Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignment | Signal Canvas | ScienceToStartup