starreeze/latr-data
收藏Hugging Face2025-10-30 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/starreeze/latr-data
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为LATR Data，是基于前瞻树状展开的强化学习可验证奖励框架下的轨迹级探索数据集。它旨在支持大型语言模型在GRPO和DAPO等算法中的训练和评估，提供多样化的推理任务轨迹，以加速策略学习和提升LLM的表现。
The dataset named LATR Data is designed for trajectory-level exploration in Reinforcement Learning with Verifiable Rewards (RLVR) based on lookahead tree-based rollouts. It aims to support the training and evaluation of large language models in algorithms like GRPO and DAPO, providing diverse trajectories for reasoning tasks to accelerate policy learning and enhance LLM performance.
提供机构：
starreeze
