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

For vision-and-language reasoning tasks, both fully connectionist, end-to-endmethods and hybrid, neuro-symbolic methods have achieved high in-distributionperformance. In which out-of-distribution settings does each paradigm excel? Weinvestigate this question on both single-image and multi-image visualquestion-answering through four types of generalization tests: a novelsegment-combine test for multi-image queries, contrast set, compositionalgeneralization, and cross-benchmark transfer. Vision-and-language end-to-endtrained systems exhibit sizeable performance drops across all these tests.Neuro-symbolic methods suffer even more on cross-benchmark transfer from GQA toVQA, but they show smaller accuracy drops on the other generalization tests andtheir performance quickly improves by few-shot training. Overall, our resultsdemonstrate the complementary benefits of these two paradigms, and emphasizethe importance of using a diverse suite of generalization tests to fullycharacterize model robustness to distribution shift.

Generalization Differences between End-to-End and Neuro-Symbolic Vision-Language Reasoning Systems