Unverified50% confidenceBenchmarkExact time
SigLIP 2 B/16 achieves an mAP of 0.672 on image retrieval tasks, higher than CLIP ViT-B/32's 0.571.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
Unverified在细粒度汽车分类的k-NN实验中，SigLIP2 SO400M达到约92%准确率，CLIP ViT-L约59%，DINOv2 Giant约41%65% similarUnverifiedWhen DINOv3 ViT-H+/16 is pruned to 50% sparsity, removing 427 million parameters, ImageNet-1k linear probe accuracy drops by only 0.8%, without weight correction or retraining.62% similarUnverified在覆盖九大维度共403道题的测试中，Muse-Glimmer-30B综合均分为76.04，比Corpus 3.6 27B的75.57高出0.47分60% similarUnverifiedOn Multimodal RewardBench 2, JEV-27B-VL averages 63.8 across four tasks, close to GPT-4.1 and Qwen3-VL-32B but behind GPT-5 and Gemini 3 Pro; its image editing score of 57.8 is a clear weak spot.60% similarUnverified在同一细粒度汽车分类任务中，CLIP ViT-L的k-NN准确率约为59%60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1071207API
curl https://kongchang.com/api/v1/knowledge/claims/1071207MCP
get_claim(id=1071207)