yuruny/agentic-sokoban-qwen2.5-3B-eval_results
收藏Hugging Face2025-12-18 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/yuruny/agentic-sokoban-qwen2.5-3B-eval_results
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含一系列步骤，每个步骤具有多个属性，包括动作、聊天完成（包含内容和角色）、完成状态、mc_return、模型响应、观察和奖励。此外，还有一个顶层的奖励特征。数据集包含一个训练分割，共有12800个示例，总大小为45273408字节。数据集用于训练目的，数据文件路径指向训练数据。
The dataset consists of a series of steps, each with multiple attributes including action, chat_completions (which contains content and role), done, mc_return, model_response, observation, and reward. Additionally, there is a top-level reward feature. The dataset includes a train split with 12,800 examples and a total size of 45,273,408 bytes. The dataset is intended for training purposes, with the data file path pointing to the training data.
提供机构：
yuruny
