vivo artificial intelligence model C-Eval performs well close to GPT-4

Kuai Technology reported on August 16 that earlier we reported on Xiaomi’s large artificial intelligence model MiLM-6B. Now, a domestic mobile phone manufacturer’s self-developed large model has also begun to attract attention

Recently, vivo’s self-developed large-scale pre-trained language model “vivo_Agent_LM_7B” appeared in C-Eval and achieved excellent results.

According to information from C-Eval, vivo_Agent_LM_7B is a large-scale pre-trained language model independently developed by vivo AI Global Research Institute, with a parameter number of 7 billion

vivo_Agent_LM_7B achieved a final score of 64.4 points in the test results, ranked fifth in C-Eval, and achieved high scores of 80-90 in multiple subjects in the social science field

In addition, its overall score is already quite close to the second-ranked GPT-4 (68.7 points).

This proves its excellence to a certain extent

Unsurprisingly, vivo_Agent_LM_7B model will eventually be used in vivo’s voice assistant Jovi, bringing a better user experience to users.

The above is the detailed content of vivo artificial intelligence model C-Eval performs well close to GPT-4. For more information, please follow other related articles on the PHP Chinese website!