GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior

2 pointsposted 10 hours ago
by eatitraw
(lesswrong.com)

No comments yet