Research map / Video / Video generation
Model Already Knows the Best Noise: Bayesian Active Noise Selection via Attention in Video Diffusion Model
Kwanyoung Kim, Sanghyun Kim
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 |
|---|---|---|---|
| add ANSE BANSA noise selection | AnimateDiff · VBench | total score: 77.98 -> 79.33 | Improved |
| add ANSE noise selection | CogVideoX-2B · VBench | total score: 81.03 -> 81.66 | Improved |
| add ANSE noise selection | CogVideoX-5B · VBench | total score: 81.52 -> 81.71 | Improved |
| add ANSE noise selection | HunyuanVideo · VBench aesthetic quality | aesthetic quality: 0.6276 -> 0.6268 | No clear effect |
| combine ANSE with FreqPrior | FreqPrior · VBench (AnimateDiff) | total score: 79.07 -> 80.43 | Improved |
| add ANSE noise selection | HunyuanVideo and Wan2.1 · MSR-VTT | FVMD: 17641.19 -> 16491.68 | Improved |
| select seed by BANSA (B) instead of random (ablation) | CogVideoX-2B · VBench | total score: 81.03 -> 81.66 | Improved |
| replace dropout with Bernoulli masking for BANSA (ablation) | BANSA · VBench | total score: 81.33 -> 81.66 | Improved |
| use entropy acquisition instead of random (ablation) | CogVideoX-2B · VBench | total score: 81.03 -> 81.13 | No clear effect |
| select highest BANSA seed (reverse criterion) (ablation) | ANSE · VBench | quality score: 82.08 -> 81.93 | Hurt |
| increase ensemble size K from 1 to 10 (ablation) | ANSE · VBench subject consistency | subject consistency: 0.9618 -> 0.9641 | Improved |
| add Bernoulli masking p=0.2 vs none (ablation) | ANSE · VBench subject consistency | subject consistency: 0.9254 -> 0.9310 | Improved |
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.
- FreqPrior — FreqPrior: Improving Video Diffusion Models with Frequency Filtering Gaussian Noise (ICLR 2025) · better on 1 of 1 shared result
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
- FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling (ICLR 2024 · similarity 0.87)
- FreqPrior: Improving Video Diffusion Models with Frequency Filtering Gaussian Noise (ICLR 2025 · similarity 0.87)
- How I Warped Your Noise: a Temporally-Correlated Noise Prior for Diffusion Models (ICLR 2024 · similarity 0.87)
- VideoMLA: Low-Rank Latent KV Cache for Minute-Scale Autoregressive Video Diffusion (NeurIPS 2026 · similarity 0.86)
- Information Theoretic Learning for Diffusion Models with Warm Start (NeurIPS 2025 · similarity 0.86)
- Diffusion Models With Learned Adaptive Noise (NeurIPS 2024 · similarity 0.85)
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.
