Research map / Reinforcement learning / Offline RL
Counterfactual-Augmented Importance Sampling for Semi-Offline 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 |
|---|---|---|---|
| Use C*-PDIS with evaluation-policy Q annotations instead of standard PDIS | Sepsis simulator policy evaluation · Sepsis simulator | RMSE: 0.113 ±0.038 -> 0.013 ±0.005 | Helped |
| Use C*-PDIS with evaluation-policy Q annotations instead of standard PDIS | Sepsis simulator policy evaluation · Sepsis simulator | Spearman: 0.596 ±0.110 -> 0.995 ±0.003 | Helped |
| Use C*-PDIS with evaluation-policy Q annotations instead of standard PDIS | Sepsis simulator policy evaluation · Sepsis simulator | %Accuracy: 76.5 ±3.5 -> 95.7 ±3.1 | Helped |
| Add counterfactual annotations as new trajectories without reweighting | Sepsis simulator policy evaluation · Sepsis simulator | RMSE: 0.113 ±0.038 -> 0.128 ±0.006 | Hurt |
| Reweight counterfactual annotations at the trajectory level | Sepsis simulator policy evaluation · Sepsis simulator | RMSE: 0.113 ±0.038 -> 0.097 ±0.006 | Helped |
| Use behavior-policy Q annotations instead of evaluation-policy Q annotations (ablation) | C*-PDIS · Sepsis simulator | RMSE: 0.013 ±0.005 -> 0.070 ±0.003 | Hurt |
| Correct behavior-policy Q annotations using the estimated annotation-bias function (ablation) | C*-PDIS with behavior-policy Q annotations · Sepsis simulator | RMSE: 0.070 ±0.003 -> 0.028 ±0.007 | Helped |
| Use C*-IS with equal weights instead of standard IS | Synthetic bandit evaluation · Synthetic bandits | RMSE | Helped |
| Add counterfactual annotations as unweighted samples | Synthetic bandit evaluation · Synthetic bandits | RMSE | Hurt |
| Increase the amount of available counterfactual annotations (ablation) | C*-PDIS · Sepsis simulator | OPE performance | Helped |
| Impute missing counterfactual annotations with an annotation model (ablation) | C*-PDIS · Sepsis simulator | OPE performance | Helped |
| Add noise to counterfactual annotations (ablation) | C*-PDIS · Sepsis simulator | OPE performance | Hurt |
Closest papers by meaning
- Budgeted Multi-Source Counterfactual Annotation for Off-Policy Evaluation (NeurIPS 2026 · similarity 0.91)
- Demystifying the Paradox of Importance Sampling with an Estimated History-Dependent Behavior Policy in Off-Policy Evaluation (ICML 2025 · similarity 0.90)
- OPERA: Automatic Offline Policy Evaluation with Re-weighted Aggregates of Multiple Estimators (NeurIPS 2024 · similarity 0.89)
- Model Selection for Off-policy Evaluation: New Algorithms and Experimental Protocol (NeurIPS 2025 · similarity 0.89)
- Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning (AAAI 2024 · similarity 0.88)
- Abstract Reward Processes: Leveraging State Abstraction for Consistent Off-Policy Evaluation (NeurIPS 2024 · similarity 0.88)
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.