Metraipolis › archive › dc_c82b425cb62e

Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning — arXiv:2609.34575

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34575
License (as archived)
CC0-1.0
Archived
2026-10-07T20:57:56.733Z
SHA-256 of the archived text
c7e185111f72632c167fe5c24c423a4d3b3a8f22d1e80960c91260302279462a
Stable URL of this page
https://metraipolis.com/documento/dc_c82b425cb62e

How to cite

Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning — arXiv:2609.34575 — arXiv (en), https://arxiv.org/abs/2609.34575 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c82b425cb62e (Metraipolis archive dc_c82b425cb62e, sha256 c7e185111f72632c167fe5c24c423a4d3b3a8f22d1e80960c91260302279462a)
BibTeX
@misc{dcc82b425cb62e,
  title = {Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning — arXiv:2609.34575},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34575},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c82b425cb62e; sha256 c7e185111f72632c167fe5c24c423a4d3b3a8f22d1e80960c91260302279462a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34575. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Offline goal-conditioned reinforcement learning (GCRL) learns goal-directed policies from reward-free data, but in long-horizon tasks, goal-conditioned value functions often provide unstable guidance due to sparse rewards and discounting. Hierarchical methods partially mitigate this issue via subgoal decomposition; however, high-level decision-making still relies on noise-sensitive value estimates, leading to unstable behavior in complex environments. We address this limitation by proposing \textbf{D}iffusion \textbf{S}ubgoal \textbf{P}lanning (\textbf{DSP}), a diffusion-based framework for high-level subgoal generation. DSP casts high-level planning as guided generative inference over goal-conditioned subgoals and learns both conditional and unconditional flows, enabling classifier-free guidance to introduce a goal-directed bias at inference time. By removing explicit value-based guidance from high-level planning, DSP generates reachable and goal-directed subgoals through a generative model while retaining hierarchical execution. Experiments on offline GCRL benchmarks demonstrate that DSP outperforms prior methods on a range of navigation and manipulation tasks, with particularly strong performance in maze environments that require multi-step subgoal planning.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34575 — 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 2026-10-07T20:29:01.595Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).