近年来,李开复创办的AI公司“零一万物”推出了一系列开源大模型,包括Yi-34B和Yi-6B,这些模型声称对学术研究完全开放,同步开放免费商用申请。然而,最近有开发者质疑称该模型使用了Meta LLaMA的架构,只对两个张量(Tensor)名称进行了修改。
据报道,Yi-34B和Yi-6B使用了与LLaMA相似的架构,包括Transformer和BERT等模型的基本结构。虽然名称不同,但它们都具有类似的性能和功能。这引发了对“零一万物”公司是否涉嫌抄袭的疑问。
值得注意的是,“零一万物”公司声称对学术研究完全开放,但有人质疑其并未真正遵循这一原则。虽然该公司的开源大模型对公众免费使用,但有人指出该公司的商业版模型使用了付费算法,这涉嫌违反了开源协议。
目前,“零一万物”公司尚未就这些指控做出任何回应。
英文标题:AI startup “零一万物” accused of copying LLaMA’s open-source model
关键词:AI startup, “零一万物”, open-source model, copying, LLaMA
新闻内容:
近年来,李开复创办的AI公司”零一万物”推出了一系列开源大模型,包括Yi-34B和Yi-6B,这些模型声称对学术研究完全开放,同步开放免费商用申请。然而,最近有developers质疑称该模型使用了Meta LLaMA的架构,只对两个张量(Tensor)名称进行了修改。
According to reports, Yi-34B and Yi-6B use similar architectures to LLaMA, including Transformer and BERT models. Although the names are different, they also have similar performance and functionality. This has led to questions about whether “零一万物” company has copied the model.
It is worth noting that “零一万物” company claims to be completely open to academic research, but there are still doubts about whether it has truly followed this principle. Although its open-source models are available to the public free of charge, someone pointed out that the company’s commercial version uses paid algorithms, which may violate the open-source agreement.
目前,”零一万物”公司尚未就这些指控做出任何回应。
【来源】https://www.ithome.com/0/732/405.htm
Views: 1