datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Gui-agent
Gui-Agent — GUI trajectories in LIBERO/VLA format
Human GUI demonstrations from four sources, unified into a single VLA-style
intermediate representation and written as LIBERO-layout HDF5, so LIBERO/VLA
dataloaders run against GUI data unchanged.
raw source ──[adapter]──> GuiEpisode ──[writer]──> LIBERO-style HDF5
per-source the IR format- what you train on
only specific
25,872 episodes / 453,264 steps / 235 GB… See the full description on the dataset page: https://huggingface.co/datasets/Yushi123/Gui-agent.DATA_SOURCEagentnet-partial-and-fail-v1
agentnet-partial-and-fail-v1
GUI state transitions (s, a, s') walked on an Ubuntu desktop by
Qwen3.8-27B, from tasks taken from AgentNet and run inside OSWorld's
Docker environment.
Walks the judge ruled partial or failed. These are the larger half and, for a world model, the more useful one: a walk that did not finish still opened dialogs, switched tabs and changed settings, and each of those is a real transition. Measured over eighty-eight walks, a failure visits 13.5 distinct… See the full description on the dataset page: https://huggingface.co/datasets/gui-wm/agentnet-partial-and-fail-v1.agentnet-success-v1
agentnet-success-v1
GUI state transitions (s, a, s') walked on an Ubuntu desktop by
Qwen3.8-27B, from tasks taken from AgentNet and run inside OSWorld's
Docker environment.
Walks the judge ruled had finished the task. Use these to score an agent: the trajectory is a worked example, and action_target gives the element each step was aiming at, so an answer can be marked right by control rather than by pixel.
This repository is the one-condition pool. Tasks are drawn from the 623… See the full description on the dataset page: https://huggingface.co/datasets/gui-wm/agentnet-success-v1.Magic-RICH
Dataset Summary
Magic-RICH is a Chinese benchmark dataset for evaluating mobile GUI agents in realistic smartphone environments. It contains 4,000 step-level samples across four subsets, covering 17 categories and over 150 popular apps. Unlike many previous GUI datasets, Magic-RICH also includes special actions such as screenshot and long screenshot to better reflect real-world interactions. This dataset is designed for evaluation only (no train/dev split) and was used in the… See the full description on the dataset page: https://huggingface.co/datasets/GUIAgent/Magic-RICH.macos-gui-agent-trainGUI-Dense-Descriptions
GUI Screenshots - Dense descrptions Dataset
gui-agent-desktop-sft-dataGUI_agentkoen-web-agent-sft-mixes
Korean-English Web Agent SFT Mixes
브라우저 GUI 에이전트 SFT 용 한국어·영어 혼합 데이터. 언어 비율만 다르고 나머지는
동일하게 통제된 4개 구성이라, 비율이 성능에 미치는 영향을 직접 비교할 수 있다.
구성
config
ko : en
스텝
궤적
ko_only_20k
10 : 0
20,001
3,042
mix_ko_en_5050
5 : 5
20,008
2,663
mix_ko_en_2080
2 : 8
20,003
2,436
en_only_20k
0 : 10
20,009
2,274
비율은 궤적 수가 아니라 스텝 수 기준이다. 스텝 하나가 학습 샘플 하나인데
한국어 궤적은 평균 6.6스텝, 영어는 8.7스텝이라, 궤적 수로 5:5 를 맞추면 실제
gradient 기여가 5:5 가 되지 않는다. 궤적은 절대 쪼개지 않는다.
네 구성의 표본은 서로 중첩된다.… See the full description on the dataset page: https://huggingface.co/datasets/GUIAgentt/koen-web-agent-sft-mixes.