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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

With the surge of large language models (LLMs), Large Vision-Language Models(VLMs)--which integrate vision encoders with LLMs for accurate visualgrounding--have shown great potential in tasks like generalist agents androbotic control. However, VLMs are typically trained on massive web-scrapedimages, raising concerns over copyright infringement and privacy violations,and making data auditing increasingly urgent. Membership inference (MI), whichdetermines whether a sample was used in training, has emerged as a key auditingtechnique, with promising results on open-source VLMs like LLaVA (AUC > 80%).In this work, we revisit these advances and uncover a critical issue: currentMI benchmarks suffer from distribution shifts between member and non-memberimages, introducing shortcut cues that inflate MI performance. We furtheranalyze the nature of these shifts and propose a principled metric based onoptimal transport to quantify the distribution discrepancy. To evaluate MI inrealistic settings, we construct new benchmarks with i.i.d. member andnon-member images. Existing MI methods fail under these unbiased conditions,performing only marginally better than chance. Further, we explore thetheoretical upper bound of MI by probing the Bayes Optimality within the VLM'sembedding space and find the irreducible error rate remains high. Despite thispessimistic outlook, we analyze why MI for VLMs is particularly challenging andidentify three practical scenarios--fine-tuning, access to ground-truth texts,and set-based inference--where auditing becomes feasible. Our study presents asystematic view of the limits and opportunities of MI for VLMs, providingguidance for future efforts in trustworthy data auditing.

Revisiting Data Auditing in Large Vision-Language Models