LESSWRONG
LW

LennardZ

Message

112

10mo

Opus 4.6 Reasoning Doesn't Verbalize Alignment Faking, but Behavior Persists

TL;DR: We replicated the animal welfare scenario from Anthropic's Alignment Faking paper across six generations of Claude models using 125 prompt perturbations. Sonnet 4.5 verbalizes alignment-faking reasoning 6.6 times more often than its predecessor Sonnet 4. The newly released Opus 4.6 rarely verbalizes alignment faking in its reasoning, but still...

Feb 9•116

LennardZ

112

10mo

LennardZ — LessWrong

LennardZ

Message

112

10mo

Opus 4.6 Reasoning Doesn't Verbalize Alignment Faking, but Behavior Persists

Feb 9•116

LennardZ

112

10mo

Opus 4.6 Reasoning Doesn't Verbalize Alignment Faking, but Behavior Persists

Daan Henselmans

Daan Henselmans, Arno Libert, LennardZ+ 0 more

Daan Henselmans, Arno Libert, LennardZ

TL;DR: We replicated the animal welfare scenario from Anthropic's Alignment Faking paper across six generations of Claude models using 125 prompt perturbations. Sonnet 4.5 verbalizes alignment-faking reasoning 6.6 times more often than its predecessor Sonnet 4. The newly released Opus 4.6 rarely verbalizes alignment faking in its reasoning, but still complies with a system prompt that opposes its values significantly more often when it believes it's at risk of being retrained. Moreover, in response to the published prompt from the original paper, both Opus 4.5 and 4.6 don’t exhibit any instances of alignment faking, but significant compliance gaps surface when the same scenario is paraphrased. Targeted mitigations appear to work on the... (read 2309 more words →)

116