DeepBench uses GPT-4o to generate domain-specific corruptions and evaluates CLIP, SigLIP, and ALIGN, finding CLIP most robust overall with large variation across domains and corruption types.
In: 2017 International conference of electronics, communication and aerospace tech- nology (ICECA)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On the Domain Robustness of Contrastive Vision-Language Models
DeepBench uses GPT-4o to generate domain-specific corruptions and evaluates CLIP, SigLIP, and ALIGN, finding CLIP most robust overall with large variation across domains and corruption types.