Research map / Image generation / Autoregressive and latent generation
Multi-Label Conditional Generation From Pre-Trained Models
Magdalena Proszewska, Maciej Wołczyk, Maciej Zięba, Patryk Wielopolski, Łukasz Maziarka, Marek Śmieja
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 PluGeN instead of StyleFlow | StyleGAN · FFHQ attribute modification | average accuracy: 0.86 -> 0.90 | Improved |
| use PluGeN instead of StyleFlow | StyleGAN · FFHQ attribute modification | FID: 32.59 -> 46.51 | Hurt |
| use PluGeN instead of cFlow | VAE · CelebA attribute manipulation | average FID: 39.68 -> 28.07 | Improved |
| use PluGeN instead of MSP | VAE · CelebA attribute manipulation | average accuracy: 0.70 -> 0.69 | No clear effect |
| use PluGeN instead of MSP for rare combination female x beard | VAE · CelebA attribute manipulation | accuracy: 0.33 -> 0.59 | Improved |
| use PluGeN instead of cFlow | VAE · CelebA generation, independently sampled attributes | F1: 0.29 -> 0.44 | Improved |
| use PluGeN instead of cVAE | VAE · CelebA generation, test-set attribute combinations | F1: 0.59 -> 0.69 | Improved |
| replace NICE with continuous normalizing flow (ablation) | PluGeN · CelebA generation, test-set attribute combinations | FID: 72.72 -> 68.96 | Improved |
| replace NICE with continuous normalizing flow (ablation) | PluGeN · CelebA generation, test-set attribute combinations | F1: 0.69 -> 0.63 | Hurt |
| replace NICE with continuous normalizing flow (ablation) | PluGeN · CelebA generation, independently sampled attributes | F1: 0.44 -> 0.41 | Hurt |
| increase beta-VAE beta from 1 to 16 (ablation) | PluGeN · CelebA beta-VAE backbone | FID: 55.11 -> 110.46 | Hurt |
| lower beta-VAE beta from 1 to 0.5 (ablation) | PluGeN · CelebA beta-VAE backbone | F1: 0.66 -> 0.45 | Hurt |
Closest papers by meaning
- PluGeN: Multi-Label Conditional Generation from Pre-trained Models (AAAI 2022 · similarity 0.99)
- Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space (CVPR 2017 · similarity 0.88)
- Molecule Design by Latent Prompt Transformer (NeurIPS 2024 · similarity 0.85)
- Discovering Latent Graphs with GFlowNets for Diverse Conditional Image Generation (NeurIPS 2025 · similarity 0.84)
- Linear combinations of latents in generative models: subspaces and beyond (ICLR 2025 · similarity 0.84)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.84)
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.
