Dental Foundation Models & Benchmarks
Dental multimodal LLMs, clinical reasoning benchmarks, expert-calibrated dental evaluation, and OmniDentBench.
牙科大模型是医疗 AI 里足够独立的一条垂直线：它需要理解口腔影像、牙科专科知识、病例推理、治疗风险、专家评分和跨地区诊疗标准。DentalGPT 提供多模态牙科模型路线，GlobalDentBench 和 OmniDentBench 则提供高难度、专家校准、临床推理导向的 benchmark 入口。
Research Storyline
DentalGPT 用大规模牙科图像、专业 caption、instruction tuning 和 GRPO 强化多模态复杂推理。
GlobalDentBench 覆盖 88 个国家/地区、14 个牙科专科和多种题型，强调专家校准和风险分析。
OmniDentBench 面向复杂临床决策和生物医学研究，提供全球牙科基准评测与 leaderboard 入口。
DentalBench、OralGPT-Omni 和 OralMLLM-Bench 等工作显示牙科领域正在形成专门的模型和评测生态。
Representative Work
Builds a specialized 7B dental MLLM with domain knowledge injection and reinforcement learning for dental visual reasoning.
PaperA multinational benchmark for LLM clinical reasoning in dentistry with expert calibration, specialty taxonomy, and risk analysis.
PaperA global dental benchmarking platform for high-difficulty clinical decision-making and biomedical research evaluation.
PlatformA bilingual dental QA benchmark and DentalCorpus for evaluating and adapting LLMs in dentistry.
PaperA dental-specialized multimodal LLM and MMOral-Uni benchmark for dental image analysis.
PaperBenchmark Layers
Intraoral photos, panoramic X-rays, periapical radiographs, and dental-specific VQA expose failures that generic medical MLLMs often miss.
GlobalDentBench moves from knowledge recall to routine reasoning and individualized reasoning with patient-specific constraints.
Dentist-in-the-loop validation and rubric-based scoring make dental evaluation more clinically meaningful than generic QA accuracy.
Dental AI needs to track unsafe treatment suggestions, irreversible harm risks, and specialty-specific failure modes.
Resource Map
Global dental benchmark platform and leaderboard.
PlatformDental multimodal model repository and evaluation setup.
RepositoryExpert-calibrated multinational dental clinical reasoning benchmark.
Repository