Research map / Reinforcement learning / Policy gradient and actor-critic
Doubly Robust Distributionally Robust Off-Policy Evaluation and Learning
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 |
|---|---|---|---|
| Replace SNIPS-based DROPE with localized doubly robust estimator (LDR^2OPE) | DROPE evaluation · Simulation, delta=0.1,0.2,0.3, known and unknown propensities | MSE of DROPE estimator | Helped |
| Use estimated (LightGBM) propensities in LDR^2OPE instead of known propensities (ablation) | LDR^2OPE · Simulation DROPE, N >= 10^4 | MSE | No clear effect |
| Doubly robust DROPL (CDR^2OPL) instead of SNIPS objective | DROPL policy learning · Simulation, delta=0.1 | distributionally robust value of learned policy: about 1% | Helped |
| Increase uncertainty radius delta for CDR^2OPL vs SNIPS | DROPL policy learning · Simulation, delta=0.1 to 0.3 | improvement over SNIPS robust value | Hurt |
| Self-normalize propensity weights in doubly robust methods | LDR^2OPE and CDR^2OPL · Simulation, small N | estimation stability | Helped |
| Use LDR^2OPE in small-N, unknown propensity, large delta regime | DROPE evaluation · Simulation, small N, unknown propensities, large delta | stability (MSE) | Hurt |
| Use CDR^2OPL (continuum regressions, O(N^2) cost) instead of SNIPS | DROPL · Simulation computational cost | running time | Hurt |
Closest papers by meaning
- Doubly Robust Distributionally Robust Offline Contextual Pricing (ICML 2026 · similarity 0.90)
- Doubly Robust Bias Reduction in Infinite Horizon Off-Policy Estimation (ICLR 2020 · similarity 0.88)
- Distributionally-Robust Policy Learning from Observational Data (NeurIPS 2026 · similarity 0.88)
- Distributionally Robust Policy Evaluation and Learning for Continuous Treatment with Observational Data (AAAI 2025 · similarity 0.86)
- Off-Policy Evaluation under Nonignorable Missing Data (ICML 2025 · similarity 0.86)
- Distributionally Robust Policy Learning under Concept Drifts (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.