Research map / Probabilistic and scientific modeling / Bayesian optimization
Support-Proximity Augmented Diffusion Estimation for Offline Black-Box Optimization
Yonghan Yang, Ye Yuan, Zipeng Sun, Linfeng Du, Bowei He, Haolun Wu, Can Chen, Xue Liu
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 calibration loss and support-proximity regularization to base diffusion surrogate (ablation) | SPADE diffusion surrogate · Ant Morphology | normalized maximum score: 0.932 -> 0.978 | Improved |
| add calibration loss and proximity regularization (ablation) | SPADE diffusion surrogate · TF Bind 10 | normalized maximum score: 0.870 -> 0.915 | Improved |
| add calibration loss (moment matching and ranking) (ablation) | SPADE with proximity only · Superconductor | normalized maximum score: 0.542 -> 0.546 | Improved |
| add calibration loss (ablation) | SPADE with proximity only · TF Bind 8 | normalized maximum score: 0.897 -> 0.923 | Improved |
| add support-proximity regularization (ablation) | SPADE with calibration only · LLM Data Mixture | normalized maximum score: 0.979 -> 1.019 | Improved |
| add support-proximity regularization (ablation) | SPADE with calibration only · TF Bind 10 | normalized maximum score: 0.895 -> 0.915 | Improved |
| add support-proximity regularization alone to base diffusion surrogate (ablation) | SPADE Base · TF Bind 8 | normalized maximum score: 0.890 -> 0.897 | Improved |
| add calibration loss alone to base diffusion surrogate (ablation) | SPADE Base · Ant Morphology | normalized maximum score: 0.932 -> 0.952 | Improved |
| use diffusion surrogate with proximity regularization instead of best baseline PGS | SPADE · Superconductor | normalized maximum score: 0.563 (PGS) vs 0.546 (SPADE) | Hurt |
| use SPADE instead of ROOT on TF Bind 8 | SPADE · TF Bind 8 | normalized maximum score: 0.986 (ROOT) vs 0.923 (SPADE) | Hurt |
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.
- ROOT — ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge (NeurIPS 2025) · better on 5 of 6 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
- Learning Surrogates for Offline Black-Box Optimization via Gradient Matching (ICML 2024 · similarity 0.90)
- ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge (NeurIPS 2025 · similarity 0.88)
- Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization (ICML 2025 · similarity 0.87)
- Guided Trajectory Generation with Diffusion Models for Offline Model-based Optimization (NeurIPS 2024 · similarity 0.87)
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization (NeurIPS 2026 · similarity 0.87)
- Incorporating Surrogate Gradient Norm to Improve Offline Optimization Techniques (NeurIPS 2024 · similarity 0.87)
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.
