Efficient Subgame Refinement for Extensive-form Games
Authors: Zhenxing Ge, Zheng Xu, Tianyu Ding, Wenbin Li, Yang Gao
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments conducted on research games as well as the challenging large game of Guan Dan demonstrate a significant improvement over the blueprint. |
| Researcher Affiliation | Collaboration | Zhenxing Ge Nanjing University Nanjing, Jiangsu, China zhenxingge@smail.nju.edu.cn; Zheng Xu Nanjing University Nanjing, Jiangsu, China xuzhengcs@smail.nju.edu.cn; Tianyu Ding Microsoft Corporation Redmond, Washington, USA tianyuding@microsoft.com; Wenbin Li Nanjing University Nanjing, Jiangsu, China liwenbin@nju.edu.cn; Yang Gao Nanjing University Nanjing, Jiangsu, China gaoy@nju.edu.cn |
| Pseudocode | Yes | The pseudo-code for GS2 utilizing the diversity-based generation function is presented in Appendix C.8. |
| Open Source Code | No | The paper mentions using "Open Spiel" (an open-source library), but it does not state that the code for their proposed method (GS2) is open-source or provide a link to it. |
| Open Datasets | No | The paper evaluates on research games (e.g., Leduc poker, Liar's Dice, Goofspiel) and Guan Dan, which are game environments. It does not mention or provide access information for specific, publicly available datasets (e.g., recorded game data) used for training models in a way that allows reproduction. |
| Dataset Splits | No | The paper does not provide specific dataset split information (e.g., percentages, sample counts, or references to predefined splits) for training, validation, or testing. The experiments are conducted within game environments, not on pre-split datasets. |
| Hardware Specification | Yes | The experiment was conducted on an Intel(R) Xeon(R) Gold 6242R CPU @ 3.10GHz |
| Software Dependencies | No | The paper mentions using "Open Spiel" and "MCCFR" but does not provide specific version numbers for these or other software components like linear programming solvers. Therefore, a reproducible description of ancillary software is not provided. |
| Experiment Setup | Yes | The computation time is limited for 1-KLSS [Zhang and Sandholm, 2021] and GS2. In these methods, the player is allotted a maximum of 3 seconds to compute an approximate Nash Equilibrium after constructing a subgame. |