English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Vision-language models trained on large, randomly collected data hadsignificant impact in many areas since they appeared. But as they show greatperformance in various fields, such as image-text-retrieval, their innerworkings are still not fully understood. The current work analyses the truezero-shot capabilities of those models. We start from the analysis of thetraining corpus assessing to what extent (and which of) the test classes arereally zero-shot and how this correlates with individual classes performance.We follow up with the analysis of the attribute-based zero-shot learningcapabilities of these models, evaluating how well this classical zero-shotnotion emerges from large-scale webly supervision. We leverage the recentlyreleased LAION400M data corpus as well as the publicly available pretrainedmodels of CLIP, OpenCLIP, and FLAVA, evaluating the attribute-based zero-shotcapabilities on CUB and AWA2 benchmarks. Our analysis shows that: (i) most ofthe classes in popular zero-shot benchmarks are observed (a lot) duringpre-training; (ii) zero-shot performance mainly comes out of models' capabilityof recognizing class labels, whenever they are present in the text, and asignificantly lower performing capability of attribute-based zeroshot learningis only observed when class labels are not used; (iii) the number of theattributes used can have a significant effect on performance, and can easilycause a significant performance decrease.

VL-Taboo: An Analysis of Attribute-based Zero-shot Capabilities of Vision-Language Models