Summary of Mastering Zero-shot Interactions in Cooperative and Competitive Simultaneous Games, by Yannik Mahlau et al.
Mastering Zero-Shot Interactions in Cooperative and Competitive Simultaneous Games
by Yannik Mahlau, Frederik Schubert, Bodo Rosenhahn
First submitted to arxiv on: 5 Feb 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Computer Science and Game Theory (cs.GT)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes Albatross, a novel algorithm that combines self-play and planning to improve decision-making in simultaneous games. Building on the success of AlphaZero in sequential games like Chess and Go, Albatross models the behavior of other agents in these games, allowing for cooperation and competition with agents of varying strengths. The algorithm learns to play the Smooth Best Response Logit Equilibrium (SBRLE), a novel equilibrium concept that enables it to adapt to different playing styles. Evaluation on various simultaneous perfect-information games shows significant improvements over AlphaZero, including exploiting weak agents in Battlesnake and achieving better performance in Overcooked. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Albatross is a new way for computers to make decisions when many people are making choices at the same time. This happens in games like Chess or Go, but it’s even harder when lots of people are playing together. To solve this problem, Albatross learns how other players will behave and makes good choices based on that information. It can work with players who are very strong or very weak, which is important for things like working together to achieve a goal or competing against each other. |
