Domestic large models are actively exploring ways to surpass GPT-4. However, existing evaluation methods have limitations, such as test leaks and insufficient credibility. In order to standardize the evaluation of large models and provide a more reliable reference for industry development, it is crucial to objectively and fairly evaluate the technical level of large models. This article will discuss the development status and challenges of domestic large models.
Domestic large models are exploring ways to surpass GPT4, and various evaluation methods reveal the capabilities of large models, but there are test leaks and credibility doubts. The China Academy of Information and Communications Technology has released a national standard plan to provide an official and authoritative standard for large model evaluation.
The national standard plan issued by the China Academy of Information and Communications Technology provides an important guarantee for the healthy development of domestic large models and marks a key step in the field of large model evaluation in my country. In the future, a more complete evaluation system will continue to promote the technological progress of domestic large models, and ultimately achieve competition and surpassing the international advanced level.