New benchmark and model advance explainable AI for content moderation

Researchers have developed SenBen, a new benchmark dataset for explainable content moderation in images, featuring scene graphs with detailed object attributes and sensitivity tags. They also created a compact 241M parameter model using a multi-task learning approach to improve autoregressive scene graph generation. This model demonstrates strong performance on the SenBen benchmark, outperforming most other models and commercial APIs in object detection and captioning while being significantly faster and more memory-efficient. AI

IMPACT Enhances explainability in AI content moderation, potentially improving accuracy and transparency in sensitive media analysis.

RANK_REASON The cluster describes a new academic paper introducing a benchmark dataset and a novel model for explainable content moderation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Fatih Cagatay Akyon
• Gemini
• SenBen
• Visual Genome

• paper
• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →