New Method Enables High-Dimensional Off-Policy Evaluation Without Sample Splitting
Researchers propose an Onsager correction for off-policy evaluation that removes the need for held-out data when reward models are selected, even as sample size and context dimension grow proportionally.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-09 05:00 · arXiv stat.ML
High-Dimensional Off-Policy Evaluation without Sample Splitting using an Onsager Correction