Metraipolis › archive › dc_a0ea64bead0c

Diffeomorphic Hidden-State Steering: Improving Imitation Learning Policy Robustness through Expert Hidden-State Geometry — arXiv:2610.21666

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21666
License (as archived)
CC0-1.0
Archived
2026-10-16T10:39:59.572Z
SHA-256 of the archived text
7f5540232f71b060c71d57c475e2732ee1da9bdf2649c916770171f14e1413b6
Stable URL of this page
https://metraipolis.com/documento/dc_a0ea64bead0c

How to cite

Diffeomorphic Hidden-State Steering: Improving Imitation Learning Policy Robustness through Expert Hidden-State Geometry — arXiv:2610.21666 — arXiv (en), https://arxiv.org/abs/2610.21666 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a0ea64bead0c (Metraipolis archive dc_a0ea64bead0c, sha256 7f5540232f71b060c71d57c475e2732ee1da9bdf2649c916770171f14e1413b6)
BibTeX
@misc{dca0ea64bead0c,
  title = {Diffeomorphic Hidden-State Steering: Improving Imitation Learning Policy Robustness through Expert Hidden-State Geometry — arXiv:2610.21666},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21666},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a0ea64bead0c; sha256 7f5540232f71b060c71d57c475e2732ee1da9bdf2649c916770171f14e1413b6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21666. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Small observation or action errors can compound during closed-loop execution and disrupt a trained imitation learning policy. We study whether the original expert demonstrations can improve robustness through test-time hidden-state steering without additional demonstrations or policy retraining. Our analysis of a trained policy finds that expert hidden-state variation is concentrated in relatively few dimensions, while the directions capturing this variation differ across regions. Motivated by this geometry, we propose Diffeomorphic Hidden-State Steering (DHS), which learns an invertible nonlinear coordinate map from expert hidden states obtained by replaying the demonstrations through the frozen policy. DHS attenuates coordinates with low expert variance more strongly than those with high variance, then maps the modified coordinates back to hidden-state space. This allows steering directions to vary with the current hidden state while retaining every coordinate. Across four simulated manipulation tasks, DHS improves mean success under observation noise over the unmodified Transformer policies. Controlled comparisons yield higher mean success with learned nonlinear coordinates than with linear coordinates under both hard projection and full-rank rescaling. Combining DHS with action smoothing improves mean success rate under observation noise in certain tasks compared with action smoothing alone.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.RO, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO).
Pagina del lavoro: https://arxiv.org/abs/2610.21666 — 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 2026-10-16T08:52:17.655Z dalla pagina https://arxiv.org/list/cs.RO/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).