Metraipolis › archive › dc_cde8f23cc588

Linear Recurrent Memory Suffices to Distil a World-Model Policy for Robot Air Hockey — arXiv:2609.39151

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39151
License (as archived)
CC0-1.0
Archived
2026-10-07T13:47:29.143Z
SHA-256 of the archived text
81dc234ac2429dbb9ea031da02603037bf7cc10a90312d5c23c78485d1bf7208
Stable URL of this page
https://metraipolis.com/documento/dc_cde8f23cc588

How to cite

Linear Recurrent Memory Suffices to Distil a World-Model Policy for Robot Air Hockey — arXiv:2609.39151 — arXiv (en), https://arxiv.org/abs/2609.39151 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_cde8f23cc588 (Metraipolis archive dc_cde8f23cc588, sha256 81dc234ac2429dbb9ea031da02603037bf7cc10a90312d5c23c78485d1bf7208)
BibTeX
@misc{dccde8f23cc588,
  title = {Linear Recurrent Memory Suffices to Distil a World-Model Policy for Robot Air Hockey — arXiv:2609.39151},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39151},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_cde8f23cc588; sha256 81dc234ac2429dbb9ea031da02603037bf7cc10a90312d5c23c78485d1bf7208},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39151. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Does memory-dependent control need nonlinear recurrent dynamics? We study simulated air-hockey defence under temporary loss of puck tracking. A DreamerV3 teacher outperforms a memoryless policy under tracking loss, while resetting the teacher's recurrent state sharply reduces performance, which demonstrates that the task requires memory. We distil this teacher into compact recurrent policies with a 64 dimensional state, with a combination of a diagonal linear recurrence and an optional rank-$k$ nonlinear innovation while retaining nonlinear observation encoders and action heads. Across five matched seeds, the purely linear recurrent model ($k=0$) matches both the GRU baseline and the teacher throughout the tested range of tracking loss. Increasing nonlinear innovation rank providing no measured benefits. This result is obtained on a fresh test split, which will be only opened after all models and analyses are frozen. The linear model requires fewer recurrent parameters and less computation than GRU, but performs comparably. These results suggest that, for this memory dependent control task, nonlinear representation learning around a simple linear memory mechanism can be sufficient, and that nonlinear recurrent dynamics are not necessarily required. These conclusions are limited to the simulated task, teacher, state dimension, and blackout horizon considered here, and to policies whose observation encoder and action head remain nonlinear.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.LG).
Categorie arXiv: Robotics (cs.RO) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.39151 — 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 2026-10-07T13:18:10.620Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).