Research map / Image generation / Fast diffusion sampling
DAPS++: Rethinking Diffusion Inverse Problems with Decoupled Posterior Annealing
Hao Chen, Renzheng Zhang, Scott Howard
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 |
|---|---|---|---|
| remove prior gradient from DAPS refinement (ablation) | DAPS-100 · FFHQ inpainting | SSIM / LPIPS: 0.716 / 0.194 -> 0.716 / 0.194 | No clear effect |
| use Tweedie initialization in DAPS (ablation) | DAPS-100 · FFHQ inpainting | SSIM / LPIPS: LPIPS 0.194 -> 0.186; time 7.4 -> 10.1 s | Mixed |
| full decoupling: Tweedie/RK4 init plus likelihood-only refinement (ablation) | DAPS-100 · FFHQ inpainting | SSIM / LPIPS: 0.716 / 0.194 -> 0.812 / 0.141; 7.4 -> 2.4 s | Improved |
| use Tweedie/RK4 init without removing prior gradient (ablation) | DAPS · FFHQ inpainting | LPIPS: 0.194 -> 0.192 | No clear effect |
| replace DAPS-1K with DAPS++-50 under heavy noise | DAPS-1K · FFHQ Gaussian blur gamma=0.4 | LPIPS: 0.633 -> 0.303 | Improved |
| DAPS++ vs DAPS-1K sampling cost | DAPS-1K · FFHQ-256 SR | seconds per image: 38.10 -> 2.31 | Improved |
| use negative polynomial exponent rho=-7 instead of rho=2 (ablation) | DAPS++ · FFHQ Gaussian blur | LPIPS: 0.233 -> 0.171 | Improved |
| increase sigma-bar threshold from 0.5 to 5.0 (ablation) | DAPS++ · FFHQ tasks | NFE / quality: NFE 100 -> 152; HDR LPIPS 0.171 -> 0.210 | Hurt |
| DAPS++-50 vs DAPS-1K on ImageNet nonlinear blur | DAPS-1K · ImageNet nonlinear blur | LPIPS: 0.212 vs 0.211 (DAPS++-100) | No clear effect |
| DAPS++ vs DAPS-1K on FFHQ nonlinear blur | DAPS-1K · FFHQ nonlinear blur | LPIPS: 0.191 -> 0.194 | Hurt |
Closest papers by meaning
- Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · similarity 0.93)
- Inference-Time Search Using Side Information for Diffusion-Based Image Reconstruction (NeurIPS 2026 · similarity 0.90)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.90)
- Improving Diffusion Posterior Samplers with Lagged Temporal Corrections for Image Restoration (NeurIPS 2026 · similarity 0.90)
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.90)
- Analyzing and Guiding Zero-Shot Posterior Sampling in Diffusion Models (NeurIPS 2026 · similarity 0.90)
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.
