VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation

Part of

Deep Learning Model Evaluation Benchmarks