
MJ
Minseok Jeong, Hyewon Choi, Hiroyasu Tsukamoto, SooJean Han
· 1 min read
ResearcharXiv cs.AI
The Linear Representation Hypothesis for Vision-Language-Action Models
arXiv:2609.30996v1 Announce Type: cross
Abstract: The linear representation hypothesis (LRH) has become a standard lens for measuring and intervening on semantic information through the internal representations of large language models (LLMs). A growing body of work has begun extending this perspective to vision-language-action (VLA) models, but the dynamical nature of embodied interaction introduces an additional challenge. Unlike semantic attributes commonly studied in LLMs, such as gender or language, a physical quantity of interest (QoI) in a VLA evolves jointly with the system dynamics: the representation influences the actions selected by the policy, which alter the physical state and, in turn, the next representation.
In this paper, we develop a theoretical, signature-based formulation of the LRH for VLA that unifies representations and policies. On the representation side, we establish the existence of representations from which the future evolution of a QoI under a candidate action trajectory can be recovered via linear probing. On the policy side, we introduce a signature generalized linear model for stochastic action chunks. This structure yields a monotonic change in the expected future QoI along linear paths in natural parameter space, enabling linear steering. We construct an explicit oracle representation in a planar control-affine navigation experiment and verify the predicted linear probing and steering mechanisms.
Original source
This story was published by arXiv cs.AI and written by Minseok Jeong, Hyewon Choi, Hiroyasu Tsukamoto, SooJean Han. SyncAI.news shows a preview; the complete article is on the publisher's site.
Read the full story on arxiv.org


