zlacker

>>goranm+(OP)
FWIW, I've seen stronger performance from gpt-4-1106-preview when I use `response_format: { type: "json_object" },` (providing a target typescript interface in context), vs the "tools" API.

More flexible, and (evaluating non-scientifically!) qualitatively better answers & instruction following -- particularly for deeply nested or complex schemas, which typescript expresses very clearly and succinctly.

Example from a hack week project earlier this month (using a TS-ish schema description that's copy/pasted from healthcare's FHIR standard): https://github.com/microsoft-healthcare-madison/hackweek-202...

Or a more complex example with one model call to invent a TS schema on-the-fly and another call to abstract clinical data into it: https://github.com/microsoft-healthcare-madison/hackweek-202...

>>JoshMa+Dg
So something I have also noticed, mostly on 3.5-Turbo, is textual responses in json take a quality hit, full stop. This has caused me to use mixed output usually. Thoughts and process in json, then "exit" to text for a conversational response.

It is likely also a behavior in gpt-4, but I haven't studied it as closely.