Metraipolis › archive › dc_3a014157e97e
Horizon-Scalable Q-Learning — arXiv:2611.10987
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.10987
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-10T18:42:59.570Z
- SHA-256 of the archived text
5505d20344d2fc2e73d4e6fb5af68f2fdd2d6178825ee7c26e1b56c3f29abace- Stable URL of this page
https://metraipolis.com/documento/dc_3a014157e97e
How to cite
Horizon-Scalable Q-Learning — arXiv:2611.10987 — arXiv (en), https://arxiv.org/abs/2611.10987 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3a014157e97e (Metraipolis archive dc_3a014157e97e, sha256 5505d20344d2fc2e73d4e6fb5af68f2fdd2d6178825ee7c26e1b56c3f29abace)
BibTeX
@misc{dc3a014157e97e,
title = {Horizon-Scalable Q-Learning — arXiv:2611.10987},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.10987},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3a014157e97e; sha256 5505d20344d2fc2e73d4e6fb5af68f2fdd2d6178825ee7c26e1b56c3f29abace},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.10987. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Learning accurate off-policy state-action value functions (Q-functions) from offline datasets becomes harder as the task horizon grows. One-step temporal-difference (TD) learning bootstraps after every transition, so small per-step errors accumulate along the horizon, and early states can be badly misvalued even when the TD loss is small. Multi-step TD shortens this chain by following the dataset trajectory for several steps; however, since the behavior policy is often suboptimal, every followed action adds a bias, and no fixed number of steps suits a whole trajectory. In this paper, we propose Horizon-Scalable Q-Learning (HQL), a simple method that enables accurate long-horizon value learning by automatically balancing between one-step and multi-step returns. HQL extends a target through a dataset action only when the critic values it at least as highly as the best-of-N actions sampled from the BC policy, and otherwise falls back to a one-step target. We show that this rule is a principled off-policy correction that needs no action densities, which makes HQL compatible with expressive flow and diffusion policies. Experiments show that HQL avoids the failures of both one-step and multi-step TD on a combination-lock task across horizons, and outperforms prior horizon-reduction methods on long-horizon OGBench and OCBench tasks. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2611.10987 — 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 2026-11-10T18:37:37.618Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).