Anthropic, 2024-12-18; retrieved 2026-09-15 tier 3 self (anthropic)confirmed
Alignment faking in large language models
https://www.anthropic.com/research/alignment-faking
Lab-published research page (self-published by Anthropic, a third party to Redwood): "A new paper from Anthropic's Alignment Science team, in collaboration with Redwood Research"; acknowledgements name the collaboration; four independent reviewers.
Signals citing this source
- Redwood Research, P (against): Co-authored alignment research with Anthropic; close talent flow with lab safety teams.“in collaboration with Redwood Research , provides the first empirical example of a large language model”
- Redwood Research, S (against): Redwood co-authored the December 2024 alignment-faking paper with Anthropic's Alignment Science team, an evaluation of Claude 3 Opus carried out with the developer whose models Redwood also evaluates.
Ledger rows citing this URL
- none