Is it better to use this work to achieve intent recognition than to invoke the large model directly? Because its format output is more stringent

noamgat / lm-format-enforcer

Enforce the output format (JSON Schema, Regex etc) of a language model

MIT License

1.01k stars 46 forks source link

Is it better to use this work to achieve intent recognition than to invoke the large model directly? Because its format output is more stringent #33

Closed gzchenjiajun closed 7 months ago

gzchenjiajun commented 7 months ago

thks

noamgat commented 7 months ago

This is too broad of a question, and relies on many factors such as the model being used, the prompt being asked, and more. My two cents: try and see.