度点。与 OpenAI CLIP 相比,V1 在描述性标题检索和区分“狗在沙发上/沙发在狗上”更准,CLIP 则在大类搜索领先 6.4 平均精度点,差距主要来自稀有分类词。较小的 COCO 测试覆盖三个概念时 V1 反超。结论是描述性标题搜索可试用 V1,大类搜索建议在自己的数据上对比 Lorentz 与 cosine。

Current read

Milvus 拿 5 万张图实测了双曲嵌入和 CLIP 的检索差距,写描述找图它更强,搜大类还是 cosine 稳,动手前可以看看。

6 sources39° AI heatLast updated 2026/11/2 15:05:00

Evidence

6 sources
Related source 1Simon Willison’s Weblog
View source
Related source 2arXiv: OpenAI
View source
Related source 4The Information
View source
Related source 5shao__meng
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.