Research map / Probabilistic and scientific modeling / Bayesian optimization
Gradient-Based Optimization for Bayesian Preference Elicitation
Ivan Vendrov, Tyler Lu, Qingqing Huang, Craig Boutilier
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 |
|---|---|---|---|
| Use continuous gradient-based query optimization with alternating deep retrieval | Bayesian preference elicitation · Synthetic pairwise comparisons | queries to reach regret 1.0: 7 -> 5 | Improved |
| Use continuous gradient-based EVOI optimization | Bayesian preference elicitation · Goodreads, first query | EVOI: 0.089 vs about 0.076 | Hurt |
| Initialize continuous optimization with RandUserTopItem instead of Balanced (ablation) | Continuous EVOI methods · Goodreads | regret | Improved |
| Use ContPartial rather than greedy attribute selection | Partial comparison elicitation · MovieLens-20M, smaller slates and single attributes | EVOI | Improved |
| Use ContPartial rather than greedy attribute selection | Partial comparison elicitation · MovieLens-20M, larger slates and multiple attributes | EVOI | Hurt |
| Add feasibility regularization to keep optimized query items near feasible items (ablation) | ContFree · MovieLens pairwise comparisons, first query | EVOI: 0.083 -> 0.08 | Hurt |
| Use unconstrained ContFree rather than alternating deep retrieval (ablation) | Continuous EVOI · Goodreads (N=10^6, m=500, k=5) | runtime (sec.): 3.14 (0.06) -> 1.00 (0.04) | Improved |
| Increase slate size from pairwise to slate size 5 | MovieLens elicitation · MovieLens | queries to reach regret 0.03: about half as many queries (3 to 4) | Improved |
Closest papers by meaning
- Preferential Bayesian Optimization (ICML 2017 · similarity 0.86)
- Preference Diffusion for Recommendation (ICLR 2025 · similarity 0.85)
- Fading to Grow: Growing Preference Ratios via Preference Fading Discrete Diffusion for Recommendation (NeurIPS 2025 · similarity 0.84)
- Mult-DPO: Multinomial Direct Preference Optimization for Recommender Systems (NeurIPS 2026 · similarity 0.84)
- BayesCNS: A Unified Bayesian Approach to Address Cold Start and Non-Stationarity in Search Systems at Scale (AAAI 2025 · similarity 0.84)
- On Softmax Direct Preference Optimization for Recommendation (NeurIPS 2024 · similarity 0.84)
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.
