Article

Can LLMs Debug Robots by Editing Their Weights?

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.

This page provides a summary and source link only.