AI RESEARCH
Via Negativa for AI Alignment: Why Negative Constraints Are Structurally Superior to Positive Preferences
arXiv CS.AI
•
ArXi:2603.16417v1 Announce Type: new Recent empirical results have nstrated that