datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CLI-Gym
CLI-Gym Verified Problem Instances
This is the official dataset release of environment-intensive (CLI) problem instances from CLI-Gym.
Each instance is designed to evaluate and train an agent’s ability to diagnose and repair real runtime environments via CLI interaction (dependencies, permissions, system configuration, networking, etc.), with executable verification through unit tests.
News
[2026-02-28] Full dataset release! (1,655 CLI-Gym problem instances).… See the full description on the dataset page: https://huggingface.co/datasets/LiberCoders/CLI-Gym.agent-task-cli-gymopen-instruct-cli-gym
💻 Code ·
🤗 Models & Data ·
📜 Paper ·
📓 Blog
[!NOTE]
For full information, go check out the Tmax paper here.
CLI Gym - Open Instruct formatted
This is cli-gym formatted for use with our fork of open-instruct.
For more information and the original data, you should check out the original source.
cligym-compaction-effort-nc0.3drctx-compaction-cligym-mk1cligym-compaction-mk2-nc1