A Closer Look at Offline RL Agents

Authors: Yuwei Fu, Di Wu, Benoit Boulet

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this work, we first introduce a set of experiments to evaluate offline RL agents, focusing on three fundamental aspects: representations, value functions and policies.
Researcher Affiliation Academia Yuwei Fu, Di Wu, Benoit Boulet McGill University yuwei.fu@mail.mail.ca, {di.wu5, benoit.boulet}@mcgill.ca
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at: https://github.com/fuyw/RIQL.
Open Datasets Yes on the standard D4RL dataset [14].
Dataset Splits Yes For each probing target, we use a 5-fold cross-validation on Dprobe to train a linear regression model with Mean Squared Error (MSE) loss.
Hardware Specification No The paper does not provide specific details about the hardware used for experiments in the provided text.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup No The paper mentions training details but does not provide specific hyperparameter values or comprehensive system-level training settings in the provided text.