Research map / Probabilistic and scientific modeling / Bayesian optimization
Generative Bayesian Optimization: Generative Models as Acquisition Functions
Rafael Oliveira, Daniel M. Steinberg, Edwin V. Bonilla
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 |
|---|---|---|---|
| train generative model directly on utilities instead of fitting an intermediate surrogate classifier (GenBO vs VSD) | GenBO · ALOHA text optimization | run time: on average 3 times faster than VSD | Improved |
| use all collected data for retraining instead of only the last batch (GenBO/VSD vs CbAS) | GenBO · ALOHA text optimization | simple regret | Improved |
| use robust preference loss with EI utility | GenBO · ALOHA text optimization | speed of improvement | Improved |
| use a pre-trained informative prior p0 instead of no prior (ablation) | GenBO · ALOHA text optimization | simple regret | No clear effect |
| vary the threshold annealing schedule, final percentile above 90% for GenBO (ablation) | GenBO · ALOHA text optimization | simple regret | No clear effect |
| use balanced forward KL loss on longest sequences (ablation) | GenBO · Ehrlich functions (M=64) | simple regret | Improved |
| use exponential regularization (exponential dot-product kernel) (ablation) | GenBO · Ehrlich functions (high dimension) | simple regret | Improved |
| increase batch size B (ablation) | GenBO · Ehrlich functions | simple regret | Improved |
| use GenBO instead of VSD/CbAS on protein stability (mRouge, M=228) | GenBO · FoldX stability | stability | Hurt |
| use GenBO variants instead of baselines on SASA task | GenBO · FoldX SASA | SASA | Improved |
Closest papers by meaning
- Batched Energy-Entropy acquisition for Bayesian Optimization (NeurIPS 2024 · similarity 0.88)
- Amortized Active Generation of Pareto Sets (NeurIPS 2025 · similarity 0.87)
- Understanding the Challenges in Iterative Generative Optimization with LLMs (NeurIPS 2026 · similarity 0.87)
- Latent Bayesian Optimization via Autoregressive Normalizing Flows (ICLR 2025 · similarity 0.87)
- BONSAI: Bayesian Optimization with Natural Simplicity and Interpretability (NeurIPS 2026 · similarity 0.87)
- Principled Preferential Bayesian Optimization (ICML 2024 · similarity 0.86)
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.
