Research map / Reinforcement learning / Visual and representation RL
Open Bandit Dataset and Pipeline: Towards Realistic and Reproducible Off-Policy Evaluation
Yuta Saito, Shunsuke Aihara, Megumi Matsutani, Yusuke Narita
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 DRos with automatic hyperparameter tuning instead of DR | OPE estimators · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.482 -> 0.316 | Improved |
| use DRos instead of IPW | OPE estimators · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.493 -> 0.316 | Improved |
| increase DRos lambda from 1 to 1,000 (ablation) | DRos · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.963 -> 0.245 | Improved |
| increase DRos lambda from 1,000 to 10,000 (ablation) | DRos · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.245 -> 0.323 | Hurt |
| increase DRos lambda from 1 to 5,000 (ablation) | DRos · Open Bandit Dataset ALL, Random -> Bernoulli TS | RMSE: 1.384 -> 0.476 | Improved |
| use automatic lambda tuning instead of best fixed lambda (ablation) | DRos · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.245 -> 0.323 | Hurt |
| use automatic lambda tuning instead of lambda=1,000 (ablation) | DRos · Open Bandit Dataset ALL, Random -> Bernoulli TS | RMSE: 0.476 -> 0.476 | No clear effect |
| use small sample (n=10,000) instead of large (n=300,000) for DM | DM · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 1.026 -> 0.797 | Improved |
| use small sample (n=10,000) instead of large (n=300,000) for DRos | DRos · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 0.316 -> 0.765 | Hurt |
| replace logistic regression with gradient boosting reward estimator (ablation) | DM · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 1.597 -> 1.026 | Improved |
| replace logistic regression with gradient boosting reward estimator (ablation) | DR · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 2.250 -> 0.482 | Improved |
| replace logistic regression with gradient boosting reward estimator (ablation) | DRos · Open Bandit Dataset ALL, Bernoulli TS -> Random | RMSE: 2.065 -> 0.316 | Improved |
Closest papers by meaning
- Cross-Domain Off-Policy Evaluation and Learning for Contextual Bandits (ICLR 2025 · similarity 0.90)
- OPERA: Automatic Offline Policy Evaluation with Re-weighted Aggregates of Multiple Estimators (NeurIPS 2024 · similarity 0.87)
- Log-Sum-Exponential Estimator for Off-Policy Evaluation and Learning (ICML 2025 · similarity 0.87)
- Budgeted Multi-Source Counterfactual Annotation for Off-Policy Evaluation (NeurIPS 2026 · similarity 0.86)
- Towards Assessing and Benchmarking Risk-Return Tradeoff of Off-Policy Evaluation (ICLR 2024 · similarity 0.86)
- Off-Policy Evaluation under Nonignorable Missing Data (ICML 2025 · 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.
