Research map / Probabilistic and scientific modeling / Bayesian optimization
Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization
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 |
|---|---|---|---|
| Use normalized designs and objective values for gradient ascent (ablation) | Gradient Ascent · Hopper Controller | performance: 1.4x | Helped |
| Add input and output normalization and dimension-scaled learning rate | Gradient Ascent · Design-Bench tasks | 100th percentile performance | Helped |
| Use the minimum prediction from an ensemble instead of a single objective model (ablation) | Gradient Ascent · ChEMBL | normalized 100th percentile performance: 0.307 ± 0.308 -> 0.653 ± 0.024 | Helped |
| Use the mean prediction from an ensemble instead of a single objective model (ablation) | Gradient Ascent · Hopper Controller | normalized 100th percentile performance: 1.035 ± 0.482 -> 1.586 ± 0.454 | Helped |
| Use ensemble minimum prediction rather than single-model prediction (ablation) | Gradient Ascent · TF Bind 10 | normalized 100th percentile performance: 0.657 ± 0.039 -> 0.649 ± 0.032 | Hurt |
| Use ensemble mean prediction rather than single-model prediction (ablation) | Gradient Ascent · Ant Morphology | normalized 100th percentile performance: 0.293 ± 0.023 -> 0.445 ± 0.080 | Helped |
| Use a generative-model offline MBO method rather than a simple CMA-ES baseline | Offline MBO methods · Design-Bench tasks | 100th percentile performance | Mixed |
| Use an offline task with distinguishable method performance | Design-Bench benchmark task selection · GFP and UTR | normalized 100th percentile performance | No clear effect |
Closest papers by meaning
- Learning Surrogates for Offline Black-Box Optimization via Gradient Matching (ICML 2024 · similarity 0.88)
- Offline Model-Based Optimization by Learning to Rank (ICLR 2025 · similarity 0.88)
- A survey and benchmark of high-dimensional Bayesian optimization of discrete sequences (NeurIPS 2024 · similarity 0.88)
- Robust Model-Based Optimization for Challenging Fitness Landscapes (ICLR 2024 · similarity 0.87)
- Offline Model-Based Optimization via Policy-Guided Gradient Search (AAAI 2024 · similarity 0.87)
- Empower Structure-Based Molecule Optimization with Gradient Guided Bayesian Flow Networks (ICML 2025 · 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.