Researchers Let Models Revise Their Own Alignment Docs for 20 Rounds
PKirgis's team had models iteratively revise their own constitutions, model specs and system prompts over 20 rounds, empirically observing how model morality self-modifies, drawing wide praise for the experimental design.
2026-10-14 ~ 2026-10-14 · 2 related posts
- Researchers let models revise their own alignment docs for 20 iterations — sethlazar · 2026-10-14
- Researchers Let AI Models Amend Their Own Constitutions 20 Times in a Row — dhadfieldmenell · 2026-10-14