# Can LLMs Debug Robots by Editing Their Weights?

- Source: arXiv cs.RO
- Published: 2026-11-09T05:00:00.000Z
- RoboHot: https://robohot.news/en/items/bnd6oldtv8y9f09fl9o42pszy
- Original: https://arxiv.org/abs/2611.09376

## Summary

The research proposes having GPT-6 Astra read a robot's base policy weights and edit only 65 rows of token embeddings, raising the success rate on π0.5's 765 held-out paraphrased instructions from 86.3% to 92.0% (48 repairs, 4 regressions, paired p=1.31×10⁻¹⁰), all without gradient updates or new demonstrations.