VideoLVR: Latent Visual Reasoning over Grounded Video Evidence
VideoLVR introduces latent visual reasoning to prevent visual forgetting and improve multimodal large language models' ability to reason over video content
VideoLVR introduces latent visual reasoning to prevent visual forgetting and improve multimodal large language models' ability to reason over video content