Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Doubly robust off-policy evaluation with shrinkage
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 |
|---|---|---|---|
| shrink importance weights with DRs-PI | DR-PI · MSLR-WEB10K ranking | MSE: 1.5 or more | Helped |
| select z=w² for reward-predictor training (ablation) | DRs · 108 atomic-action evaluation conditions | conditions where predictor dominates all others: 35 | Helped |
| use MRDR reward-predictor training with shrinkage (ablation) | DRs · 108 atomic-action evaluation conditions | conditions where predictor dominates all others: 18 (2) | Hurt |
| choose pessimistic rather than optimistic shrinkage with z=1 (ablation) | DRs · 108 atomic-action evaluation conditions | conditions where one shrinkage type dominates the other: 58 versus 28 | Mixed |
| choose optimistic rather than pessimistic shrinkage with zero reward predictor (ablation) | DRs · 108 atomic-action evaluation conditions | conditions where one shrinkage type dominates the other: 51 versus 21 | Helped |
| use DRs-direct model selection instead of DRs-upper (ablation) | DRs · stochastic-reward atomic-action evaluation | aggregate MSE | Helped |
| use DRs-direct for off-policy learning | DM, DR, and IPS · UCI off-policy learning | learned-policy value | Helped |
| use DRs-direct for off-policy learning | DM, DR, and IPS · optdigits off-policy learning | learned-policy value | No clear effect |
| shrink DR-PI weights | DR-PI · MSLR-WEB10K ranking | MSE: 1.5 or more | Helped |
Closest papers by meaning
- Optimal and Adaptive Off-policy Evaluation in Contextual Bandits (ICML 2017 · similarity 0.92)
- On the Design of Estimators for Bandit Off-Policy Evaluation (ICML 2019 · similarity 0.91)
- Off-Policy Learning in Large Action Spaces: Optimization Matters More Than Estimation (ICML 2026 · similarity 0.91)
- Off-Policy Evaluation for Large Action Spaces via Conjunct Effect Modeling (ICML 2023 · similarity 0.91)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.90)
- Log-Sum-Exponential Estimator for Off-Policy Evaluation and Learning (ICML 2025 · similarity 0.89)
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.