datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
robotwin_3d
RoboTwin 2.0 — 3D (RGB + Depth)
Bimanual manipulation data from the RoboTwin 2.0 simulator, in LeRobot v2.1 format, with per-camera ground-truth depth alongside RGB.
Tasks
50
Episodes
27,500 (550 per task, contiguous)
Frames
6,183,813
Robot
ALOHA-style bimanual, 14-DoF
Control rate
50 Hz
Cameras
3 (cam_high, cam_left_wrist, cam_right_wrist)
Resolution
240 × 320
Language instructions
1,039,891 unique corpus-wide; 100 entries per episode
Total size… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/robotwin_3d.self_repair_gripper_dagger
self_repair_gripper_dagger
Robot self-repair, DAgger rollouts with operator corrections on the same task as self_repair_gripper_bc.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
2,154
Frames
609,385
Duration
~5.6 h @ 30 fps
Tasks
1
Robot
yam (bimanual)
Cameras
cam_high… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/self_repair_gripper_dagger.sort_utensils
sort_utensils
Bimanual utensil sorting: pick up a utensil, set it on the pink plate, then place the other utensil in the container.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
152
Frames
128,010
Duration
~1.2 h @ 30 fps
Tasks
1
Robot
yam (bimanual)
Cameras
cam_high… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/sort_utensils.put_plate_on_the_rack
put_plate_on_the_rack
Pick plates off the table one by one and slot each into a dish rack.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
301
Frames
286,658
Duration
~2.7 h @ 30 fps
Tasks
1
Robot
yam (bimanual)
Cameras
cam_high, cam_left_wrist, cam_right_wrist
RGB
640×360… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/put_plate_on_the_rack.soft_bag_zipping
soft_bag_zipping
Deformable-object manipulation: unzip a soft bag, insert pens one at a time, then zip the bag closed.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
534
Frames
1,047,427
Duration
~9.7 h @ 30 fps
Tasks
1
Robot
yam (bimanual)
Cameras
cam_high, cam_left_wrist… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/soft_bag_zipping.self_repair_gripper_bc
self_repair_gripper_bc
Robot self-repair, human teleoperation (BC): install a gripper into an empty holder, drive a screw with a screwdriver, then clear the table.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
802
Frames
1,278,804
Duration
~11.8 h @ 30 fps
Tasks
1
Robot
yam… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/self_repair_gripper_bc.kitchen_organization
kitchen_organization
Multi-task kitchen tidying (3 tasks): placing plates, cups and spoons onto a dish rack, including a hand-to-hand pass.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
248
Frames
146,025
Duration
~1.4 h @ 30 fps
Tasks
3
Robot
yam (bimanual)
Cameras… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/kitchen_organization.sharpa-cylinder-rotation
Sharpa rotate dataset (format v1)
In-hand cylinder rotation trajectories collected in FlexPhysics with a frozen
SharpaWave ProprioAdapt policy. Each NPZ is one episode.
Repository layout and Hub preview
This repository contains eight cylinder-scale subsets: scale_0.400,
scale_0.429, scale_0.457, scale_0.486, scale_0.514, scale_0.543,
scale_0.571, and scale_0.600. Each currently contains 128 episode NPZs
(1,024 episodes in total). Folder names round the exact… See the full description on the dataset page: https://huggingface.co/datasets/FlexPhysics/sharpa-cylinder-rotation.stack-cubes-flexiv-0801This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_flexiv_rizon4_rt",
"total_episodes": 170,
"total_frames": 891288,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:170"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet"… See the full description on the dataset page: https://huggingface.co/datasets/Xense/stack-cubes-flexiv-0801.mod-arithmetic
Modular Arithmetic Dataset
Synthetic dataset of modular-arithmetic problems of the form a mod b,
paired with the result and a hypothesis about the most suitable tokenizer.
Tokenizer hypothesis
For a mod b where b = 2^k × 5^j (no other prime factors), only the
rightmost max(k, j) digits of a determine the answer, because
10^max(k,j) ≡ 0 (mod b). A tokenizer that groups digits right-to-left
in chunks of that size exposes the relevant information as a single token.
For all… See the full description on the dataset page: https://huggingface.co/datasets/flexitok/mod-arithmetic.librig2p-nostress-space-cmuGrapheme-to-Phoneme training, validation and test setslibrig2p-nostress-spaceGrapheme-to-Phoneme training, validation and test setsflexa-pour1-v3This is a FiftyOne dataset with 1 sample.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/flexa-pour1-v3")
# Launch the App
session = fo.launch_app(dataset)
Dataset Card for flexa-pour1 (FiftyOne, v3.0 conversion)
The… See the full description on the dataset page: https://huggingface.co/datasets/harpreetsahota/flexa-pour1-v3.FlexBench
FlexBench: Strictness-Adaptive Moderation Benchmark
Overview
FlexBench is a benchmark for evaluating strictness-adaptive content moderation under policy shifts. Each sample is annotated with a 5-tier risk severity label (BENIGN / LOW / MODERATE / HIGH / EXTREME). Following the accompanying paper, we derive three deployment-oriented binary classification tasks—strict, moderate, and loose—by… See the full description on the dataset page: https://huggingface.co/datasets/Tommy-DING/FlexBench.flexa-pour1
flexa-pour1
H1 humanoid + Shadow Hand replay of pouring task from iPhone egocentric capture.
Pipeline
iPhone rear camera (Record3D .r3d) → HaMeR 2D hand detection + MANO pose → Record3D LiDAR depth fusion → 3D wrist trajectories → Jacobian IK for H1 arms → DexMV finger retargeting → MuJoCo replay → LeRobot export
Data
Robot: Unitree H1 + Shadow Hand (48 DOF: 10 arm + 38 finger)
Task: Pouring
Episodes: 1
Frames: 294
FPS: 30
Format: LeRobot v2.1 (Parquet + MP4)… See the full description on the dataset page: https://huggingface.co/datasets/FlexaTechnologies/flexa-pour1.demo-plate-flex-washer-full-1This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"trossen_subversion": "v1.0",
"robot_type": "trossen_ai_solo",
"total_episodes": 10,
"total_frames": 7187,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits":{
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet"… See the full description on the dataset page: https://huggingface.co/datasets/andysigler/demo-plate-flex-washer-full-1.pick_and_place
Pick-and-place dataset (format v1)
Robot pick-and-place trajectories collected with a scripted task-space expert and joint-space PD control. Each NPZ is one episode. The current release contains one UR10 with Robotiq-85 dataset: 32 valid episodes, 32 successful episodes, and 32 rendered videos.
Repository layout and Hub preview
README.md
robots/<robot_id>/
README.md
metadata.json # Collection configuration, schema, ordering, and units… See the full description on the dataset page: https://huggingface.co/datasets/FlexPhysics/pick_and_place.multilingual-addition
Multilingual Addition Dataset
Synthetic dataset of addition problems of the form a+b=answer, where a
and b are written-form representations of integers in 21 languages, plus
a 22nd split using raw digit strings.
Task format
Each sample contains:
field
type
description
a_str
str
written-form (or digit) representation of a
a_digit
int
integer value of a
b_str
str
written-form (or digit) representation of b
b_digit
int
integer value of b
answer
str… See the full description on the dataset page: https://huggingface.co/datasets/flexitok/multilingual-addition.flex-damage-impactsfinemath-4plus-flexible-ordering.01.00_BATCHdemo-plate-flex-washer-full-2This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"trossen_subversion": "v1.0",
"robot_type": "trossen_ai_solo",
"total_episodes": 1,
"total_frames": 197,
"total_tasks": 1,
"total_videos": 2,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits":{
"train": "0:1"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/andysigler/demo-plate-flex-washer-full-2.bi_flexiv_rt_pick_and_placeThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_flexiv_rizon4_rt",
"total_episodes": 3,
"total_frames": 981,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:3"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/xensedyl/bi_flexiv_rt_pick_and_place.pick-mug-v5
Flexa Pick Mug V5
Robot training dataset generated from a single iPhone 15 Pro LiDAR recording of a human picking up a mug.
Pipeline
iPhone .r3d recording -> GroundingDINO (object detection) -> HaMeR (3D hand mesh) -> LiDAR depth fusion -> Spatial trajectory retargeting (V5 hybrid) -> IK solve -> LeRobot HDF5
Results
18.1cm clean mug lift in MuJoCo Franka Panda simulation
0 table collisions
93.7% hand pose coverage (282/301 frames)
Data Format… See the full description on the dataset page: https://huggingface.co/datasets/FlexaTechnologies/pick-mug-v5.flexiv_build_block_awbc_5_5flexa-stack2
flexa-stack2
H1 humanoid + Shadow Hand replay of bimanual book stacking from iPhone egocentric capture.
Pipeline
iPhone rear camera (Record3D .r3d) → HaMeR 2D hand detection + MANO pose → Record3D LiDAR depth fusion → 3D wrist trajectories → Jacobian IK for H1 arms → DexMV finger retargeting → MuJoCo replay → LeRobot export
Data
Robot: Unitree H1 + Shadow Hand (48 DOF: 10 arm + 38 finger)
Task: Bimanual book stacking
Episodes: 1
Frames: 476
FPS: 30
Format:… See the full description on the dataset page: https://huggingface.co/datasets/FlexaTechnologies/flexa-stack2.grab_flexiv_one_cube_1flexible_molecules_JCP2021
Cite this dataset Vassilev-Galindo, V., Fonseca, G., Poltavsky, I., and Tkatchenko, A. flexible molecules JCP2021. ColabFit, 2023. https://doi.org/10.60732/71f8031b
This dataset has been curated and formatted for the ColabFit Exchange
This dataset is also available on the ColabFit Exchange:
https://materials.colabfit.org/id/DS_i23sbm1o45sj_0
Visit the ColabFit Exchange to search additional datasets by author, description, element content and more.… See the full description on the dataset page: https://huggingface.co/datasets/colabfit/flexible_molecules_JCP2021.grab_flexiv_one_cube_3flexa-picknplace1
flexa-picknplace1
H1 humanoid + Shadow Hand replay of pick-and-place manipulation from iPhone egocentric capture.
Pipeline
iPhone rear camera (Record3D .r3d) → HaMeR 2D hand detection + MANO pose → Record3D LiDAR depth fusion → 3D wrist trajectories → Jacobian IK for H1 arms → DexMV finger retargeting → MuJoCo replay → LeRobot export
Data
Robot: Unitree H1 + Shadow Hand (48 DOF: 10 arm + 38 finger)
Task: Pick and place
Episodes: 1
Frames: 690
FPS: 30
Format:… See the full description on the dataset page: https://huggingface.co/datasets/FlexaTechnologies/flexa-picknplace1.FlexMS
NeurIPS 2026 mass spectrometry dataset
This repository was prepared automatically for Hugging Face Datasets.
Available configs
gnps_random
gnps_scaffold
mist_canopus_split_0
mist_canopus_split_1
mist_canopus_split_2
massbank_scaffold
massspecgym
Usage
from datasets import load_dataset
ds = load_dataset("your-username/your-dataset-name", "gnps_random")
Replace the repo id and config name as needed.
Notes
valid and val files are mapped to the… See the full description on the dataset page: https://huggingface.co/datasets/unifin/FlexMS.