Follow Me If You Can:Auditing Understanding-to-Generation Transfer

Understanding-only training can change image generation through shared parameters, but the resulting scores support different claims about transfer. We develop an audit framework separating source validity, target influence, and attribute correspondence. Fixed-checkpoint image and instruction controls test the source gain; paired trained and frozen states measure generation influence; conditioning swaps probe the requested value expressed by the update. We characterize a blind spot of norm-matched swaps: changes along a shared direction yield identical interventions even when their amplitudes depend on the prompt. Across six unified architectures, counting and spatial experiments reveal different answer–generation response patterns. In a focused counting audit, a $+4.98$ percentage-point answer gain becomes $+0.37/-0.85$ under digit requests at fixed weights. A standard low-rank adaptation (LoRA) reference also improves answers without a consistent generation-score gain. Joint readouts distinguish recipient improvement from donor-value following: a $+4.88$ percentage-point recipient gain accompanies a $-0.59$ donor change. Mixed same-budget prompt-contrast calibration bounds inference from non-following. The audit ties each interpretation of transfer to the comparison that supports it.