Datasets for LLM reinforcement learning in sandboxed environments in our paper: Computer Environments Elicit General Agentic Intelligence in LLMs

from datasets import load_dataset

ds = load_dataset("daixuancheng/llm-in-sandbox-rl", "instruct_pretrain", split="train")

ds = load_dataset("daixuancheng/llm-in-sandbox-rl", "math_mini"… See the full description on the dataset page: https://huggingface.co/datasets/daixuancheng/llm-in-sandbox-rl.