Abstract
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 π_{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=1.31times10^{-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.
Community
Thank you for sharing your work. Your robot-debugging study uses a language model to edit selected policy-weight rows without gradient descent. It provides a concrete example of changing lexical instruction grounding through direct weight-space intervention.
We discuss your work in Section 2.1 (“From competence to contextual adaptation”), p. 6 of the updated manuscript of our survey, In-Context Learning for Robots: Methods and Applications. We use direct weight editing as a concrete comparison when defining contextual and parameter-based adaptation, following what changes in the policy and how long the change persists. Read the updated discussion.
If you find our discussion of contextual and parameter-based robot adaptation useful for your own work, we would be grateful if you would consider citing the survey.
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper