frosthead/ChessCOT
收藏Hugging Face2025-03-26 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/frosthead/ChessCOT
下载链接
链接失效反馈官方服务：
资源简介：
ChessCOT数据集是一个为棋类模型训练设计的数据库，使用Chain of Thought (CoT)方法，目的是让模型在走棋前能够对所有可能的走法和它们的后果进行推理，以预测最佳走法。数据集包含约449万种棋盘位置，每个位置都有对应的思考过程和最佳走法预测。数据集使用UCI格式，覆盖2500至3200 ELO等级的棋局，并能够预测黑白两方的最佳走法。
The ChessCOT dataset is designed for training chess models using a Chain of Thought (CoT) approach, aiming to make the model reason about all possible moves and their consequences before making a move to predict the best move. The dataset contains about 4.49 million chess positions, each with the corresponding thought process and the models prediction of the best move. The dataset uses the UCI format, covering games from 2500 to 3200 ELO ratings, and is capable of predicting the best moves for both White and Black.
提供机构：
frosthead
