近日,大模型竞技场LMSYS榜单迎来了一次重要更新。国内大模型公司零一万物旗下的Yi-Large千亿参数闭源大模型,以其卓越的表现,一跃成为总榜第七,国产大模型中的佼佼者。这一成绩,不仅标志着国产大模型在技术实力上的一大突破,也与国际顶尖水平持平。
据了解,Yi-Large的成绩几乎与GPT-4-0125-preview相当。此模型由零一万物团队精心研发,以其强大的语言处理能力,赢得了广泛的认可。同时,国内另一大模型公司清华系智谱华章的GLM-4-0116也首次杀入总榜,排名第15位。
值得一提的是,LMSYS榜单此次更新对规则进行了修改,大模型在亮明身份后便不能再参与投票,有效杜绝了刷分的可能性。这一改变,使得榜单的公正性得到了进一步的保障。
在Yi-Large之前排名前六的模型中,有四个来自GPT,一个谷歌的Gemini,还有一个Anthropic的Claude。这表明,虽然国产大模型在国际竞争中取得了显著成绩,但与国际顶尖水平相比,仍有不小的差距。未来,国内大模型企业还需不断努力,推动技术进步。
此次Yi-Large在LMSYS榜单上的出色表现,充分展示了零一万物在人工智能领域的技术实力。在国内外竞争日益激烈的背景下,国产大模型的发展正迎来新的机遇与挑战。
英语如下:
Title: “Yi-Large, a Trillion-Parameter Model, Ranks Seventh Globally, Leading Domestic AI”
Keywords: LMSYS leaderboard update, Yi-Large, the zero-one-thing model, ranks first in Chinese, seventh overall. The LMSYS leaderboard, the arena of large model competition, was updated suddenly today: the zero-one-thing company’s Yi-Large, a closed-source large model with a trillion parameters, jumped to the seventh place in the overall ranking, also becoming the first domestic model on the list. Its performance is nearly on par with GPT-4-0125-preview. At the same time, the Tsinghua-affiliated domestic large model company Zhipu Hanzhang’s GLM-4-0116 also entered the overall ranking, ranking fifteenth. This result comes from the real blind test votes of over 11.7 million global users. Moreover, the Large Model Competition Arena recently revised the rules, which state that once a large model reveals its identity, it cannot continue to vote, eliminating the possibility of score padding. Looking at the top six models before Yi-Large’s ranking, four are from GPT, one is Google’s Gemini, and one is Anthropic’s Claude. Source: Quantumbit.
News Content: ### LMSYS Leaderboard Update: Zero-One-Thing’s Yi-Large Ranked Seventh Overall, Leads Domestic Models
Recently, the LMSYS leaderboard, the competition arena for large models, has undergone a significant update. The zero-one-thing company’s Yi-Large, a trillion-parameter closed-source large model, has become the seventh in the overall ranking and the outstanding domestic large model thanks to its excellent performance. This achievement not only marks a significant breakthrough in the technical strength of domestic large models but also keeps pace with international top-level performance.
It is understood that Yi-Large’s performance is almost equivalent to GPT-4-0125-preview. This model is carefully developed by the zero-one-thing team, known for its powerful language processing capabilities, and has won widespread recognition. At the same time, another domestic large model company affiliated with Tsinghua, Zhipu Hanzhang’s GLM-4-0116, has also entered the overall ranking for the first time, ranking fifteenth.
It is worth mentioning that this update to the LMSYS leaderboard has changed the rules, which state that large models cannot vote after revealing their identity, effectively preventing score padding. This change has further guaranteed the fairness of the leaderboard.
Among the top six models before Yi-Large, four are from GPT, one is Google’s Gemini, and one is Anthropic’s Claude. This indicates that although domestic large models have made significant achievements in international competition, there is still a considerable gap compared to international top-level performance. In the future, domestic large model enterprises need to continue their efforts to promote technological progress.
The impressive performance of Yi-Large on the LMSYS leaderboard fully demonstrates zero-one-thing’s technical strength in the field of artificial intelligence. In the context of increasingly fierce competition at home and abroad, the development of domestic large models is welcoming new opportunities and challenges.
【来源】https://www.qbitai.com/2024/05/145283.html
Views: 1