Learning to Assist Humans without Inferring Rewards

Open in new window