Metraipolis › archive › dc_528f979aee28

Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision — arXiv:2610.10332

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.10332
License (as archived)
CC0-1.0
Archived
2026-10-08T11:26:22.658Z
SHA-256 of the archived text
50359be125f498a2b3a7488a9fce799c87dc7754647690219c934e7988f0dbd1
Stable URL of this page
https://metraipolis.com/documento/dc_528f979aee28

How to cite

Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision — arXiv:2610.10332 — arXiv (en), https://arxiv.org/abs/2610.10332 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_528f979aee28 (Metraipolis archive dc_528f979aee28, sha256 50359be125f498a2b3a7488a9fce799c87dc7754647690219c934e7988f0dbd1)
BibTeX
@misc{dc528f979aee28,
  title = {Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision — arXiv:2610.10332},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.10332},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_528f979aee28; sha256 50359be125f498a2b3a7488a9fce799c87dc7754647690219c934e7988f0dbd1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.10332. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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. A central design choice is what to retain from teacher trajectories that contain reasoning, actions, and information about task progress. We introduce Task-Progress Distillation (TPD), an offline approach that pairs each demonstrated action with a short label describing the current task stage. The student learns these compact targets and selects actions by jointly scoring admissible stage--action pairs, which a deterministic harness executes in the environment. 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. Explicit stages provide an additional benefit at 200 demonstrations, improving success from 48.0\% to 67.7\% over action-only supervision. With more demonstrations, the action-only student closes the gap, and both approaches reach 76.9\% at 808 demonstrations. Shared-history analyses link part of TPD's local advantage to better decisions when moving between subgoals, particularly from object acquisition to processing. These results show that compact supervision can train effective small task agents, while explicit task progress provides additional guidance at an intermediate demonstration budget.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.10332 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T10:24:06.352Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).