Research map / Reinforcement learning / Policy gradient and actor-critic
Reliable Decision Support using Counterfactual Models
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 the counterfactual objective instead of the standard supervised objective | GP risk model · Simulated risk prediction, Regime B | Risk Score Δ from A: 0.083 -> 0.001 | Helped |
| Use the counterfactual objective instead of the standard supervised objective | GP risk model · Simulated risk prediction, Regime B | Kendall's τ from A: 0.857 -> 0.998 | Helped |
| Use the counterfactual objective instead of the standard supervised objective | GP risk model · Simulated risk prediction, Regime B | AUC: 0.832 -> 0.872 | Helped |
| Train under a policy that violates the CGP assumptions | CGP · Simulated risk prediction, Regime C | Risk Score Δ from A: 0.001 -> 0.128 | Hurt |
| Train under a policy that violates the CGP assumptions | CGP · Simulated risk prediction, Regime C | Kendall's τ from A: 0.998 -> 0.562 | Hurt |
| Train under a policy that violates the CGP assumptions | CGP · Simulated risk prediction, Regime C | AUC: 0.872 -> 0.829 | Hurt |
| Train on a different action policy | Baseline GP · Simulated risk prediction, Regime B | Risk Score Δ from A: 0.000 -> 0.083 | Hurt |
| Train on a different action policy | Baseline GP · Simulated risk prediction, Regime B | AUC: 0.853 -> 0.832 | Hurt |
| Add treatment-effect modeling to the three-GP mixture (ablation) | CGP outcome model · MIMIC-II creatinine, 24-hour horizon | MAE: 0.07 (0.06, 0.08) | Helped |
| Add treatment-effect modeling to the three-GP mixture (ablation) | CGP outcome model · MIMIC-II creatinine, 24-48 hours | MAE: 0.09 (0.08, 0.10) | Helped |
| Use a mixture of GPs instead of a single GP while modeling treatment effects (ablation) | CGP outcome model · MIMIC-II creatinine, 24-hour horizon | MAE: 0.04 (0.04, 0.05) | Helped |
| Use a mixture of GPs instead of a single GP while modeling treatment effects (ablation) | CGP outcome model · MIMIC-II creatinine, 24-48 hours | MAE: 0.03 (0.02, 0.04) | Helped |
Closest papers by meaning
- Variational Counterfactual Intervention Planning to Achieve Target Outcomes (ICML 2025 · similarity 0.87)
- Designing Decision Support Systems using Counterfactual Prediction Sets (ICML 2024 · similarity 0.86)
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching (NeurIPS 2025 · similarity 0.85)
- Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agents (ICML 2025 · similarity 0.84)
- Learning Counterfactual Outcomes Under Rank Preservation (NeurIPS 2025 · similarity 0.84)
- Controlling Counterfactual Harm in Decision Support Systems Based on Prediction Sets (NeurIPS 2024 · similarity 0.84)
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.