[–] werdnabae 13d ago ↗ Really interesting work on reproducing reward hacking behaviors across frontier AI models.
1 comment
[ 3.0 ms ] story [ 2.8 ms ] thread