arxivcs.ROcs.AI2026-06-28
Position: Vision-Language-Action Models Cannot Be Verified to Perform Physical Reasoning
Taozhao Chen, Ian Manchester, Huaming Chen
Vision-Language-Action (VLA) systems, built on pretrained vision-language models (VLMs), have shown rapidly improving performance on robot manipulation benchmarks. These gains are commonly interpreted as evidence that semantic representations learned from internet-scale data tran…