PureTD model achieves strong backgammon performance with no search

Researchers have developed PureTD, a reinforcement learning model for backgammon money games that achieves near state-of-the-art performance without requiring search during evaluation. Trained solely through self-play, PureTD demonstrates that pure reinforcement learning is sufficient for developing strong backgammon players. The model significantly outperforms open-source engines like GNUbg and Open Sage in cubeful money games, even when those engines employ a one-ply look-ahead search. AI

IMPACT Demonstrates the efficacy of pure self-play reinforcement learning for complex strategy games, potentially influencing AI development in other domains.

RANK_REASON The cluster contains a research paper detailing a new reinforcement learning model for backgammon. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• GNUbg
• Hugging Face
• IArxiv
• Open Sage
• PureTD
• TD-Gammon

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →