26 papers
StageBreaker: Shifting Video RL Post-Training Scheduling from Stages down to Data Dependencies
Li Peng, Chenghuan Huang, Ziyang Ma +3
Reinforcement learning (RL) post-training has emerged as the key paradigm for improving autoregressive video models. Emerging vision-language reward models further strengthen post-…
EchoCache: Energy-Guided Cross-Modal Caching for Efficient Audio-Driven Video Generation
Jiayu Chen, Xiaoyu Wu, Rongshan Gao +6
Audio-driven video generation (A2V) has achieved promising progress in synthesizing temporally coherent and audio-visually aligned videos, yet its inference remains expensive due t…
When and Where to Look: Adaptive Visual Evidence Scheduling for Efficient Long Video Understanding
Ke Li, Jiayu Chen, Maoliang Li +5
Efficient long-video understanding requires vision--language models (VLMs) to reason over a small number of frames selected as sparse visual evidence. Existing relevance-based meth…
Token Radius Attention for Efficient Video Generation
Jiayu Chen, Zhikun Jiang, Maoliang Li +6
Video Diffusion Transformers (VDiTs) enable high-fidelity generation but incur quadratic cost from dense 3D self-attention. Existing head- and block-level sparse methods share comp…
EcoVideo: Entropy-Orchestrated Video Generation Paradigm in Cloud-Edge Dynamics
Jiayu Chen, Hengyi Zhang, Maoliang Li +5
DiT video generation is latency-intensive due to iterative full-frame denoising, while prior cloud-edge methods largely rely on static inter-step decoupling and cannot leverage int…
MoECa: Aligning Feature Reuse with Expert Decomposition in Diffusion Transformers
Maoliang Li, Haojing Chen, Jiayu Chen +4
Diffusion Transformers with Mixture-of-Experts (DiT-MoE) improve model capacity under sparse activation, but diffusion inference is still bottlenecked by redundant computation acro…
Pyramid Forcing: Head-Aware Pyramid KV Cache Policy for High-Quality Long Video Generation
Jiayu Chen, Junbei Tang, Wenbiao Zhao +6
Autoregressive video generation enables streaming and open-ended long video synthesis, but still suffers from long-term degradation caused by accumulated errors. Existing KVCache s…
Interactive State Space Model with Cross-Modal Local Scanning for Depth Super-Resolution
Chen Wu, Ling Wang, Zhuoran Zheng +4
Guided depth super-resolution (GDSR) reconstructs HR depth maps from LR inputs with HR RGB guidance. Existing methods either model each modality independently or rely on computatio…
Beyond Spatial Compression: Interface-Centric Generative States for Open-World 3D Structure
Xiang Chen, Alexander Binder
Current 3D tokenizers largely treat representation as spatial compression: compact codes reconstruct surface geometry, but leave component ownership and attachment validity implici…
NTIRE 2026 3D Restoration and Reconstruction in Real-world Adverse Conditions: RealX3D Challenge Results
Shuhong Liu, Chenyu Bao, Ziteng Cui +103
This paper presents a comprehensive review of the NTIRE 2026 3D Restoration and Reconstruction (3DRR) Challenge, detailing the proposed methods and results. The challenge seeks to…
RoboECC: Multi-Factor-Aware Edge-Cloud Collaborative Deployment for VLA Models
Zihao Zheng, Hangyu Cao, Jiayu Chen +6
Vision-Language-Action (VLA) models are mainstream in embodied intelligence but face high inference costs. Edge-Cloud Collaborative (ECC) deployment offers an effective fix by easi…
HeiSD: Hybrid Speculative Decoding for Embodied Vision-Language-Action Models with Kinematic Awareness
Zihao Zheng, Zhihao Mao, Sicheng Tian +8
Vision-Language-Action (VLA) Models have become the mainstream solution for robot control, but suffer from slow inference speeds. Speculative Decoding (SD) is a promising accelerat…
KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models
Zihao Zheng, Zhihao Mao, Maoliang Li +6
Vision-Language-Action (VLA) models build a token-domain robot control paradigm, yet suffer from low speed. Speculative Decoding (SD) is an optimization strategy that can boost inf…
DIRECT: Video Mashup Creation via Hierarchical Multi-Agent Planning and Intent-Guided Editing
Ke Li, Maoliang Li, Jialiang Chen +4
Video mashup creation represents a complex video editing paradigm that recomposes existing footage to craft engaging audio-visual experiences, demanding intricate orchestration acr…
DyQ-VLA: Temporal-Dynamic-Aware Quantization for Embodied Vision-Language-Action Models
Zihao Zheng, Hangyu Cao, Sicheng Tian +12
Vision-language-action (VLA) models achieve favorable task performance, yet runtime errors in closed-loop execution evolve with alternating updates of actions and observations. Exi…
ActECC: Action Chunk Temporal Harnessing for Edge-Cloud Collaborative VLA Inference System
Zihao Zheng, Sicheng Tian, Zhihao Mao +10
VLA models have emerged as the mainstream for embodied intelligence, which decouples reasoning and execution via action chunks. To enable real-time deployment of VLA models, existi…
ToProVAR: Efficient Visual Autoregressive Modeling via Tri-Dimensional Entropy-Aware Semantic Analysis and Sparsity Optimization
Jiayu Chen, Ruoyu Lin, Zihao Zheng +4
Visual Autoregressive(VAR) models enhance generation quality but face a critical efficiency bottleneck in later stages. In this paper, we present a novel optimization framework for…
HeRo: Adaptive Orchestration of Agentic RAG on Heterogeneous Mobile SoC
Maoliang Li, Jiayu Chen, Zihao Zheng +5
With the increasing computational capability of mobile devices, deploying agentic retrieval-augmented generation (RAG) locally on heterogeneous System-on-Chips (SoCs) has become a…
STaRR: Spatial-Temporal Token-Dynamics-Aware Responsive Remasking for Diffusion Language Models
Xinhao Sun, Huaijin Zhao, Maoliang Li +4
Diffusion Language Models (DLMs) enable parallel decoding via iterative denoising, where remasking strategies play a critical role in balancing inference speed and output quality.…
MIRAGE: Runtime Scheduling for Multi-Vector Image Retrieval with Hierarchical Decomposition
Maoliang Li, Ke Li, Yaoyang Liu +5
To effectively leverage user-specific data, retrieval augmented generation (RAG) is employed in multimodal large language model (MLLM) applications. However, conventional retrieval…
Agent.xpu: Efficient Scheduling of Agentic LLM Workloads on Heterogeneous SoC
Xinming Wei, Jiahao Zhang, Haoran Li +6
Personal LLM agents increasingly combine foreground reactive interactions with background proactive monitoring, forming long-lived, stateful LLM flows that interleave prefill and t…
FedHQ: Hybrid Runtime Quantization for Federated Learning
Zihao Zheng, Ziyao Wang, Xiuping Cui +6
Federated Learning (FL) is a decentralized model training approach that preserves data privacy but struggles with low efficiency. Quantization, a powerful training optimization tec…
EaqVLA: Encoding-aligned Quantization for Vision-Language-Action Models
Feng Jiang, Zihao Zheng, Xiuping Cui +3
With the development of Embodied Artificial intelligence, the end-to-end control policy such as Vision-Language-Action (VLA) model has become the mainstream. Existing VLA models fa…
DynaMo: Runtime Switchable Quantization for MoE with Cross-Dataset Adaptation
Zihao Zheng, Xiuping Cui, Size Zheng +4
As the Mix-of-Experts (MoE) architecture increases the number of parameters in large models, there is an even greater need for model quantization. However, existing quantization me…