Research map / Video / Video generation
HiAR: Efficient Autoregressive Long Video Generation via Hierarchical Denoising
Kai Zou, Dian Zheng, Hongbo Liu, Tiankai Hang, Bin Liu, Nenghai Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace fully denoised context (t_c=0) with context at output noise level t_{j+1} via hierarchical denoising | HiAR · VBench (20s videos) | Drift score: 0.355 -> 0.257 | Improved |
| use hierarchical denoising with pipelined parallel inference | HiAR · VBench (20s videos) | Throughput (fps): 17 -> 30 | Improved |
| use hierarchical denoising with pipelined parallel inference | HiAR · VBench (20s videos) | Latency (s per chunk): 0.69 -> 0.30 | Improved |
| hierarchical denoising and forward-KL regulariser over Self-Forcing | HiAR · VBench (20s videos) | Total score: 0.805 -> 0.821 | Improved |
| use input-level context t_c=t_j instead of t_{j+1} (ablation) | HiAR · Context noise ablation (20s) | Quality score: 0.846 -> 0.799 | Hurt |
| use input-level context t_c=t_j instead of t_{j+1} (ablation) | HiAR · Context noise ablation (20s) | Drift score: 0.257 -> 0.184 | Improved |
| use t_c=t_{j+1} instead of t_c=0 context (ablation) | HiAR · Context noise ablation (20s) | Quality score: 0.829 -> 0.846 | Improved |
| apply forward-KL regulariser in causal mode instead of bidirectional attention (ablation) | HiAR · Forward-KL ablation (20s) | Dynamic score: 0.686 -> 0.625 | Hurt |
| increase constrained steps K from 1 to 4 (ablation) | HiAR · Forward-KL ablation (20s) | Drift score: 0.257 -> 0.306 | Hurt |
| increase constrained steps K from 1 to 2 (ablation) | HiAR · Forward-KL ablation (20s) | Dynamic score: 0.686 -> 0.693 | No clear effect |
| add forward-KL regulariser (ablation) | HiAR · Forward-KL ablation (20s) | Dynamic score: 0.445 -> 0.686 | Improved |
| add forward-KL regulariser (ablation) | HiAR · Forward-KL ablation (20s) | Drift score: 0.218 -> 0.257 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Causal Forcing — Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation (ICML 2026) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Equilibrium Forcing: Adaptive Video Generation Without Noise Conditioning (NeurIPS 2026 · similarity 0.90)
- FIFO-Diffusion: Generating Infinite Videos from Text without Training (NeurIPS 2024 · similarity 0.89)
- Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion (NeurIPS 2025 · similarity 0.89)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · similarity 0.89)
- Denoising Autoregressive Transformers for Scalable Text-to-Image Generation (ICLR 2025 · similarity 0.88)
- Spectral Progressive Diffusion for Efficient Image and Video Generation (NeurIPS 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
