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].
Pessimism Meets Invariance: Provably Efficient Offline Mean-Field Multi-Agent RL
Authors: Minshuo Chen, Yan Li, Ethan Wang, Zhuoran Yang, Zhaoran Wang, Tuo Zhao
NeurIPS 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments are provided.We perform experiments on the multi-agent particle environment (MPE, Lowe et al. (2017)), a popular benchmark used in prior work (Mordatch and Abbeel, 2018; Liu et al., 2020a). |
| Researcher Affiliation | Academia | Minshuo Chen1 Yan Li1 Ethan Wang1 Zhuoran Yang2 Zhaoran Wang3 Tuo Zhao1 1Georgia Tech 2University of California, Berkeley 3Northwestern University |
| Pseudocode | Yes | Algorithm 1 Pessimistic Mean-Field Value Iteration (SAFARI) |
| Open Source Code | No | Sample code is also available at (followed by an empty link in the PDF). Extension to online setting is provided in a longer technical report version, which is available upon request. |
| Open Datasets | Yes | We perform experiments on the multi-agent particle environment (MPE, Lowe et al. (2017)), a popular benchmark used in prior work (Mordatch and Abbeel, 2018; Liu et al., 2020a). |
| Dataset Splits | No | No explicit statement of train/validation/test dataset splits was found. The paper mentions using “n = 500 sample episodes of experience data” for training, but does not detail how this data is partitioned for validation purposes or if there’s a specific validation set. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory specifications) used for experiments were provided. |
| Software Dependencies | No | No specific software dependencies with version numbers (e.g., library versions, programming language versions) were provided. |
| Experiment Setup | Yes | Both the policy and critic networks are implemented as traditional MLPs, with 64 and 512 nodes in a single hidden layer, respectively, and we use parameter sharing for policy networks. |