Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Optimal and Adaptive Off-policy Evaluation in Contextual Bandits
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 |
|---|---|---|---|
| Use SWITCH-DR, applying the reward model to actions with large importance weights and DR to the rest | Off-policy value estimation · UCI contextual-bandit simulations, deterministic and noisy rewards | relative MSE | Helped |
| Use DR instead of IPS with a reward model | Off-policy value estimation · UCI contextual-bandit simulations | MSE | Helped |
| Replace IPS with SWITCH-DR when the direct method performs poorly | Off-policy value estimation · UCI contextual-bandit simulations, including optdigits and yeast | MSE | No clear effect |
| Use IPS or DR without a consistent reward model | Off-policy value estimation · Agnostic contextual bandits | minimax MSE | No clear effect |
| Use a single automatically selected importance-weight threshold rather than a multi-threshold MAGIC combination | SWITCH threshold selection · UCI contextual-bandit simulations | relative MSE | Helped |
| Use DM rather than IPS for large importance weights | SWITCH · Off-policy evaluation with large or heavy-tailed importance weights | variance | Helped |
| Increase reward noise | IPS and DR · UCI contextual-bandit simulations with noisy rewards | MSE | Hurt |
| Set the SWITCH threshold to zero, yielding DM (ablation) | SWITCH · Off-policy evaluation | MSE | Mixed |
| Set the SWITCH threshold to infinity, yielding IPS (ablation) | SWITCH · Off-policy evaluation | MSE | Hurt |
| Replace DM imputation on large-weight actions with zero, yielding Trimmed IPS (ablation) | SWITCH · UCI contextual-bandit simulations | relative MSE | Hurt |
Closest papers by meaning
- Marginal Density Ratio for Off-Policy Evaluation in Contextual Bandits (NeurIPS 2023 · similarity 0.93)
- Off-Policy Risk Assessment in Contextual Bandits (NeurIPS 2021 · similarity 0.92)
- Cross-Domain Off-Policy Evaluation and Learning for Contextual Bandits (ICLR 2025 · similarity 0.90)
- Log-Sum-Exponential Estimator for Off-Policy Evaluation and Learning (ICML 2025 · similarity 0.90)
- A General Framework for Off-Policy Learning with Partially-Observed Reward (ICLR 2025 · similarity 0.90)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.90)
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.