Research map / Label-efficient and robust learning / Domain adaptation and generalization
AlignFlow: Cycle Consistent Learning from Multiple Domains via Normalizing Flows
Aditya Grover, Christopher Chute, Rui Shu, Zhangjie Cao, Stefano Ermon
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 |
|---|---|---|---|
| combine adversarial and MLE training | AlignFlow · Facades | MSE: 0.7129 -> 0.5801; 0.3286 -> 0.2512 | Improved |
| combine adversarial and MLE training | AlignFlow · Maps | MSE: 0.0245 -> 0.0209; 0.0953 -> 0.0897 | Improved |
| combine adversarial and MLE training | AlignFlow · CityScapes | MSE: 0.1252 -> 0.1130; 0.1200 -> 0.1462 | Mixed |
| add the MLE objective to adversarial training (ablation) | AlignFlow · Facades | MSE: 0.6727 -> 0.5801; 0.2679 -> 0.2512 | Improved |
| add adversarial training to MLE training (ablation) | AlignFlow · Facades | MSE: 0.9014 -> 0.5801; 0.5960 -> 0.2512 | Improved |
| add the MLE objective to adversarial training (ablation) | AlignFlow · Maps | MSE: 0.0385 -> 0.0209; 0.1123 -> 0.0897 | Improved |
| add adversarial training to MLE training (ablation) | AlignFlow · Maps | MSE: 0.0452 -> 0.0209; 0.1746 -> 0.0897 | Improved |
| add the MLE objective to adversarial training (ablation) | AlignFlow · CityScapes | MSE: 0.2569 -> 0.1130; 0.2196 -> 0.1462 | Improved |
| add adversarial training to MLE training (ablation) | AlignFlow · CityScapes | MSE: 0.2526 -> 0.1130; 0.2272 -> 0.1462 | Improved |
| replace CycleGAN translation with AlignFlow | CyCADA adaptation pipeline · MNIST→USPS | classification accuracy: 95.6 ± 0.2 -> 96.2 ± 0.2 | Improved |
| replace CycleGAN translation with AlignFlow | CyCADA adaptation pipeline · USPS→MNIST | classification accuracy: 96.5 ± 0.1 -> 96.7 ± 0.1 | Improved |
| replace CycleGAN translation with AlignFlow | CyCADA adaptation pipeline · SVHN→MNIST | classification accuracy: 90.4 ± 0.4 -> 91.0 ± 0.3 | Improved |
Closest papers by meaning
- Flowing Backwards: Improving Normalizing Flows via Reverse Representation Alignment (AAAI 2026 · similarity 0.89)
- Align Your Flow: Scaling Continuous-Time Flow Map Distillation (NeurIPS 2025 · similarity 0.87)
- Boosting Alignment for Post-Unlearning Text-to-Image Generative Models (NeurIPS 2024 · similarity 0.86)
- Improved Techniques for Training Consistency Models (ICLR 2024 · similarity 0.86)
- How to build a consistency model: Learning flow maps via self-distillation (NeurIPS 2025 · similarity 0.86)
- Aligning Few-Step Generative Model via Amortizing Sample-Based Variational Inference (NeurIPS 2026 · 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.
