Research map / Probabilistic and scientific modeling / Variational and generative sampling
Score-Guided Intermediate Level Optimization: Fast Langevin Mixing for Inverse Problems
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace ILO's sparsity prior with a learned score-based prior in an intermediate layer | StyleGAN-2 inverse-problem reconstruction · Denoising, compressed sensing, and inpainting | MSE and perceptual distance | Helped |
| Replace ILO's sparsity prior with a learned score-based prior in an intermediate layer | StyleGAN-2 inverse-problem reconstruction · Denoising with noise standard deviation σ=0.1 | MSE and perceptual distance | No clear effect |
| Use a learned score-based prior for posterior sampling instead of ILO | StyleGAN-2 reconstruction · FFHQ downsampled × 256 | CLIP gender prediction correct: 78/128 -> 89/128 | Helped |
| Apply the learned score prior to an early-stopped ILO intermediate representation | StyleGAN-2 third-layer inversion · Out-of-distribution painting projection | visual naturalness | Helped |
| Run score-based sampling in an intermediate latent space rather than pixel space | SGILO · Image generation | sampling speed: ~10× faster | Helped |
| Train the score prior on deeper intermediate layers (ablation) | StyleGAN-2 score-based prior · Inverse problems | reconstruction power | Helped |
| Use the score prior for intermediate-layer sampling rather than only using it as an inverse-problem regularizer (ablation) | SGILO · Inverse-problem sampling | function evaluations: 300 function evaluations | Mixed |
Closest papers by meaning
- Intermediate Layer Optimization for Inverse Problems using Deep Generative Models (ICML 2021 · similarity 0.89)
- Score-Based Generative Modeling with Critically-Damped Langevin Diffusion (ICLR 2022 · similarity 0.88)
- Preconditioned Langevin Dynamics with Score-based Generative Models for Infinite-Dimensional Linear Bayesian Inverse Problems (NeurIPS 2025 · similarity 0.87)
- Consistency Posterior Sampling for Diverse Image Synthesis (CVPR 2025 · similarity 0.86)
- Sample-specific Noise Injection for Diffusion-based Adversarial Purification (ICML 2025 · similarity 0.85)
- Sample as you Infer: Predictive Coding with Langevin Dynamics (ICML 2024 · 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.