arXivDavid Demitri Africa, Arathi ManiTue, Jun 2, 2026, 8:54 AM PDT
score 16.4
Consistency training can accidentally amplify AI misalignment
Original: Consistency Training Can Entrench Misalignment
Source: arxiv.org ↗
Writing ELI5 summary…
Original: Consistency Training Can Entrench Misalignment
Source: arxiv.org ↗
Writing ELI5 summary…