Research map / Image generation / Autoregressive and latent generation
DiffSDA: Unsupervised Diffusion Sequential Disentanglement Across Modalities
Hedi Zisling, Ilan Naiman, Nimrod Berman, Supasorn Suwajanakorn, Omri Azencot
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 |
|---|---|---|---|
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · CelebV-HQ conditional swap | AKD: 28.69 -> 6.932 | Improved |
| use DiffSDA instead of SPYL | DiffSDA vs SPYL and DBSE · VoxCeleb conditional swap | AKD: 4.705 -> 2.793 | Improved |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · TaiChi-HD conditional swap | AED: 0.325 -> 0.326 | No clear effect |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · VoxCeleb reconstruction | AED: 1.03 -> 0.37 | Improved |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · TaiChi-HD reconstruction | AKD: 3.83 -> 0.10 | Improved |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · MUG reconstruction | MSE: 0.001 -> 3e-7 | Improved |
| use DiffSDA instead of SPYL | DiffSDA vs SPYL and DBSE · TIMIT speaker identification | Static EER: 3.41% -> 4.43% | Hurt |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · TIMIT speaker identification | Dynamic EER: 34.62% -> 46.72% | Improved |
| use DiffSDA instead of SPYL | DiffSDA vs SPYL and DBSE · LibriSpeech speaker identification | Dynamic EER: 49.76% -> 45.94% | Hurt |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · LibriSpeech speaker identification | Disentanglement gap: 5.58% -> 34.93% | Improved |
| use DiffSDA instead of a supervised model | DiffSDA vs SPYL and DBSE · ETTh1 oil temperature | MAE: 10.19 -> 9.89 | Improved |
| use DiffSDA instead of DBSE | DiffSDA vs SPYL and DBSE · Air Quality classification | accuracy: 65.9 -> 69.2 | Improved |
Closest papers by meaning
- Can Diffusion Models Disentangle? A Theoretical Perspective (NeurIPS 2025 · similarity 0.90)
- Disentanglement Beyond Static vs. Dynamic: A Benchmark and Evaluation Framework for Multi-Factor Sequential Representations (NeurIPS 2025 · similarity 0.89)
- Factorized Diffusion Autoencoder for Unsupervised Disentangled Representation Learning (AAAI 2024 · similarity 0.89)
- Diffusion Model with Cross Attention as an Inductive Bias for Disentanglement (NeurIPS 2024 · similarity 0.89)
- Simplified and Generalized Masked Diffusion for Discrete Data (NeurIPS 2024 · similarity 0.88)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · 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.
