Comment by fragmede
2 years ago
Pretty clever to use a specific length as a test for quality of output, since text itself is subjective. Another one might be to see if it's lazy with code generation with and without positive/negative reinforcement.
2 years ago
Pretty clever to use a specific length as a test for quality of output, since text itself is subjective. Another one might be to see if it's lazy with code generation with and without positive/negative reinforcement.
Except that LLMs are notoriously bad at counting characters.