Model-Based Opponent Modeling

Authors: XiaoPeng Yu, Jiechuan Jiang, Wanpeng Zhang, Haobin Jiang, Zongqing Lu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we show that MBOM achieves more effective adaptation than existing methods in a variety of tasks, respectively with different types of opponents, i.e., fixed policy, naïve learner, and reasoning learner.
Researcher Affiliation Academia Xiaopeng Yu Jiechuan Jiang Wanpeng Zhang Haobin Jiang Zongqing Lu School of Computer Science, Peking University
Pseudocode Yes Algorithm 1 MBOM
Open Source Code No The code for the main experiments will be released after acceptance.
Open Datasets Yes Triangle Game is an asymmetric zero-sum game implemented on Multi-Agent Particle Environments (MPE) [21].
Dataset Splits No The paper describes interacting with environments like MPE and Google Research Football but does not provide explicit training, validation, and test dataset splits with percentages or sample counts, as these are simulated environments rather than static datasets.
Hardware Specification Yes All the experiments were run on NVIDIA GeForce RTX 3090 GPUs.
Software Dependencies Yes We use PyTorch (v1.8.0) to implement our method and baselines.
Experiment Setup Yes More details about experimental settings and hyperparameters are available in Appendix E.