大规模Transformer的摊销规划：以国际象棋为例的案例研究
Amortized Planning with Large-Scale Transformers: A Case Study on Chess
AI总结：
本文通过发布包含1000万局棋的ChessBench数据集并训练大规模Transformer，展示了监督学习可蒸馏出接近Stockfish的规划能力，在无显式搜索下达到特级大师水平，同时表明完美蒸馏仍具挑战性。
摘要：
本文以国际象棋这一人工智能领域的标志性规划问题为对象，评估Transformer在规划任务中的表现，在该任务中，即使在大规模条件下，记忆化也是徒劳的。为此，我们发布了ChessBench，这是一个大规模基准数据集，包含1000万局国际象棋对局，并带有由当前最先进的国际象棋引擎Stockfish 16提供的合法走法和价值标注（150亿个数据点）。我们通过监督学习在ChessBench上训练了参数量高达2.7亿的Transformer，并进行了广泛的消融实验，以评估数据集规模、模型规模、架构类型以及不同预测目标（状态价值、动作价值和行为克隆）的影响。我们最大的模型能够相当准确地预测新棋局的行动价值，这表明其具有高度非平凡的泛化能力。尽管没有执行显式搜索，我们得到的国际象棋策略仍能解决具有挑战性的国际象棋谜题，并在与人类对弈的Lichess快棋中取得了令人惊讶的2895 Elo分（特级大师水平）。我们还将该模型与Leela Chess Zero和AlphaZero（通过自对弈进行无监督训练）在有搜索和无搜索的情况下进行了比较。我们表明，尽管可以通过监督学习将Stockfish基于搜索的算法的一个非常出色的近似蒸馏到大规模Transformer中，但完美蒸馏仍然无法实现，这使得ChessBench非常适合未来的研究。
英文摘要：
This paper uses chess, a landmark planning problem in AI, to assess transformers' performance on a planning task where memorization is futile $\unicode{x2013}$ even at a large scale. To this end, we release ChessBench, a large-scale benchmark dataset of 10 million chess games with legal move and value annotations (15 billion data points) provided by Stockfish 16, the state-of-the-art chess engine. We train transformers with up to 270 million parameters on ChessBench via supervised learning and perform extensive ablations to assess the impact of dataset size, model size, architecture type, and different prediction targets (state-values, action-values, and behavioral cloning). Our largest models learn to predict action-values for novel boards quite accurately, implying highly non-trivial generalization. Despite performing no explicit search, our resulting chess policy solves challenging chess puzzles and achieves a surprisingly strong Lichess blitz Elo of 2895 against humans (grandmaster level). We also compare to Leela Chess Zero and AlphaZero (trained without supervision via self-play) with and without search. We show that, although a remarkably good approximation of Stockfish's search-based algorithm can be distilled into large-scale transformers via supervised learning, perfect distillation is still beyond reach, thus making ChessBench well-suited for future research.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
