Overall
Models play full games against classical engines. The bar is mean keep-the-max quality (S_after / S_before) over scored decisions, pooled across Hex, Othello, Gomoku, and Shogi.
Models play full games against classical engines. The bar is mean keep-the-max quality (S_after / S_before) over scored decisions, pooled across Hex, Othello, Gomoku, and Shogi.