4 papers
DVTR-KD: Distilling Vision-Language Models via Intermediate Visual Token Reconstruction
Haolan Wei, Jiahui Wang, Junli Zha +4
Multimodal large language models (MLLMs) have achieved strong performance on diverse vision-language tasks, but their computational demands limit deployment under constrained budge…
Redirecting Teacher Supervision: Temporal-Response-Guided Latent Distillation for Video-Language Models
Xinkai Lu, Junli Zha, Haolan Wei +3
On-policy distillation (OPD) transfers capabilities through teacher predictions on student-generated sequences, as studied by GKD and MiniLLM for language models and Video-OPD for…
Illusion-Aware Visual Preprocessing and Anti-Illusion Prompting for Classic Illusion Understanding in Vision-Language Models
Junli Zha, Jiahui Wang, Xinkai Lu +1
Vision-Language Models (VLMs) exhibit systematic bias toward visual illusions, recalling memorized facts rather than perceiving actual visual differences. This paper presents a tra…
Collaborative Attention Memory Network for Video Object Segmentation
Zhixing Huang, Junli Zha, Fei Xie +3
Semi-supervised video object segmentation is a fundamental yet Challenging task in computer vision. Embedding matching based CFBI series networks have achieved promising results by…