中国人工智能公司百川智能近日发布了一款超千亿参数的大语言模型Baichuan 3，该模型在多个权威通用能力评测中表现出色，尤其在中文任务上更是超越了GPT-4。
Baichuan 3在CMMLU、GAOKAO和AGI-Eval等评测中展现了出色的能力。在数学和代码专项评测如MATH、HumanEval和MBPP中同样表现出色，证明了Baichuan 3在自然语言处理和代码生成领域的强大实力。
据百川大模型公众号消息，Baichuan 3的发布标志着中国在人工智能领域取得了重要突破，其强大的语言处理能力将为各行各业带来更多创新应用。
英文标题：Baichuan 3: A Trillion-Parameter Language Model by Baichuan Intelligence Surpasses GPT-4 in Chinese Evaluation
关键词：Baichuan 3, Language model, GPT-4
新闻内容：
Chinese artificial intelligence company Baichuan Intelligence has recently released the Baichuan 3, a trillion-parameter large language model, which has demonstrated exceptional performance in several authoritative general-purpose capability evaluations, particularly surpassing GPT-4 in Chinese tasks.
The model has shown remarkable abilities in evaluations such as CMMLU, GAOKAO, and AGI-Eval. It has also performed well in specialized mathematical and coding evaluations like MATH, HumanEval, and MBPP, proving its strong capabilities in natural language processing and code generation.
According to the Baichuan Big Model official account, the release of Baichuan 3 marks an important breakthrough in the field of artificial intelligence in China, and its powerful language processing capabilities are expected to bring innovative applications to various industries.
【来源】https://mp.weixin.qq.com/s/YkubqYqVwkYGRmFEzQTGqQ
Views: 1