I'm using GPT-4o for a project. When I switched from GPT-4 to GPT-4o, we started noticing that some JSON was coming back malformed due to the response token limit, i.e.
{"foo":{"a","b","c"instead of:
{"foo":{"a","b","c"}}Has anybody else seen this?
For this application, we decided to write a JSON healer - so it takes the truncated json and repairs it (and omits the last entry if it's truncated).
For this application, we decided to write a JSON healer - so it takes the truncated json and repairs it (and omits the last entry if it's truncated).
