yuruny/agentic-sudoku-NonMarkov_qwen2.5-3B-it-5e-6_9x9_6-6_gt-SFT_ans1-9k-eval_results

该数据集包含steps和reward两个主要特征。steps特征下包含多个子特征，如action（动作）、chat_completions（聊天补全，包含content和role）、done（是否完成）、mc_return（MC返回）、model_response（模型响应）、observation（观察）和reward（奖励）。数据集分为一个train训练集，包含12,800个样本，总大小为28,834,538字节。

The dataset includes two main features: steps and reward. The steps feature contains several sub-features such as action, chat_completions (including content and role), done, mc_return, model_response, observation, and reward. The dataset is split into a train set with 12,800 examples and a total size of 28,834,538 bytes.