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

Reporting bias arises when people assume that some knowledge is universallyunderstood and hence, do not necessitate explicit elaboration. In this paper,we focus on the wide existence of reporting bias in visual-language datasets,embodied as the object-attribute association, which can subsequentially degrademodels trained on them. To mitigate this bias, we propose a bimodalaugmentation (BiAug) approach through object-attribute decoupling to flexiblysynthesize visual-language examples with a rich array of object-attributepairing and construct cross-modal hard negatives. We employ large languagemodels (LLMs) in conjunction with a grounding object detector to extract targetobjects. Subsequently, the LLM generates a detailed attribute description foreach object and produces a corresponding hard negative counterpart. Aninpainting model is then used to create images based on these detailed objectdescriptions. By doing so, the synthesized examples explicitly complementomitted objects and attributes to learn, and the hard negative pairs steer themodel to distinguish object attributes. Our experiments demonstrated that BiAugis superior in object-attribute understanding. In addition, BiAug also improvesthe performance on zero-shot retrieval tasks on general benchmarks like MSCOCOand Flickr30K. BiAug refines the way of collecting text-image datasets.Mitigating the reporting bias helps models achieve a deeper understanding ofvisual-language phenomena, expanding beyond mere frequent patterns to encompassthe richness and diversity of real-world scenarios.

Towards reporting bias in visual-language datasets: bimodal augmentation by decoupling object-attribute association