Research map / Image generation / Text-to-image diffusion
Reconciling Visual Perception and Generation in Diffusion Models
Liulei Li, Yi Yang, Wenguan 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 |
|---|---|---|---|
| add generative visual perception learning (ablation) | GenRep (LDM) · ObjectNet / ADE20K | top-1 / mIoU: 45.4 -> 47.8 top-1; 27.8 -> 30.9 mIoU | Improved |
| add semantic-driven generation learning (ablation) | GenRep (LDM) · ImageNet 256 | FID: 13.27 -> 7.45 | Improved |
| add gradient alignment (ablation) | GenRep (LDM) · ObjectNet / ADE20K / ImageNet | top-1 / mIoU / FID: 49.4 -> 51.1 top-1; 31.5 -> 32.5 mIoU; 7.23 -> 6.92 FID | Improved |
| add generative distillation loss L_gen_distil (ablation) | GenRep (LDM) · ObjectNet | ECE: 0.237 -> 0.208 | Improved |
| thin samples with interval k=2 instead of k=1 (ablation) | GenRep (LDM) · ObjectNet | top-1: 50.3 -> 51.1 | Improved |
| increase thinning interval to k=5 (ablation) | GenRep (LDM) · ObjectNet | top-1: 51.1 -> 48.9 | Hurt |
| increase burn-in to m=125 from m=50 (ablation) | GenRep (LDM) · ObjectNet | top-1: 51.1 -> 47.6 | Hurt |
| add importance re-weighting (linear scaling) of samples (ablation) | GenRep (LDM) · ObjectNet | top-1: 51.1 -> 49.2 | Hurt |
| add both noise adjustment and representation alignment (ablation) | GenRep (LDM) · ImageNet 256 | FID: 7.16 -> 6.92 | Improved |
| symmetric gradient alignment instead of generation-only alignment (ablation) | GenRep (LDM) · ObjectNet / ADE20K / ImageNet | top-1 / mIoU / FID: 48.7 -> 50.1 top-1; 30.3 -> 32.5 mIoU; 6.79 -> 6.92 FID | Mixed |
| use larger thinning interval k=4 with fixed 75 samples (ablation) | GenRep (LDM) · ObjectNet | top-1: 51.1 -> 51.9 | Improved |
| task weights wp=0.8, wg=0.2 instead of 0.7/0.3 (ablation) | GenRep (LDM) · ObjectNet / ImageNet 256 | top-1 / FID: 51.1 -> 51.5 top-1; 6.92 -> 7.12 FID | Mixed |
Closest papers by meaning
- Boosting Generative Image Modeling via Joint Image-Feature Synthesis (NeurIPS 2025 · similarity 0.89)
- Scaling Properties of Diffusion Models For Perceptual Tasks (CVPR 2025 · similarity 0.89)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.89)
- Aligning Generative Denoising with Discriminative Objectives Unleashes Diffusion for Visual Perception (ICLR 2025 · similarity 0.89)
- DICEPTION: A Generalist Diffusion Model for Visual Perceptual Tasks (NeurIPS 2025 · similarity 0.88)
- Dual Diffusion for Unified Image Generation and Understanding (CVPR 2025 · similarity 0.88)
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.
