DentAgent multi-agent AI beats dental specialists by 17.3% in diagnosis

A new research paper introduces DentAgent, an evidence-centric multi-agent framework designed to tackle multimodal dental reasoning. Developed by Zijie Meng and colleagues from Zhejiang University and Ant Group, DentAgent addresses a key gap in dental AI: most existing systems are modality- or task-specific. While vision-language models can answer dental questions flexibly, their responses often lack traceable evidence. DentAgent solves this by coordinating five specialized agents—each handling a different modality such as domain knowledge, radiographs, intraoral photographs, and 3D dental data—under an Orchestrator.

Each specialist uses domain-specific tools to convert raw observations into structured evidence records. These are managed in an Evidence Blackboard, which acts as a shared state tracking coverage, gaps, and conflicts before final response generation. This standardized evidence representation unifies isolated dental capabilities into a single agentic workflow. The paper reports that DentAgent achieved leading performance across four benchmarks, surpassing senior specialists by 17.3 percentage points on multi-label diagnosis. The framework's traceable evidence and strong accuracy highlight its potential as a technical foundation for large-scale population oral health assessment and management, making AI-driven dental care more transparent and scalable.