Research map / Reinforcement learning / MDPs and planning
Breaking the Order Barrier: Off-Policy Evaluation for Confounded POMDPs
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes off-policy evaluation estimators for confounded POMDPs using bridge functions, aiming to break the order barrier in error bounds with respect to horizon T and the observation-action space size.
Problem
Off-policy evaluation in partially observable MDPs with unobserved confounding, where existing bridge-function-based estimators have statistical error bounds that depend poorly on horizon length T and the size of the observation-action space |O||A|.
Approach
Develops bridge-function-based OPE estimators for confounded POMDPs whose statistical error bounds improve on prior order-dependent rates in horizon and observation-action space size, as suggested by the title ("Breaking the Order Barrier").
Limitations
- Full text unavailable; only the abstract was used, so specific rates and experiments are unknown.
Experimental setup
Builds on: bridge function methods for confounded POMDPs
Closest papers by meaning
- A Minimax Learning Approach to Off-Policy Evaluation in Confounded Partially Observable Markov Decision Processes (ICML 2022 · similarity 0.94)
- Statistical Tractability of Off-policy Evaluation of History-dependent Policies in POMDPs (ICLR 2025 · similarity 0.92)
- A Policy Gradient Method for Confounded POMDPs (ICLR 2024 · similarity 0.91)
- Model-based Reinforcement Learning for Confounded POMDPs (ICML 2024 · similarity 0.91)
- Efficient and Sharp Off-Policy Evaluation in Robust Markov Decision Processes (NeurIPS 2024 · similarity 0.91)
- On the Curses of Future and History in Future-dependent Value Functions for Off-policy Evaluation (NeurIPS 2024 · similarity 0.90)
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.