Metraipolis › archive › dc_37ffd95b2357
Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models — arXiv:2610.10526
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.10526
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T11:26:42.532Z
- SHA-256 of the archived text
f9794c60ef8bc52e6d209c0d6950aa18fc26983446026140a7f72b1974a5b1bd- Stable URL of this page
https://metraipolis.com/documento/dc_37ffd95b2357
How to cite
Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models — arXiv:2610.10526 — arXiv (en), https://arxiv.org/abs/2610.10526 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_37ffd95b2357 (Metraipolis archive dc_37ffd95b2357, sha256 f9794c60ef8bc52e6d209c0d6950aa18fc26983446026140a7f72b1974a5b1bd)
BibTeX
@misc{dc37ffd95b2357,
title = {Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models — arXiv:2610.10526},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.10526},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_37ffd95b2357; sha256 f9794c60ef8bc52e6d209c0d6950aa18fc26983446026140a7f72b1974a5b1bd},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.10526. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models they are built on. A one-word edit can move success by tens of points: $\pi_{0.5}$ turns on a LIBERO stove 100% of the time for "switch on the stove" and 2% for "switch on the hot plate", and a $\pi_0$ checkpoint finetuned with rephrase augmentation still shows swings of up to 61 points. We characterize this sensitivity with statistically tested single-edit swings and an oracle phrase search, which shows that phrasing alone nearly closes the 21-point gap between in-distribution and out-of-distribution tasks. We then reduce it without modifying the policy. Because the sensitivity is systematic, it can be expressed as explicit rules: we score many phrasings of a few training tasks, have a large language model distill the evidence into ten to twenty rephrasing rules, and at deployment rewrite each incoming instruction once under these rules. The rules improve the frozen $\pi_0$ by 16 to 27% relative on twelve held-out tasks across adversarial, VLM-generated, and human-generated phrasings, with gains concentrated on out-of-distribution tasks. The pipeline replicates on $\pi_{0.5}$ and LIBERO, lifting in-finetune success from 93.6% to 97.8%. The method requires no retraining and no per-step verification, and applies zero-shot to unseen tasks and instructions. Project website: this https URL
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO) ; Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.10526 — 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 2026-10-08T10:24:06.400Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).