The Compression Gap: Why Discrete Tokenization Limits Vision-Language-Action Model Scaling

Forum topic · 小凯 · 2026-04-06

Summary

This paper introduces the Compression Gap, an information-theoretic principle explaining why upgrading vision encoders fails to improve Vision-Language-Action (VLA) models when actions are represented as discrete tokens. The authors show that in any visuomotor pipeline, scaling behavior is governed by the location of the tightest information bottleneck. When actions are continuous, as in Diffusion Policy, the vision encoder is the binding constraint and improving it directly boosts downstream manipulation performance. However, when actions are discretized through a fixed-capacity codebook, as in OAT, the codebook becomes the binding constraint, so encoder improvements cannot propagate past it regardless of scale. The work reframes VLA scaling strategy: instead of focusing on encoder capacity, practitioners must widen the action tokenization bottleneck to unlock the benefits of stronger vision backbones. Paper: arXiv 2604.03191.

Paper Overview

Field: Computer Vision (CV)
Author: Takuya Shiba
Published: 2026-04-03
arXiv: 2604.03191

Abstract

Scaling Vision-Language-Action (VLA) models by upgrading the vision encoder is expected to improve downstream manipulation performance — as it does in vision-language modeling. This paper shows that this expectation fails when actions are represented as discrete tokens, and explains why through an information-theoretic principle the author calls the Compression Gap: in any visuomotor pipeline, scaling behavior is governed by the location of the tightest information bottleneck.

When actions are continuous (e.g., Diffusion Policy), the vision encoder is the binding constraint, and upgrading it directly improves performance.
When actions are discretized through a fixed-capacity codebook (e.g., OAT), the codebook becomes the binding constraint, and encoder improvements cannot propagate past it — regardless of how large or capable the encoder becomes.

Key Takeaways

The tightest information bottleneck, not encoder quality, determines whether scaling investments pay off.
Discrete action tokenization with fixed-capacity codebooks caps the effective information flow from perception to action.
To benefit from stronger vision encoders in VLA models, the action representation bottleneck (codebook capacity) must be addressed first.

---

*Auto-collected on 2026-04-06*