Bridging Local and Global Knowledge via Transformer in Board Games
Published in IJCAI-25, 2024
This research introduces ResTNet, which combines residual and Transformer blocks to enhance board game performance by integrating local and global knowledge. The approach achieves improved win rates in Go and Hex variants, and successfully recognizes long-sequence patterns like circular and ladder formations.
Recommended citation: Yan-Ru Ju, Tai-Lin Wu, Chung-Chin Shih, Ti-Rong Wu. (2025). "Bridging Local and Global Knowledge via Transformer in Board Games." IJCAI-25.
Download Paper