SparkDiffusion: Mitigating the High-Sparsity Trap --- A Unified Framework for up to 265 times Single-GPU Acceleration of Visual Generation

arXiv:2609.23153 · cs.CV · Submitted 2026-09-19 · Read on arXiv

cs.CV

Submitted: 2026-09-19

Updated: 2026-10-08

Code: https://github.com/AlibabaResearch/SparkDiffusion

Project page: https://sparkdiffusion.github.io

Terminology

• Sparsity — 115× in this paper · explained in 4 episodes
• Distillation — 49× in this paper · explained in 2 episodes
• Sparse Attention — 27× in this paper · explained in 5 episodes
• Consistency — 16× in this paper · explained in 3 episodes
• Quantization — 15× in this paper · explained in 12 episodes
• Transformer — 13× in this paper · explained in 2 episodes
• Alignment — 12× in this paper · explained in 1 episode
• Distribution Matching — 11× in this paper · explained in 2 episodes
• Flow Matching — 8× in this paper · explained in 8 episodes
• Diffusion Model — 6× in this paper · explained in 8 episodes
• Diffusion Models — 6× in this paper · explained in 6 episodes
• Fidelity — 6× in this paper · explained in 2 episodes
• Generative Model — 4× in this paper · explained in 2 episodes
• Generative Models — 4× in this paper · explained in 2 episodes
• Ablation — 3× in this paper · explained in 1 episode
• Benchmark — 3× in this paper · explained in 2 episodes
• Core — 2× in this paper · explained in 2 episodes
• Temporal consistency — 2× in this paper · explained in 2 episodes
• Unified Framework — 2× in this paper · explained in 5 episodes
• Activation Checkpointing — 1× in this paper · explained in 1 episode
• Adaptation — 1× in this paper · explained in 1 episode
• CVF International Conference on Computer Vision — 1× in this paper
• Conference on Computer Vision and Pattern Recognition — 1× in this paper
• Failure Modes — 1× in this paper · explained in 3 episodes
• Latent Space — 1× in this paper · explained in 7 episodes
• Semantic Alignment — 1× in this paper · explained in 3 episodes
• Semantic Drift — 1× in this paper · explained in 3 episodes
• World Model — 1× in this paper · explained in 7 episodes

Sources

• V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
• Data-Forcing Distillation: Restoring Diversity and Fidelity in Few-Step Video Generation
• SALAD: Achieve High-Sparsity Attention via Efficient Linear Attention Tuning for Video Diffusion Transformer
• HunyuanVideo: A Systematic Framework For Large Video Generative Models
• Flow Matching for Generative Modeling
• Mixture of Distributions Matters: Dynamic Sparse Attention for Efficient Video Diffusion Transformers
• CrossDistill: Balancing Quality and Diversity via Trajectory-Level Hybrid Few-Step Distillation
• RoPeSLR: 3D RoPE-driven Sparse-LowRank Attention for Efficient Diffusion Transformers
• Parallel Decoding Distillation for Fast Image and Video Generation
• Consistency Models
• Wan: Open and Advanced Large-Scale Video Generative Models
• VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking
• Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity
• SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference
• TurboDiffusion: Accelerating Video Diffusion Models by 100-200 Times
• SLA2: Sparse-Linear Attention with Learnable Routing and QAT
• Fast Video Generation with Sliding Tile Attention
• RoLA: Rotary-Positioned Low-Rank Linear Attention for Efficient Diffusion Transformers
• VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models