智东西快讯：今日，中文多模态大模型SuperCLUE-V基准8月榜单发布，腾讯混元大模型在国内大模型中排名第一，总分71.95，表现优异。测评涵盖国内外最具代表性的12个多模态理解大模型，腾讯混元大模型在多模态基础能力和应用能力方面均获得高分，显示出其在中文多模态理解领域的领先地位。评估标准涵盖理解准确性、回应相关性和推理深度等维度，腾讯混元大模型的表现已逼近海外顶尖模型，仅略低于GPT-4，优于CLaude3.5-Sonnet和Gemini-1.5-Pro。这一成绩标志着中国大模型在多模态理解领域的技术实力正逐步提升，与国际先进水平接近。
英语如下：
News Title: “Tencent’s Hybrid-Ernie Model Tops List: Chinese Multimodal Understanding Benchmark”
Keywords: Hybrid-Ernie, Domestic, First Place
News Content:
TechNode News: Today, the August list of the SuperCLUE-V benchmark for Chinese multimodal large models was released. Tencent’s Hybrid-Ernie model ranked first among domestic models, with a total score of 71.95, demonstrating excellent performance. The assessment covered 12 multimodal understanding large models that are most representative both domestically and internationally. The Hybrid-Ernie model from Tencent received high scores in both basic multimodal capabilities and application capabilities, indicating its leading position in the field of Chinese multimodal understanding. The evaluation criteria included dimensions such as understanding accuracy, response relevance, and reasoning depth. The performance of Tencent’s Hybrid-Ernie model has approached that of top overseas models, trailing only GPT-4 but outperforming CLaude3.5-Sonnet and Gemini-1.5-Pro. This achievement marks an increasing technical strength of Chinese large models in the field of multimodal understanding, approaching international advanced levels.
【来源】https://zhidx.com/news/42266.html
Views: 2