Trajectories That Segment Themselves: Agent-Declared Boundaries as a Training Unit

Jingxi Wei

Long-horizon coding-agent trajectories are poorly matched to the credit units available to train on: a single action has no stable value, an episode label merges productive explora…