LangGap: Diagnosing and Closing the Language Gap in Vision-Language-Action Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hou, Yuchen, Zhao, Lin
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908858529611776
author Hou, Yuchen
Zhao, Lin
author_facet Hou, Yuchen
Zhao, Lin
contents Vision-Language-Action (VLA) models achieve over 95% success on standard benchmarks. However, through systematic experiments, we find that current state-of-the-art VLA models largely ignore language instructions. Prior work lacks: (1) systematic semantic perturbation diagnostics, (2) a benchmark that forces language understanding by design, and (3) linguistically diverse training data. This paper constructs the LangGap benchmark, based on a four-dimensional semantic perturbation method -- varying instruction semantics while keeping the tabletop layout fixed -- revealing language understanding deficits in π0.5. Existing benchmarks like LIBERO assign only one task per layout, underutilizing available objects and target locations; LangGap fully diversifies pick-and-place tasks under identical layouts, forcing models to truly understand language. Experiments show that targeted data augmentation can partially close the language gap -- success rate improves from 0% to 90% with single-task training, and 0% to 28% with multi-task training. However, as semantic diversity of extended tasks increases, model learning capacity proves severely insufficient; even trained tasks perform poorly. This reveals a fundamental challenge for VLA models in understanding diverse language instructions -- precisely the long-term value of LangGap.
format Preprint
id arxiv_https___arxiv_org_abs_2603_00592
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LangGap: Diagnosing and Closing the Language Gap in Vision-Language-Action Models
Hou, Yuchen
Zhao, Lin
Robotics
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
Vision-Language-Action (VLA) models achieve over 95% success on standard benchmarks. However, through systematic experiments, we find that current state-of-the-art VLA models largely ignore language instructions. Prior work lacks: (1) systematic semantic perturbation diagnostics, (2) a benchmark that forces language understanding by design, and (3) linguistically diverse training data. This paper constructs the LangGap benchmark, based on a four-dimensional semantic perturbation method -- varying instruction semantics while keeping the tabletop layout fixed -- revealing language understanding deficits in π0.5. Existing benchmarks like LIBERO assign only one task per layout, underutilizing available objects and target locations; LangGap fully diversifies pick-and-place tasks under identical layouts, forcing models to truly understand language. Experiments show that targeted data augmentation can partially close the language gap -- success rate improves from 0% to 90% with single-task training, and 0% to 28% with multi-task training. However, as semantic diversity of extended tasks increases, model learning capacity proves severely insufficient; even trained tasks perform poorly. This reveals a fundamental challenge for VLA models in understanding diverse language instructions -- precisely the long-term value of LangGap.
title LangGap: Diagnosing and Closing the Language Gap in Vision-Language-Action Models
topic Robotics
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2603.00592