Metraipolis › archive › dc_a403105e021f

Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies — arXiv:2608.04692

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.04692
License (as archived): CC0-1.0
Archived: 2026-10-31T11:35:44.084Z
SHA-256 of the archived text: d6f9cb675dfc1cb358089af540ea0aa7c8ea19d459edc4742cd75867aca10bae
Stable URL of this page: https://metraipolis.com/documento/dc_a403105e021f

How to cite

Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies — arXiv:2608.04692 — arXiv (en), https://arxiv.org/abs/2608.04692 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a403105e021f (Metraipolis archive dc_a403105e021f, sha256 d6f9cb675dfc1cb358089af540ea0aa7c8ea19d459edc4742cd75867aca10bae)

BibTeX

@misc{dca403105e021f,
  title = {Suppression Sticks, Locality Is Fragile: A Closed-Loop Target-and-Control Audit of Task-Vector Negation in VLA Policies — arXiv:2608.04692},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.04692},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a403105e021f; sha256 d6f9cb675dfc1cb358089af540ea0aa7c8ea19d459edc4742cd75867aca10bae},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.04692. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Task-vector arithmetic offers a closed-form way to modify a model, yet its behavioral locality remains unclear in closed-loop robot control. We present a target-and-control audit of per-skill task-vector subtraction from multitask vision-language-action (VLA) policies. Across all ten LIBERO-Goal skills, subtraction produces three qualitatively different regimes: target-control separation for five skills, resistance for three, and global collapse for two. On held-out initial states, the five suppressible targets remain at 0% success; however, mean baseline-normalized control retention is only 52%, and each target-suppressing edit materially harms at least one nominally unrelated control. Additional Goal panels show separation across tested policies with continuous-regression, discrete-token, and flow-matching action heads, whereas we observe no clean separation on Spatial and control collapse on the tested Object and Long-horizon panels. Mean task-vector cosine does not account for this variation. A matched-norm control identifies a local sign asymmetry around one Goal anchor, while multi-vector outcomes vary with anchor and scale. Retain-aware gradient baselines provide data-dependent comparators but require removal-time data and optimization; subtraction is data- and gradient-free only at edit time, assuming precomputed expert deltas. Finally, a single-skill relearning probe is consistent with behavioral masking, not certified unlearning. These results characterize task-vector subtraction as a fast but brittle intervention and underscore the need for closed-loop target-and-control evaluation when assessing locality in embodied model editing.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-06 in cs.LG).
Categorie arXiv: Robotics (cs.RO) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.04692 — 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 2026-10-31T11:24:48.837Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-06?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).