6 papers
Towards Memory-Efficient Autoregressive Video Generation via Instance-Specific Parametric Absorption
Xiaomeng Fu, Jia Li, Yiming Hu +5
Autoregressive (AR) streaming models have emerged as a powerful paradigm for long video generation. However, the linearly growing Key-Value (KV) cache poses a significant bottlenec…
Train Short, Inference Long: Training-free Horizon Extension for Autoregressive Video Generation
Jia Li, Xiaomeng Fu, Xurui Peng +7
Autoregressive video diffusion models have emerged as a scalable paradigm for long video generation. However, they often suffer from severe extrapolation failure, where rapid error…
Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod
Ao Xiao, Bangzheng He, Baoquan Zhang +125
Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentr…
Improving Variable-Length Generation in Diffusion Language Models via Length Regularization
Zicong Cheng, Ruixuan Jia, Jia Li +3
Diffusion Large Language Models (DLLMs) are inherently ill-suited for variable-length generation, as their inference is defined on a fixed-length canvas and implicitly assumes a kn…
DEER: Draft with Diffusion, Verify with Autoregressive Models
Zicong Cheng, Guo-Wei Yang, Jia Li +3
Efficiency, as a critical practical challenge for LLM-driven agentic and reasoning systems, is increasingly constrained by the inherent latency of autoregressive (AR) decoding. Spe…
TATAA: Programmable Mixed-Precision Transformer Acceleration with a Transformable Arithmetic Architecture
Jiajun Wu, Mo Song, Jingmin Zhao +3
Modern transformer-based deep neural networks present unique technical challenges for effective acceleration in real-world applications. Apart from the vast amount of linear operat…