Research map / Image generation / Autoregressive and latent generation
Multi-objective Deep Data Generation with Correlated Property Control
Shiyu Wang, Xiaojie Guo, Xuanyang Lin, Bo Pan, Yuanqi Du, Yinkai Wang, Yanfang Ye, Ashley Ann Petersen, Austin Leitgeb, Saleh AlKhalifa, Kevin Minbiole, William Wuest and 2 more
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 CorrVAE instead of Bayesian optimization on all latent w | CorrVAE · dSprites size | MSE: 0.0033 -> 0.0016 | Improved |
| use CorrVAE instead of Bayesian optimization | CorrVAE · Pendulum light position | MSE: 19.2387 -> 15.3900 | Improved |
| use CorrVAE instead of Bayesian optimization | CorrVAE · Pendulum shadow position | MSE: 17.2858 -> 6.0250 | Improved |
| use CorrVAE instead of Bayesian optimization | CorrVAE · dSprites x+y position | MSE: 0.0062 vs 0.0066 | No clear effect |
| use ground-truth mask (CorrVAE-1) instead of learned mask (ablation) | CorrVAE · Pendulum light position | MSE: 15.3900 -> 11.2878 | Improved |
| use ground-truth mask (CorrVAE-1) instead of learned mask (ablation) | CorrVAE · dSprites x+y position | MSE: 0.0066 -> 0.0023 | Improved |
| use ground-truth mask (CorrVAE-1) instead of learned mask (ablation) | CorrVAE · Pendulum shadow length | MSE: 10.2600 -> 2.4626 | Hurt |
| use CorrVAE instead of CSVAE | CorrVAE · dSprites x+y position | MSE: 0.3563 -> 0.0066 | Improved |
| use CorrVAE instead of CSVAE | CorrVAE · dSprites size | MSE: 0.0006 -> 0.0016 | Hurt |
| use CorrVAE instead of Semi-VAE | CorrVAE · Pendulum pendulum angle | MSE: 9.9455 -> 36.37 | Hurt |
| use CorrVAE instead of PCVAE | CorrVAE · QAC and QM9 | uniqueness: 89.2% -> 44.5% | Hurt |
| use CorrVAE instead of Semi-VAE | CorrVAE · QAC | MSE of MolWeight: 433447.6 -> 356701.5 | Improved |
Closest papers by meaning
- Property Controllable Variational Autoencoder via Invertible Mutual Dependence (ICLR 2021 · similarity 0.87)
- Multi-View Data Generation Without View Supervision (ICLR 2018 · similarity 0.86)
- Controllable and Compositional Generation with Latent-Space Energy-Based Models (NeurIPS 2021 · similarity 0.86)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · similarity 0.85)
- Efficient Generative Modeling with Residual Vector Quantization-Based Tokens (ICML 2025 · similarity 0.84)
- Compositional Generalization via Forced Rendering of Disentangled Latents (ICML 2025 · 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.
