Research map / Image generation / Autoregressive and latent generation
ProReGen: Progressive Residual Generation under Attribute Correlations
Ruby Shrestha, Ajay Gopi, Casey Meisenzahl, Bipin Lekhak, Linwei Wang
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 |
|---|---|---|---|
| train g_mjr then g_res in two stages instead of simultaneously (ablation) | ProReGen-GAN · Colored-MNIST (95% correlation) | minority correctness: 0.3557 -> 0.9256 | Improved |
| two-stage training instead of single-stage (ablation) | ProReGen-GAN · Colored-MNIST (95% correlation) | minority FID: 65.0227 -> 17.2562 | Improved |
| two-stage training instead of single-stage (ablation) | ProReGen-GAN · Colored-MNIST (95% correlation) | majority correctness: 0.9289 -> 0.9592 | No clear effect |
| perturb x1 to simulate error in m(x1) (80% of samples) (ablation) | ProReGen-GAN · Colored-MNIST (95% correlation) | overall correctness: 0.9256 -> 0.8589 | Hurt |
| use one conv block instead of two in the residual sub-network (ablation) | ProReGen-GAN · Colored-MNIST (95% correlation) | minority correctness: 0.9256 -> 0.6476 | Hurt |
| invert causal direction from digit->color to color->digit (ablation) | ProReGen-GAN · Colored-MNIST | minority correctness: 0.9396 -> 0.0811 | Hurt |
| add pseudo-supervision (causal-cHVAE) to c-VAE | c-VAE · Colored-MNIST | minority vs majority correctness | Mixed |
| re-weight (upsample) minority samples | c-VAE, c-GAN, c-DM · Colored-MNIST | correctness and coverage | Mixed |
| add ProReGen to naive c-GAN/c-DM | c-GAN, c-DM · Corrupted-CIFAR10 | minority correctness | Improved |
| add pseudo-supervision (causal-GAN) to c-GAN | c-GAN · MNIST-Correlation | minority correctness | No clear effect |
Closest papers by meaning
- Pretrained Reversible Generation as Unsupervised Visual Representation Learning (ICCV 2025 · similarity 0.86)
- Efficient Generative Modeling with Residual Vector Quantization-Based Tokens (ICML 2025 · similarity 0.86)
- Counterfactual Generative Networks (ICLR 2021 · similarity 0.86)
- Machine Unlearning for Image-to-Image Generative Models (ICLR 2024 · similarity 0.85)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (NeurIPS 2024 · similarity 0.85)
- Plug-and-Play Context Feature Reuse for Efficient Masked Generation (NeurIPS 2025 · 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.
