Get the latest tech news
OK, I can partly explain the LLM chess weirdness now
(“make LLMs play better with one weird trick”)
You are a chess grandmaster.You will be given a list of legal moves and a partially completed game.After seeing it, you should choose the next move.Use standard algebraic notation, e.g. "e4" or "Rdf8" or "R1a3".NEVER give a turn number.NEVER explain your choice. Meanwhile, in section A.2 of this paper(h/t Gwern) some OpenAI authors mention that GPT-4 was trained on chess games in PGN notation, filtered to only include players with Elo at least 1800. But I suspect not, because a smart model would recognize that if the sequence of moves so far is high skill then the player isn’t a total idiot and probably won’t throw away their queen.
Or read this on Hacker News
