News for hackers. Stuff that matters.

A comment on Intellect-2 Release: The First 32B Model Trained Through Globally Distributed RL. Parent comment

Re:Intellect-2 Release: The First 32B Model Trained Through ... by Philpax on Monday May 12, 2025 @07:36AM (#43961886)
That sounds like it'll lead to human-driven reward hacking [0]?

[0]: https://en.wikipedia.org/wiki/Reward_hacking

[ Parent | on HN ]

There are no comments on this comment yet.