Research map / Image generation / Autoregressive and latent generation
Cooperative Training of Fast Thinking Initializer and Slow Thinking Solver for Conditional Learning
Jianwen Xie, Zilong Zheng, Xiaolin Fang, Song Chun Zhu, Ying Nian Wu
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 |
|---|---|---|---|
| add solver Langevin refinement on top of initializer output (ablation) | CCoopNets · CIFAR-10 | Inception score: 6.63 -> 7.30 | Improved |
| use more Langevin refinement steps (l=1 or 8 vs 16+) (ablation) | CCoopNets · fashion-MNIST | training stability | Improved |
| increase latent dimension d of the initializer (ablation) | CCoopNets · fashion-MNIST | FID | Improved |
| increase Langevin steps l from 1 to 64 (d=8) (ablation) | CCoopNets · fashion-MNIST | time per epoch (s): 8.98 -> 86.93 | Hurt |
| increase latent dimension d from 8 to 128 (l=16) (ablation) | CCoopNets · fashion-MNIST | time per epoch (s): 26.88 -> 27.76 | No clear effect |
| use a larger Langevin step size (ablation) | CCoopNets · fashion-MNIST | learning speed | Improved |
| replace naive concatenation initializer with U-Net with dropout noise (ablation) | CCoopNets initializer · image-to-image translation | noise usage | Improved |
| feed ground truth solutions and add l1 loss to the initializer update (ablation) | CCoopNets · CMP Facade | visual quality | Improved |
| CCoopNets vs pix2pix training cost | CCoopNets · CMP Facade | time per epoch (s): 30.9 -> 32.7 | Hurt |
| CCoopNets vs pix2pix | pix2pix · CMP Facades inpainting | PSNR: 19.34 -> 20.47 | Improved |
| CCoopNets vs pix2pix | pix2pix · Paris streetview inpainting | PSNR: 15.17 -> 21.17 | Improved |
| CCoopNets vs pix2pix in human preference | pix2pix · CMP Facade | preference ratio: 0.280 -> 0.720 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CVAE-GAN — CVAE-GAN: Fine-Grained Image Generation Through Asymmetric Training (ICCV 2017) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Energy-Based Transformers are Scalable Learners and Thinkers (ICLR 2026 · similarity 0.85)
- A Simple Approach to Unifying Diffusion-based Conditional Generation (ICLR 2025 · similarity 0.84)
- Manifold Preserving Guided Diffusion (ICLR 2024 · similarity 0.84)
- Random Conditioning for Diffusion Model Compression with Distillation (CVPR 2025 · similarity 0.83)
- Revisiting Diffusion Models: From Generative Pre-training to One-Step Generation (ICML 2025 · similarity 0.83)
- Improved Distribution Matching Distillation for Fast Image Synthesis (NeurIPS 2024 · similarity 0.83)
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.
