The paper list of the 86-page paper "The Rise and Potential of Large Language Model Based Agents: A Survey" by Zhiheng Xi et al. - GitHub - WooooDyy/LLM-Agent-Paper-List: The paper list of the 86-page paper "The Rise and Potential of Large Language Model Based Agents: A Survey" by Zhiheng Xi et al.
by
Facebook AI Research (FAIR) published a paper on Recursive Belief-based Learning (ReBeL), their new AI for playing imperfect-information games that can defeat top human players in poker. The algorithm combines reinforcement learning (RL) with state-space search and converges to a Nash equilibrium for any two player zero sum game. Code for training the algorithm to play Liar s Dice has been open-sourced.
FAIR researchers Noam Brown and Anton Bakhtin gave an overview of the system in a blog post. ReBeL is a general-purpose algorithm for use in any two player zero sum game, even those with
imperfect information, where players do not have complete knowledge of the game state. By modeling the probability distribution of the beliefs that players may have about game state, ReBeL can apply AI techniques used in perfect-information games. The algorithm is proven to converge to the optimal policy for the game, and FAIR s implementation for Heads-Up No-Limit (HUNL) Texas Hold Em poker