Jeuk Kang

Running VLA benchmarks requires setup and dependencies. This launch removes that overhead with a no-setup execution layer for one-line benchmark execution, enabling parallel, cross-benchmark evaluation at scale.