yep, we in fact link to this work... "Even 5 years later, no pure RL algorithms have cracked reasoning and memory games; on the contrary, approaches that have

yep, we in fact link to this work... "Even 5 years later, no pure RL algorithms have cracked reasoning and memory games; on the contrary, approaches that have done well at them have either used instructions <link> or demonstrations <link> just as we mentioned would make sense to do in the board game allegory."