Metraipolis › archive › dc_405dbeb5392a

Can LLMs Debug Robots by Editing Their Weights? — arXiv:2611.09376

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.09376
License (as archived)
CC0-1.0
Archived
2026-11-09T22:35:01.027Z
SHA-256 of the archived text
f08ad9396ce6d8bccd41ea0ef2c458b72bc0178413a0438d8cb3b7b7b7447af4
Stable URL of this page
https://metraipolis.com/documento/dc_405dbeb5392a

How to cite

Can LLMs Debug Robots by Editing Their Weights? — arXiv:2611.09376 — arXiv (en), https://arxiv.org/abs/2611.09376 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_405dbeb5392a (Metraipolis archive dc_405dbeb5392a, sha256 f08ad9396ce6d8bccd41ea0ef2c458b72bc0178413a0438d8cb3b7b7b7447af4)
BibTeX
@misc{dc405dbeb5392a,
  title = {Can LLMs Debug Robots by Editing Their Weights? — arXiv:2611.09376},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.09376},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_405dbeb5392a; sha256 f08ad9396ce6d8bccd41ea0ef2c458b72bc0178413a0438d8cb3b7b7b7447af4},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.09376. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A language model can repair a robot policy's lexical instruction grounding by editing just 65 token-embedding rows. We bring weight-space programming to robot foundation policies: GPT-6 Astra reads weights and uses a CPU workspace and budgeted simulator feedback, without gradient updates or new demonstrations. Three independent sessions on $\pi_{0.5}$ all improve success on 765 held-out paraphrases; the best raises success from 86.3\% to 92.0\% (48 repairs, four regressions; exact paired $p=1.31\times10^{-10}$). Among the 335 instructions containing edited tokens, success rises from 74.3\% to 87.5\%; the other 430 episodes have zero outcome changes. Matched random-direction edits do not help. Full replacement using the editor's own lexicon explains much of the gain but produces more regressions; its direct difference from the editor is not significant at 0.05. Across 13 further robot sessions, the editor declines 12 candidates; the sole retained edit does not show a significant held-out gain. Separately, benchmark metadata lowers FastWAM success from 70.0\% to 48.6\%, while the evaluated VLA's aggregate rate is unchanged. The evidence supports precise lexical repair, not general acquisition of missing robot capabilities.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Robotics (cs.RO) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2611.09376 — 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 2026-11-09T22:30:47.327Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).