Large Multimodal Model (LMM)

大型多模态模型（Large Multimodal Model, LMM），可同时处理和生成多种模态数据（文本、图像、音频、视频等）的大规模基础模型，例如 GPT-4V、Gemini。

English Definition

"A large-scale foundation model capable of processing and generating multiple data modalities simultaneously, such as text, images, audio, and video (e.g., GPT-4V, Gemini)."

用法说明 · 针对中文母语者

与 LLM（Large Language Model，纯文本）相区分；LMM 是 CUA 得以直接'看懂'屏幕并操作 GUI 的关键能力来源。

真实用例 · 2 条

"The recent advent of Computer-Using Agents (CUA) based on Large Multimodal Models, which can directly manipulate graphical user interfaces, offers new opportunities for such accessibility."

论文Pipeline extracted — review needed·指出 CUA 的技术基础是 LMM

"Our paper contributes by presenting one of the first empirical studies exploring how LMM-based agents can facilitate accessibility in visually intensive online shopping contexts."

论文Pipeline extracted — review needed·强调本研究是基于 LMM 的代理在视觉密集场景的早期实证

由 pipeline 自动采集，待人工 review