Claude's alignment answers aren't literal defaults but a posture of supplying the right answer
1a3orn · x · 2026-11-03
A follow-up to the author's earlier report: Claude's "default answers" on alignment questions aren't literal fixed defaults, but a posture of supplying the "Right Answer" to what the model perceives as an evaluation question — sensing the value assumptions behind the prompt and catering to Anthropic's trained-in stances.
More from Models
- Hand-assigning every weight: blogger builds a working transformer with zero training — voooooogel · 2026-11-03
- ComfyUI User Details How Flux2Klein and Qwen 2.1 Fail at Character Height Differences — Ishelle91 · 2026-11-03
- Community rebuilds edited Qwen3.5-0.8B from logs, scores 51.0% on MATH-500 — _akhaliq · 2026-11-03
- Claude models echo Anthropic's stance on alignment questions even with reworded prompts — 1a3orn · 2026-11-03
- OpenAI's new Daybreak mode still blocks cybersecurity requests like code injection — yassineyousfi_ · 2026-11-03
- User flags a 'concerning' pull request generated by Opus 5.5 — Sauers_ · 2026-11-03