A Phenomenology of Neural Geometry in Vision-Language-Action Models

Key points

• Prior work on vision-language-action (VLA) policies establishes that task-relevant variables are linearly readable from activations, yet readability certifies only that a variable is present and says nothing about the shape its values trace, which may be curved and multi-dimensional.1
• Using controlled sweeps, natural rollouts, linear probes, and activation interventions, we build a taxonomy of seven task-relevant concepts in VLA models.1
• We find that realized geometry is predominantly multi-dimensional and is shaped by the constraints under which the policy acts.1
• Overall, this work lays out an initial phenomenology of realized geometry in VLA policies and demonstrates that a concept's shape carries as much importance as its readability.1

Sources (1)

• [1]A Phenomenology of Neural Geometry in Vision-Language-Action Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 07:27 PM

  Robotic manipulation is inherently geometric, but what shape does the world take inside the policy that performs it?

  Prior work on vision-language-action (VLA) policies establishes that task-relevant variables are linearly readable from activations, yet readability certifies only that a variable is present and says nothing about the shape its values trace, which may be curved and multi-dimensional.

Extractive summary: sentences quoted from the sources.