Paraphrase Consensus Is Not Self-Consistency: Instruction Reliance and Robustness in VLA Policies
From the abstract
Vision-language-action (VLA) policies fine-tuned on LIBERO are trained and evaluated on the same ten instruction strings, and rewording the instruction changes base OpenVLA's success by -38.0 points. Test-time prompt ensembling, which improves language models, does not recover this loss in any setting we ran.
From the abstract. Our summary is in progress.
The full text is not republished here because the paper's license does not allow it. Read the original on arXiv.