Metraipolis › archive › dc_4b455ca1cc43
DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents — arXiv:2610.19376
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.19376
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T20:01:19.500Z
- SHA-256 of the archived text
5eeaba396c9777a17df2ad40b6f4c82663f83c1776dfc546bc1db85309f54247- Stable URL of this page
https://metraipolis.com/documento/dc_4b455ca1cc43
How to cite
DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents — arXiv:2610.19376 — arXiv (en), https://arxiv.org/abs/2610.19376 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4b455ca1cc43 (Metraipolis archive dc_4b455ca1cc43, sha256 5eeaba396c9777a17df2ad40b6f4c82663f83c1776dfc546bc1db85309f54247)
BibTeX
@misc{dc4b455ca1cc43,
title = {DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents — arXiv:2610.19376},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.19376},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4b455ca1cc43; sha256 5eeaba396c9777a17df2ad40b6f4c82663f83c1776dfc546bc1db85309f54247},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.19376. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Language-model agents that run long tasks often redo work they have already done: they re-issue a tool call they made earlier and get back exactly the same observation, a step that adds no new information. We identify this failure as a loop. Loops appear natively in Qwen, Llama, and Gemini agents and become far more common once the agent's history is compressed to fit its context window. To reduce these loops, we propose DeLoop, a self-supervised training method. DeLoop builds on a simple observation: the agent's own rollout can provide the supervision needed to learn from its loops. We use a looping action as a negative example and the first later non-looping action in the same rollout as a positive example. DeLoop learns from these pairs through direct preference optimization and does not require task rewards, gold answers, or LLM judges. On CodeGym under compression, DeLoop reduces the share of Qwen3-8B tasks containing loops from 57.3% to 6.1%. It also uses about one third as many steps per task, with similar task success. On Qwen3-32B, DeLoop reduces loops and steps while raising task success from 37.7% to 44.8%. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.19376 — 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 2026-10-15T19:35:48.218Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).