Multi-Agent Training beyond Zero-Sum with Correlated Equilibrium Meta-Solvers

Authors: Luke Marris, Paul Muller, Marc Lanctot, Karl Tuyls, Thore Graepel

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct several experiments using CE meta-solvers for JPSRO and demonstrate convergence on n-player, general-sum games.
Researcher Affiliation Collaboration 1Deep Mind 2University College London 3Universit e Gustave Eiffel.
Pseudocode Yes Algorithm 1 Two-Player PSRO ... Algorithm 2 JPSRO
Open Source Code No The paper does not provide an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes All games used are available in Open Spiel (Lanctot et al., 2019).
Dataset Splits No The paper describes training multi-agent systems within game environments, but does not specify explicit training/validation/test dataset splits with percentages or sample counts for reproduction.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU/GPU models, memory, or cloud instance types) used for running experiments.
Software Dependencies No The paper mentions software like CVXPY, OSQP, and Open Spiel, but does not provide specific version numbers for these dependencies, which are necessary for reproducible setup.
Experiment Setup Yes We use an exact BR oracle, and exactly evaluate policies in the meta-game by traversing the game tree to precisely isolate the MS s contribution to the algorithm. ... Random solvers were evaluated with five seeds and we plot the mean. ... Experiments were ran for up to 6 hours, after which they were terminated.