Selecting Long-Horizon Trajectories for Reliable and Efficient Terminal-Agent Training

2026-10-05

AI Summary

This study addresses the reliability-cost imbalance in terminal agent training caused by undefined supervision horizon lengths, proposing a selective long-horizon refinement method. Grounded in a bias-complexity theoretical analysis that reveals performance saturation, this work establishes a new paradigm demonstrating trajectory selection's superiority over full-scale training. A two-stage progressive strategy is designed to first warm-start the model with short prefixes and subsequently perform selective continuation optimization via probabilistic filtering. This approach significantly enhances task success rates and stability while reducing training overhead by 30%, substantially outperforming existing baselines across multiple benchmarks.

Abstract

Terminal agents are commonly trained by imitating long teacher trajectories, yet how much of each trajectory to supervise remains unexplored. We study the \emph{supervision horizon}, the number of trajectory tokens retained for training, and show that it is a key design axis for reliability and cost. Reliability improves with longer horizons but saturates: on Terminal-Bench, a 12K-token horizon solves more tasks than 16K ($29\pm0.7$ vs.\ $26\pm0.8$) while requiring 30\% less training time. The horizon also shapes agent behavior: short horizons cause premature termination, intermediate horizons yield productive error recovery, and long horizons induce over-persistence. We analyze this saturation through a bias--complexity bound, in which longer supervision reduces temporal supervision bias but increases finite-sample estimation error from more heterogeneous late-stage histories. Guided by this analysis, we propose \emph{selective long-horizon refinement}, which first trains on short prefixes and then refines only on continuations that are most likely under the warm-start model. It consistently outperforms full long-horizon training. At 16K, it raises successful attempts from $110\pm2.7$ to $126\pm2.1$ and tasks solved in at least six of eight attempts from $9\pm0.7$ to $14\pm0.6$; with half of the long-horizon data, it still reaches $122\pm2.4$ while cutting training time by 23\%. The gains transfer across benchmarks, from $64\pm2.6$ to $73\pm2.1$ on Terminal-Bench v2.0 and from $137\pm2.7$ to $155\pm2.2$ on OpenThoughts-TBLite. For long-horizon supervision, selecting the right trajectories matters more than training on all of them.

Problem

• terminal agent training
• supervision horizon
• trajectory selection
• imitation learning
• training efficiency

Innovation

• Supervision Horizon
• Selective Long-Horizon Refinement
• Bias-Complexity Bound
• Terminal Agents
• Trajectory Selection