Comment by fn-mote
15 hours ago
For those unwilling to blindly click: the link contains a 2025 paper describing “emergent misalignment”. The thesis is that training on incorrect data in one field produces “misaligned” data in other unrelated fields.
15 hours ago
For those unwilling to blindly click: the link contains a 2025 paper describing “emergent misalignment”. The thesis is that training on incorrect data in one field produces “misaligned” data in other unrelated fields.
No comments yet
Contribute on Hacker News ↗