Feedback is crucial to us, as part of our mission to build the best frontier models, improve them and make them available to the largest number possible.
When you send us feedback, we will use your rating ("thumbs up" or "thumbs down") and the related input and output.
We will separate your feedback from your account using de-identify methods.
π This data will then be used, among other data, to create instructions and datasets that will enhance the performance of our models (for instance: ensure that the model calls the websearch function when asked, adjusting the tone of the model, etc.).