NVIDIA PivotOPD Teaches Multi-Turn AI Agents to Recover From Pivotal Mistakes Business October 8, 2026 NVIDIA researchers, with Princeton University and the University of Maryland, have introduced PivotOPD, an on-policy distillation method for multi-turn LLM…
