datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
robotwin_aloha_clean_hdf5without depth data. processed data. generated from python scripts/process_data.py $task_name $setting $expert_data_num
robotwin-2.0-480p
RoboTwin 2.0 Clean-50 at 480p
This dataset contains 2,500 rerendered RoboTwin ALOHA-AgileX demonstrations:
50 tasks with 50 episodes per task. The head and left and right wrist cameras
use native 640×480 rendering. The auxiliary front camera retains its native
320×240 resolution. Images are JPEG-encoded; they are not upscaled.
The source demonstrations come from
TianxingChen/RoboTwin2.0.
Rerendering uses RoboTwin revision 6dde57155eafa3e4ebf6ad1f93a7cf7d5d41a755
and the… See the full description on the dataset page: https://huggingface.co/datasets/sai-prasanna/robotwin-2.0-480p.robotwin-aloha-clean
RoboTwin ICIL Aloha Clean
Training data for RoboTwin-ICIL, a
one-demonstration in-context imitation learning benchmark, in robotwin-icil-hdf5-multiview-v1 format.
Contents
50 tasks x 50 trajectories = 2,500 aloha-agilex trajectories, all of them training data. Each
episode has an HDF5 trajectory, three aligned 320x240 H.264 videos (head, left wrist, right wrist)
and an instruction JSON. The HDF5 holds joint commands, actions, end-effector poses, camera
calibration… See the full description on the dataset page: https://huggingface.co/datasets/robotensor/robotwin-aloha-clean.worldarena2026-robotwin-data
WorldArena-RoboTwin Dataset
Dataset: DavidxWang/worldarena2026-robotwin-dataProvider: davidwang (WorldArena Challenge 2026 Track 1 team)Last Updated: 2026-05-04Downstream Model: ABot-PhysWorld A2V VACE Training
Overview
This dataset provides action-conditioned robot manipulation videos synthesized by RoboTwin 2.0 simulation, formatted specifically for ABot-PhysWorld A2V (Action-to-Video) VACE training. It covers all 50 task categories from the CVPR 2026 WorldArena… See the full description on the dataset page: https://huggingface.co/datasets/DavidxWang/worldarena2026-robotwin-data.robotwin-eval-subset-v2-seen-unseen
RoboTwin frozen evaluation subset v2: seen + unseen
Default project evaluation subset from 2026-10-02 onward: 50 tasks × 2 settings × 10 episodes = 1,000 closed-loop episodes.
Group
Environment
Instruction pool
Episodes
clean2clean
clean
seen
500
clean2random
randomized
unseen
500
This repository publishes the immutable task specifications already released in GitHub PR #5. It is a development evaluation subset, not training demonstrations, simulator binaries… See the full description on the dataset page: https://huggingface.co/datasets/ar-mine/robotwin-eval-subset-v2-seen-unseen.openwam-robotwin-features
OpenWAM Robotwin Features
Features and similarity measurements extracted for the OpenWAM Robotwin
downstream evaluation.
Contents
agilex_temporal_features/: Agilex temporal DepthV3, DINOv2 and related features.
franka_temporal_features/: Franka temporal features.
cka/, cka_rbf/: similarity measurements.
manifests/: source manifests defining task, episode, video and frame alignment.
The temporal feature extraction alignment is 33 frames with stride 4. This… See the full description on the dataset page: https://huggingface.co/datasets/lishanyang/openwam-robotwin-features.robotwin_pill_rara
RoboTwin 2.0 move_pillbottle_pad: target demonstrations + retrieved pre-training subset (RARA, one task)
Everything needed to adapt a pre-trained Diffusion Policy to one RoboTwin 2.0 target task with FT, Co-FT and RARA,
packaged for reproduction on a single task.
path
content
dp_zarr/narrow_move_pillbottle_pad.zarr
50 target demonstrations (7,324 frames), demonstration configuration (object position range 0.5 of the task window)… See the full description on the dataset page: https://huggingface.co/datasets/ShuaKang/robotwin_pill_rara.robotwin_base