Research map / Bandits and online learning / Contextual bandits and Thompson sampling
On the Design of Estimators for Bandit Off-Policy Evaluation
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 |
|---|---|---|---|
| subtract IPS control variate from DR (DR & IPS CV) | DR · UCI multiclass bandit datasets (9 datasets) | RMSE: small gains on smaller datasets (e.g., glass 0.2497 to 0.2271); ties on large datasets | Helped |
| subtract IPS control variate from MRDR (MRDR & IPS CV) | MRDR · UCI multiclass bandit datasets (9 datasets) | RMSE: small gains (e.g., ecoli 0.1673 to 0.1665); ties on large datasets | Helped |
| IPS control variate added to DR/MRDR on large datasets (>5500 obs) | DR and MRDR · UCI letter, optdigits, page-blocks, pendigits, satimage | RMSE: CV coefficient near zero; identical RMSE to base | No clear effect |
| optimized IPS plus lambda*t control variate (REG/IPS base) on K-armed Bernoulli bandit | IPS base estimator · simulation with 5 and 10 actions | MSE | Helped |
| plug-in sample-surrogate control variate versus oracle lambda | K-armed bandit CV · simulation | MSE | Mixed |
Closest papers by meaning
- Optimal and Adaptive Off-policy Evaluation in Contextual Bandits (ICML 2017 · similarity 0.92)
- Doubly robust off-policy evaluation with shrinkage (ICML 2020 · similarity 0.91)
- Adaptive Estimator Selection for Off-Policy Evaluation (ICML 2020 · similarity 0.91)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.89)
- Log-Sum-Exponential Estimator for Off-Policy Evaluation and Learning (ICML 2025 · similarity 0.89)
- Leveraging (Biased) Information: Multi-armed Bandits with Offline Data (ICML 2024 · 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.