multimodalfactbullishAn image-blind LLM-as-a-Judge validation approach can ensure generated captions provide reliable semantic anchoring for Data-Free Knowledge Distillation by intentionally obscuring the original imageComputer Vision02 Aug 2026http://arxiv.org/abs/2607.28269v1