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].
Asymptotic Extinction in Large Coordination Games
Authors: Desmond Chan, Bart De Keijzer, Tobias Galla, Stefanos Leonardos, Carmine Ventre
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We compare how our theoretical results fare against numerical experiments of finite-sized, coordination games. We used the default Sci Py Runga-Kutta 4(5) solver (Virtanen et al. 2020) with max stepsize set to 0.5 to be approximate continuous Q-Learning (4) as closely as possible. ... Figure 6 displays a heat map, displaying the likelihood Q-Learning convergences to a unique fixed point, given the chosen parameters. |
| Researcher Affiliation | Academia | 1 King s College London 2 Institute for Cross-Disciplinary Physics and Complex Systems (IFISC, CSIC-UIB) EMAIL , bart.de EMAIL, EMAIL, EMAIL , EMAIL |
| Pseudocode | No | The paper presents mathematical equations and descriptions of the Q-Learning process and effective dynamics, but it does not include any clearly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain an explicit statement about releasing code, nor does it provide a link to a code repository or mention code in supplementary materials for the methodology described. |
| Open Datasets | No | To generate a game, we draw the payoff matrix, Π, from a multivariate Gaussian with mean 0 which treats all players symmetrically 1.The covariance matrix of the distribution is determined by parameter Γ ( 1, p 1), which captures the pairwise correlations between the players payoffs. ... In each game, the payoffs matrices are randomly-generated and held fixed. |
| Dataset Splits | No | The paper describes how games are generated by drawing payoff matrices from multivariate Gaussians. It mentions that "100 random initial strategies are drawn and simulated for up to 5000 time units" and "40 independent games are generated," but these refer to simulation setups and random initializations, not explicit training/test/validation splits of a static dataset. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the numerical experiments. |
| Software Dependencies | Yes | We used the default Sci Py Runga-Kutta 4(5) solver (Virtanen et al. 2020) with max stepsize set to 0.5 to be approximate continuous Q-Learning (4) as closely as possible. |
| Experiment Setup | Yes | We used the default Sci Py Runga-Kutta 4(5) solver (Virtanen et al. 2020) with max stepsize set to 0.5 to be approximate continuous Q-Learning (4) as closely as possible. A point is classified as fixed when the derivative of (4) drops below |10 8|. To determine if a given game converges to a unique fixed point, 100 random initial strategies are drawn and simulated for up to 5000 time units, or until it reaches a fixed point. If all 100 final points, are within a relative distance of 0.01 of each other, we assume there is a unique fixed point. For each Γ and T, 40 independent games are generated, and we record the proportion of games for which Q-Learning converges to a unique fixed point. |