ArtAnno: LLM Agent-Driven System for Bidirectional Human-AI Artwork Annotation

A recent scholarly article presents ArtAnno, an annotation system for artworks that employs a multi-agent framework to enhance human-AI collaboration in both directions. This system aims to tackle the difficulty of deriving implicit meanings from art, which often necessitates a deep understanding of cultural contexts. The framework, named Bidirectional Human-AI Augmentation (BiHAA), creates a closed-loop system that fosters the development of skills and expertise via real-time exchanges between human annotators and AI. The research drew insights from a formative study involving 20 annotators from various backgrounds. ArtAnno features a Proactive Agentic Support Module, enabling AI to assist humans through semantic mining and labor. The paper, identified by arXiv as 2608.05026, is classified as a cross-type announcement.