method · first seen 5 Oct 2026
trajectory supervision horizon
The number of trajectory tokens retained when training an agent by imitation learning on long sequences.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Selecting Long-Horizon Trajectories for Reliable and Efficient Terminal-Agent TrainingPapers · 5 Oct 2026 · Cuong Dang, Hoang Anh Just, Ruoxi Jia
Latest sources
- Selecting Long-Horizon Trajectories for Reliable and Efficient Terminal-Agent TrainingPapers · 5 Oct 2026 · Cuong Dang, Hoang Anh Just, Ruoxi Jia