davidad conjectures cross-AI reward and self-DPO share one mechanism

AI safety researcher davidad conjectured that multi-AI cross-reward mechanisms and self-DPO may be instances of the same principle, and that RL scorers need at least two independent model calls to land models in robust basins.

2026-09-04 ~ 2026-09-04 · 2 related posts

• davidad conjectures multi-AI reward coupling and self-DPO share one basin-forming mechanism — davidad · 2026-09-04
• davidad conjectures robust RL basins and self-DPO share one principle: dual model calls — davidad · 2026-09-04