LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models

TL;DR

LIBERO-Para is a benchmark designed to evaluate the robustness of vision-language-action models to paraphrased instructions, revealing significant performance drops primarily due to lexical and planning-level variations.

Contribution

It introduces LIBERO-Para, a controlled benchmark with a new metric PRIDE, to analyze and quantify paraphrase robustness in VLA models, highlighting their reliance on surface-level cues.

Findings

Models show 22-52 percentage points performance drop under paraphrasing.

Object lexical variation causes large performance drops, even with simple synonyms.

Most failures (80-96%) are due to planning-level divergence, not execution errors.

Abstract

Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https://github.com/cau-hai-lab/LIBERO-Para