Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

MoMA: Model-based Mirror Ascent for Offline Reinforcement Learning

Authors: Mao Hong, Zhiyue Zhang, Yue Wu, Yanxun Xu

TMLR 2024 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The effectiveness of Mo MA is demonstrated via numerical studies. (...) Numerical results in both synthetic dataset and D4RL benchmark are included in section 6.
Researcher Affiliation Academia Mao Hong EMAIL Department of Applied Mathematics and Statistics Johns Hopkins University
Pseudocode Yes Algorithm 1 Mo MA: Model-based mirror ascent for offline RL (...) Algorithm 2 Mo MA: A Practical Algorithm (...) Algorithm 3 Mo MA: A Practical Algorithm in the continuous-action case
Open Source Code No The paper does not contain any explicit statements about releasing source code or provide a link to a code repository. The Open Review link is for peer review, not code.
Open Datasets Yes We perform numerical studies on both an illustrative synthetic dataset and Mu Jo Co (Todorov et al., 2012) benchmark datasets (...) We consider the medium, medium-replay, and medium-expert datasets for the Hopper, Half Cheetah, and Walker2D tasks (all v0).
Dataset Splits No The paper mentions using 'medium, medium-replay, and medium-expert datasets' and a 'partially covered offline dataset' but does not provide explicit training/test/validation split percentages, absolute sample counts for splits, or detailed splitting methodologies within the main text or appendices for reproducibility.
Hardware Specification Yes We train and evaluate Mo MA as well as baseline algorithms on one A100 GPU for all D4RL experiments
Software Dependencies No The paper does not explicitly mention any specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions or other libraries).
Experiment Setup Yes The hyperparameters of Mo MA used in the random walk experiment are summarized in table 3. (...) The hyperparameters of Mo MA used in the D4RL experiments are summarized in Table 4.