Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision

Key points

• Learning from large-model demonstrations offers a way to train small agents that can complete recurring tasks without calling a large model at every step.1
• A central design choice is what to retain from teacher trajectories that contain reasoning, actions, and information about task progress.1
• On ALFWorld, a 1.7B student trained with 404 demonstrations achieves 72.4% mean unseen task success with either TPD or action-only supervision, compared with 48.3% for a reasoning-trained student using constrained action selection.1
• These results show that compact supervision can train effective small task agents, while explicit task progress provides additional guidance at an intermediate demonstration budget.1

Sources (1)

• [1]Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 04:15 PM

  We introduce Task-Progress Distillation (TPD), an offline approach that pairs each demonstrated action with a short label describing the current task stage.

  Learning from large-model demonstrations offers a way to train small agents that can complete recurring tasks without calling a large model at every step.

Extractive summary: sentences quoted from the sources.