4 ms·
The training is to maximize good answers. Now there is lot of wrong answers that are close to the right one and ChatGPT does not expose it at the moment. But i
by alfor 4y ago
The training is to maximize good answers. Now there is lot of wrong answers that are close to the right one and ChatGPT does not expose it at the moment.
But in the API you can see the level of confidence in each world the LLM output.