8 papers
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…
HACK++: Towards More Effective Head-Aware Key-Value Compression for Efficient Visual Autoregressive Modeling
Ziran Qin, Yuchen Jiang, Mingbao Lin +4
Visual Autoregressive (VAR) models adopt a next-scale prediction paradigm, offering high-quality generation with substantially fewer decoding steps. However, existing VAR models su…
VidLaDA: Bidirectional Diffusion Large Language Models for Efficient Video Understanding
Zhihao He, Tieyuan Chen, Kangyu Wang +6
Current Video Large Language Models (Video LLMs) typically encode frames via a vision encoder and employ an autoregressive (AR) LLM for understanding and generation. However, this…
CAKE: Cascading and Adaptive KV Cache Eviction with Layer Preferences
Ziran Qin, Yuchen Cao, Mingbao Lin +5
Large language models (LLMs) excel at processing long sequences, boosting demand for key-value (KV) caching. While recent efforts to evict KV cache have alleviated the inference bu…
Head-Aware KV Cache Compression for Efficient Visual Autoregressive Modeling
Ziran Qin, Youru Lv, Mingbao Lin +4
Visual Autoregressive (VAR) models adopt a next-scale prediction paradigm, offering high-quality content generation with substantially fewer decoding steps. However, existing VAR m…
Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens
Ziran Qin, Youru Lv, Mingbao Lin +4
Autoregressive (AR) visual generation has emerged as a powerful paradigm for image and multimodal synthesis, owing to its scalability and generality. However, existing AR image gen…