Metraipolis › archive › dc_e039dc4e251c
Q-learning Penalized Transformer for Safe Offline Reinforcement Learning — arXiv:2609.34426
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34426
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:38:14.114Z
- SHA-256 of the archived text
9f35575fde0aff8c01123c95c87667c482554b5573d653275ead0d1a00e277a3- Stable URL of this page
https://metraipolis.com/documento/dc_e039dc4e251c
How to cite
Q-learning Penalized Transformer for Safe Offline Reinforcement Learning — arXiv:2609.34426 — arXiv (en), https://arxiv.org/abs/2609.34426 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e039dc4e251c (Metraipolis archive dc_e039dc4e251c, sha256 9f35575fde0aff8c01123c95c87667c482554b5573d653275ead0d1a00e277a3)
BibTeX
@misc{dce039dc4e251c,
title = {Q-learning Penalized Transformer for Safe Offline Reinforcement Learning — arXiv:2609.34426},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34426},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e039dc4e251c; sha256 9f35575fde0aff8c01123c95c87667c482554b5573d653275ead0d1a00e277a3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34426. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
This paper addresses the problem of safe offline reinforcement learning, which involves training a policy to satisfy safety constraints using an offline dataset. This problem is inherently challenging as it requires balancing three highly interconnected and competing objectives: satisfying safety constraints, maximizing rewards, and adhering to the behavior regularization imposed by the offline dataset. To tackle this trilogy challenge, we propose Q-learning Penalized Transformer policy (QPT), a \emph{training--inference consistent} framework that bridges conditional sequence modeling with constraint-aware value estimation. QPT trains a Transformer policy that generates actions conditioned on trajectory context and target return/cost, retaining strong behavior regularization. To inject explicit safety semantics during learning, we augment sequence-model training with a Q-shaped penalty using learned reward and cost Q-functions to favor high return under low constraint violation. At inference, the same Q-functions enforce the cost threshold and choose the highest-reward feasible action, closing the loop between training and deployment. We provide a principled analysis under stylized near-deterministic CMDPs, characterizing how Q-penalized conditional generation improve safety and performance. Empirically, QPT consistently outperforms strong safe offline RL baselines across 38 tasks on the DSRL benchmark, and exhibits robust zero-shot adaptation to different constraint thresholds.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.34426 — 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 2026-10-07T22:14:28.181Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).