Content filtering of results

Kieran Paton 0

The content filtering of results seems a little over sensitive. For instance, it will trigger in the playground if you ask for 'breeds of cats', due to the result maine coon, I assume - if you ask for 'breeds of cats, not including maine coon' it works perfectly (so doesn't seem to trigger on the input).

The example is just to illustrate the point however - is there anything planned to give more control over content filtering in future, so that false positives like this could be assessed and handled? Perhaps something like the moderations api on OpenAI https://platform.openai.com/docs/guides/moderation/overview that returns a score for different categories.

romungi-MSFT 42,311 Reputation points Microsoft Employee

2023-03-06T11:45:38.72+00:00
Kieran Paton Thanks for adding this example. We will pass this feedback to the product team for review.

At the moment, the policy on content filters allows a request to be submitted by customers to opt out of logging and human review process by submitting this form. You can choose the required option on question no. 24 of the form to opt out of these features with the option to modify following:

abuse monitoring

content filtering

I hope this helps!!
Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.