上海人工智能实验室（上海AI实验室）近日发布了一款名为“书生·视觉大模型”（InternVL）的新一代视觉大模型。该模型由上海AI实验室联合清华大学、香港中文大学、商汤科技等机构共同研发，其视觉编码器的参数量达到了60亿（InternVL-6B）。
新一代“书生·视觉基础”模型的最大亮点是首次提出了对比-生成融合的渐进式对齐技术，实现了在互联网级别数据上视觉大模型与语言大模型的精细对齐。这一技术的提出，不仅大幅提升了模型的性能，也进一步推动了人工智能技术的发展。
据上海AI实验室介绍，InternVL模型在视觉核心任务上开源领先，其性能在多个指标上均达到了国际领先水平。这一成果的取得，标志着我国在人工智能领域的研究和技术应用又迈出了重要的一步。
英文标题Title：Shanghai AI Lab Unveils New Generation of Visual Large Model
英文关键词Keywords：AI Lab, Visual Large Model, Technological Progress
News content:
The Shanghai Artificial Intelligence Laboratory (Shanghai AI Lab) recently released a new generation of visual large model named “Scholar Visual Large Model” (InternVL). The model, developed in collaboration with Tsinghua University, the Chinese University of Hong Kong, and SenseTime, has a visual encoder parameter count of 6 billion (InternVL-6B).
The highlight of the new “Scholar Visual Baseline” model is the first proposal of the comparative-generative fusion progressive alignment technology, achieving fine alignment of visual large models and language large models on internet-scale data. This technology not only significantly improves model performance but also further promotes the development of artificial intelligence technology.
According to Shanghai AI Lab, InternVL model is leading in open-source vision core tasks, with its performance reaching the international leading level in multiple indicators. This achievement marks another important step in China’s research and application of artificial intelligence.
【来源】https://mp.weixin.qq.com/s/bdfAJRqOF9tUk8Vy9KC_XQ
Views: 1