Or just rerun that preference learning approach, but deliberately, with the filt...

gwern on April 8, 2022 | parent | context | favorite | on: Dall-E 2

Or just rerun that preference learning approach, but deliberately, with the filters as the reward function and the reward being for being filtered.