Hacker News new | ask | show | jobs
by globular-toast 502 days ago
These models have been trained with the original paper. It would be more interesting to come up with a different "attack" that hasn't already been written down.