2024 08 23

Introducing CaRDiff, a framework for video saliency prediction using MLLM CoT reasoning and diffusion model.