7 papers
VGGA: Geometry-Grounded Autoencoder for Video Generation
Chaofan Gan, Yian Zhao, Shijie Li +4
Autoencoders are the core component of latent video generative models, compressing high-dimensional videos into compact latent spaces for downstream generation. However, existing v…
Awakening Diffusion Transformers: Eliciting Stronger Generation and Understanding via Massive Activation Modulation
Chaofan Gan, Zicheng Zhao, Yuanpeng Tu +6
Massive Activations (MAs) have been widely observed in Transformer-based models, yet their structure and functional roles in Diffusion Transformers (DiTs) remain insufficiently und…
Accurate and Efficient Long-Term Memory for LLM Agents
Zicheng Zhao, Xinyang Guo, Luyao Lv +3
LLM agents augmented with persistent memory can recall past interactions, but existing systems suffer from two limitations: flat, unstructured storage loses relational context need…
Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning
Shijie Li, Yilin Gao, Siyuan Yang +7
Multimodal Large Language Models (MLLMs) are often constrained by a language-space bottleneck, forcing complex visual reasoning into discrete tokens which can lose perceptual nuanc…
From Priors to Perception: Grounding Video-LLMs in Physical Reality
Zicheng Zhao, Chaofan Gan, Shijie Li +1
Video Large Language Models (Video-LLMs) excel in general video understanding but often base physical judgments on event expectations rather than observations. We find that they no…
Massive Activations are the Key to Local Detail Synthesis in Diffusion Transformers
Chaofan Gan, Zicheng Zhao, Yuanpeng Tu +5
Diffusion Transformers (DiTs) have recently emerged as a powerful backbone for visual generation. Recent observations reveal \emph{Massive Activations} (MAs) in their internal feat…
CogStream: Context-guided Streaming Video Question Answering
Zicheng Zhao, Kangyu Wang, Shijie Li +3
Despite advancements in Video Large Language Models (Vid-LLMs) improving multimodal understanding, challenges persist in streaming video reasoning due to its reliance on contextual…