Research map / Learning theory and algorithms / Prediction and decision losses
Autofocused oracles for model-based design
Clara Fannjiang, Jennifer Listgarten
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 iterative importance-weighted oracle retraining to CbAS | CbAS · Superconductors | median ground-truth expectation (K): 51.5 -> 76.4 | Improved |
| Add iterative importance-weighted oracle retraining to DbAS | DbAS · Superconductors | median ground-truth expectation (K): 57.0 -> 78.9 | Improved |
| Add iterative importance-weighted oracle retraining to RWR | RWR · Superconductors | median ground-truth expectation (K): 43.4 -> 71.4 | Improved |
| Add iterative importance-weighted oracle retraining to FB | FB · Superconductors | median ground-truth expectation (K): 49.2 -> 64.2 | Improved |
| Add iterative importance-weighted oracle retraining to CEM-PI | CEM-PI · Superconductors | median ground-truth expectation (K): 34.5 -> 67.0 | Improved |
| Add iterative importance-weighted oracle retraining to CMA-ES | CMA-ES · Superconductors | median ground-truth expectation (K): 42.1 -> 50.2 | Improved |
| Add iterative importance-weighted oracle retraining to RWR | RWR · Superconductors | Spearman correlation: 0.92 -> 0.65 | Hurt |
| Add iterative importance-weighted oracle retraining to RWR | RWR · Superconductors | RMSE (K): 7.4 -> 12.7 | Hurt |
| Add iterative importance-weighted oracle retraining to DbAS | DbAS · Superconductors | Spearman correlation: 0.01 -> 0.01 | No clear effect |
| Use unflattened importance weights (alpha = 1) instead of flattening weights to alpha = 0.2 during autofocusing (ablation) | Superconductors experiments across MBO methods · Superconductors | design metrics and oracle-ground-truth agreement | Mixed |
| Use autofocusing without importance-weight variance control (ablation) | CMA-ES · Superconductors | design metrics | Mixed |
| Apply autofocusing in the illustrative example across training-distribution and label-noise settings | CbAS · One-dimensional multimodal design problem | ground-truth design objective | Improved |
Closest papers by meaning
- Reliable Algorithm Selection for Machine Learning-Guided Design (ICML 2025 · similarity 0.84)
- Implicitly Guided Design with PropEn: Match your Data to Follow the Gradient (NeurIPS 2024 · similarity 0.83)
- A Dataset for Distilling Knowledge Priors from Literature for Therapeutic Design (NeurIPS 2025 · similarity 0.81)
- Adaptive Experimental Design to Accelerate Scientific Discovery and Engineering Design (AAAI 2025 · similarity 0.81)
- AutoSciLab: A Self-Driving Laboratory for Interpretable Scientific Discovery (AAAI 2025 · similarity 0.81)
- Protein Design with Agent Rosetta: A Case Study for Specialized Scientific Agents (ICML 2026 · similarity 0.81)
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.
