Solving Hanabi: Estimating Hands by Opponent's Actions in Cooperative Game with Incomplete Information

By Hirotaka Osawa. 2015

48 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

The Hanabi Challenge: A New Frontier for AI Research (2019, background): “Osawa [43] described some of the unique elements of Hanabi, and showed that simulated strategies that try to recognise the intention of the other players performed better than a fixed set of static strategies in two-player Hanabi.”
Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning (2018, background): “Osawa (2015) deﬁned a number of heuristics for the two-player case that reason over possible hands given the other player’s action.”
Diverse Agents for Ad-Hoc Cooperation in Hanabi (2019, background): “Many of the early approaches [10]–[12] for playing Hanabi with AI were variations of a simple strategy for self-play which prioritizes playing cards that are believed to be playable,”
Evolving Agents for the Hanabi 2018 CIG Competition (2018, background): “Osawa’s [5] best performing agent enhances a sequence of rules with a search of all possible previous states that are consistent with the other player’s last action.”
Re-determinizing Information Set Monte Carlo Tree Search in Hanabi (2019, background): “[20] investigate 2-player Hanabi strategies, and compare some heuristic rules with an approach that models the other player’s likely next action assuming they are using the same strategy.”
Implicit Communication of Actionable Information in Human-AI teams (2019, background): “Previous work in building AIs that can collaborate with humans, both for Hanabi [26, 60] and for other games [77], has generally centered onmodeling teammates’ mental states and exhaustively searching all potential states.”
Practical Specification of Belief Manipulation in Games (2021, background): “Osawa (2015) developed several AI agents for the game, noting that having a model of what the other agent knows improves the score that is obtained by the agents.”
Playing Hanabi Near-Optimally (2017, background): “Osawa [10] describes experiments with two players and five cards per player.”
Evaluating and modelling Hanabi-playing agents (2017, background): “Osawa [15] devised a number of rule-based agents for the 2-player version of the game, the mechanisms for which are described in Sections II-A2 and II-A3.”
Aspects of the Cooperative Card Game Hanabi (2016, background): “The results from [6] for the two-player version, using online learning, even seem to be a little better than for the three-player version, although it could be that the two-player version allows for higher final scores.”
I See What You See: Integrating Eye Tracking into Hanabi Playing Agents (2018, background): “Van den Bergh et al. (2015) present several agents using simple if-then rules defined by experts.”