a very "plausible" model

#10
by J22 - opened

I have tested this model with ChatLLM.cpp. The conversation goes quite well. Some impressions:

  1. Performance on Chinese is really bad.
  2. Strong context understanding, such as correcting math results, step by step analysis of quick sort, etc.
  3. Generate plausible information from time to time, for example, insisting a famous programmer to be of Geman desent in several round of conversations, in which case even a (fake) Youtube link is given as a reference.

Sign up or log in to comment