Deepmind Researchers Probe Image-Language Transformers and Propose SVO Probes for Verb Understanding

Fine-tuning is required for a range of functions performed by multimodal image–language transformers. Researchers worldwide are interested in whether these algorithms can detect verbs or merely use the nouns in a phrase. A dataset of image-sentence pairs with 447 verbs that are either visible or widely encountered in the pretraining data was compiled to perform […]

The post Deepmind Researchers Probe Image-Language Transformers and Propose SVO Probes for Verb Understanding appeared first on MarkTechPost.

ai paper summary ai shorts applications artificial intelligence country deepmind editors pick featured guest post image language language model machine learning natural language generation (nlg) natural language processing natural language understanding (nlu) probe researchers tech tech news technology transformers uk unicorns