datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
gpqa_subtask
GPQA Subtask
This is an splitted version of the GPQA dataset, where different domains and subdomains are in different files.
Samples per subdomain in main:
biology 78
physics 187
chemistry 183
Samples per subdomain in diamond:
physics 86
chemistry 93
biology 19
Samples per subdomain in extended:
biology 105
physics 227
chemistry 214
Dataset Card for GPQA
GPQA is a multiple-choice, Q&A dataset of very hard questions written and validated by experts in biology… See the full description on the dataset page: https://huggingface.co/datasets/PNYX/gpqa_subtask.libero_10_subtasksThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "libero_panda",
"total_episodes": 500,
"total_frames": 138090,
"total_tasks": 10,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 10.0,
"splits": {
"train": "0:500"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": null… See the full description on the dataset page: https://huggingface.co/datasets/KeWangRobotics/libero_10_subtasks.route_subtasks_dual_overhead_pi05
Route Subtask Episodes with Dual Overhead Views
This is a LeRobot v2.1 training dataset at 100 Hz. Every output episode contains
exactly one semantic subtask. Route boundaries are inferred from maximal
contiguous runs of metadata.route.active_subgoal; partial source episodes are
kept as the suffix subtasks they contain.
The dataset has 523 split route episodes and 135699 frames.
Camera streams
observation.images.video_overhead_vector: overhead image with the… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_subtasks_dual_overhead_pi05.route_plus_steer_place_subtasks_dual_overhead_pi05_224_plus_dagger
224px Route + Steering Placement + DAgger with Dual Overhead Views
This is a LeRobot v2.1 training dataset at 100 Hz. Every output episode contains
one constant prompt. Base route boundaries are inferred from maximal
contiguous runs of metadata.route.active_subgoal; partial source episodes are
kept as the suffix subtasks they contain.
The base data contains 495 route subtask episodes (128742 frames) and 202 steering/placement episodes (50099 frames). All 66 route DAgger clips… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_plus_steer_place_subtasks_dual_overhead_pi05_224_plus_dagger.route_subtasks_dual_overhead_pi05_448
Route Subtask Episodes with Dual Overhead Views
This is a LeRobot v2.1 training dataset at 100 Hz. Every output episode contains
one constant prompt. Base route boundaries are inferred from maximal
contiguous runs of metadata.route.active_subgoal; partial source episodes are
kept as the suffix subtasks they contain.
The dataset has 495 split route episodes and 128742 frames.
Camera streams
observation.images.video_overhead_vector: overhead image with the fixed… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_subtasks_dual_overhead_pi05_448.libero_10_image_subtaskroute_plus_steer_place_subtasks_dual_overhead_pi05
Route + Steering Placement Subtask Episodes with Dual Overhead Views
This is a LeRobot v2.1 training dataset at 100 Hz. Every output episode contains
exactly one semantic subtask. Route boundaries are inferred from maximal
contiguous runs of metadata.route.active_subgoal; partial source episodes are
kept as the suffix subtasks they contain.
The route portion has 495 split episodes and 128742 frames. The steering supplement contributes 202 episodes and 50099 frames without… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_plus_steer_place_subtasks_dual_overhead_pi05.NADI2026_subtask1.1_Robust_ASR
Dataset Card for "NADI2026_subtask1.1_Robust_ASR"
More Information needed
route_red_yellow_vector_subtasks_pi05
Route Red-Yellow Vector Subtasks for pi0.5
This is a LeRobot v2.1 transformation of DistantSky/route at commit aced1e96f6b8bf98ffaa0754407636e82442b084.
The dataset contains 212 real-robot episodes, 135699 frames, three cameras, and 14-dimensional actions at 100 Hz.
Conditioning
Only observation.images.video_overhead is modified. The left and right videos are byte-identical to the source dataset.
The overhead image receives one fixed selected-connector pose glyph… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_red_yellow_vector_subtasks_pi05.route_subtasks_dual_overhead_pi05_448_dagger_interventions_balanced
Route Subtasks + Balanced DAgger Interventions, 448px
This is a new LeRobot v2.1 dataset derived from
ajaysri/route_subtasks_dual_overhead_pi05_448 and a subsequent
DAgger collection. The original dataset is not modified.
The base contributes 495 episodes and 128,742 frames. Only
frames recorded while the human collector was actively intervening are added;
autonomous policy-control frames and policy_target_action are excluded from
the training targets. The DAgger action column… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_subtasks_dual_overhead_pi05_448_dagger_interventions_balanced.caselawqa-subtasks-8kNADI2026_subtask2_MixedASR
Dataset Card for "NADI2026_subtask2_MixedASR"
More Information needed
libero_10_subtasksNADI2026_subtask1.1_Robust_ASR_test
Dataset Card for "NADI2026_subtask1.1_Robust_ASR_test"
More Information needed
SemEval2024-Task8-SubtaskAUnofficial Mirror of SemEval2024-Task8-SubtaskA Dataset (https://github.com/mbzuai-nlp/SemEval2024-task8)
We have also separately extracted the Chinese subset from the multilingual dataset and stored it in this repository (only the training set includes the Chinese subset).
Our MAGA uses the monolingual test set and the Chinese training set of the SemEval2024-Task8-SubtaskA dataset as external test sets to verify the generalization ability of R-B MAGA.
openarm-restock-sequences-canonical-30fps-subtasks-gripper-vlm
restock-sequences-canonical-30fps
LeRobot v2.1 dataset: 226 episodes, 306218 frames at 30 fps.
Robot: openarm_bimanual
Cameras: observation.images.context, observation.images.wrist_left, observation.images.wrist_right
State/action dim: 16
Load it with the v2.1 tag, which is the revision the training path pins.
palmx_2025_subtask1_culture🏷️ PalmX 2025 — General Culture Evaluation (PalmX-GC)
Dataset Summary
PalmX-GC evaluates a model’s grasp of general Arab culture—customs, history, geography, arts, cuisine, notable figures, and everyday life across the 22 Arab League countries.
Every item is written in Modern Standard Arabic (MSA). The dataset powers Subtask 1 of the PalmX 2025 shared task.
Dataset Structure
Split
# MCQs
Release Date
Notes
Train
2000
10 Jun 2025
With gold answers
Dev
500
10… See the full description on the dataset page: https://huggingface.co/datasets/UBC-NLP/palmx_2025_subtask1_culture.AlexandriaX_Subtask_3
UBC-NLP/AlexandriaX_Subtask_3
This dataset contains the train/dev splits for AlexandriaX Subtask 3, a dialectal Arabic MT evaluation task.
Participants receive machine-translated outputs with source and reference-side information where applicable. The goal is to detect and classify translation errors using LQM-inspired annotations.
The task has two objectives:
Error span prediction: identify the exact word-level span in the translated text where an error occurs.
Error… See the full description on the dataset page: https://huggingface.co/datasets/UBC-NLP/AlexandriaX_Subtask_3.wgo_bench_lerobot_subtask_qwen_fullThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 10,
"features": {
"observation.images.cam": {
"dtype": "video",
"shape": [
384,
384,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height":… See the full description on the dataset page: https://huggingface.co/datasets/pepijn223/wgo_bench_lerobot_subtask_qwen_full.NADI2026_subtask3_TTS_testsuper_poulain_plan_subtasksThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "omx_follower",
"total_episodes": 50,
"total_frames": 32650,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:50"
},
"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/pepijn223/super_poulain_plan_subtasks.finnlp2026-subtask3-polyfiqa
FinNLP 2026 Subtask 3 — PolyFiQA
Files:
train.parquet: 76 public labeled PolyFiQA examples supplied by the organizers.
test.parquet: 76 hidden-answer participant examples with stable row IDs.
SOURCE_DATASET_CARD.md: source PolyFiQA dataset documentation and citation.
task_id identifies a source financial-report document and repeats four times. The added id column is the unique submission key. Test answers remain in the private competition repository.
route_red_yellow_vector_split_subtasks_example
Route Red-Yellow Vector Split-Subtask Example
This is a LeRobot v2.1 example derived from
ajaysri/route_red_yellow_vector_subtasks_pi05.
Source episode(s) 2 are split into one output episode per maximal
contiguous run of metadata.route.active_subgoal.
The example has 4 output episodes and 948 frames.
All three videos are cut on the same frame boundaries as the Parquet data and
re-encoded as H.264 at 100 Hz.
Why the split count is inferred
The complete 212-episode… See the full description on the dataset page: https://huggingface.co/datasets/ajaysri/route_red_yellow_vector_split_subtasks_example.finnlp2026-subtask1-greek-ner
FinNLP 2026 Subtask 1 — Greek Financial NER
Files:
train.parquet: public numeric-NER training data supplied by the organizers.
validation.parquet: public numeric-NER validation data.
test.parquet: 200 hidden-label participant examples with stable row IDs.
SOURCE_DATASET_CARD.md: source Plutus dataset documentation and citation.
The test set has 200 rows: 100 textual-NER and 100 numeric-NER prompts over the same 100 Greek financial passages. It contains only id, query, and… See the full description on the dataset page: https://huggingface.co/datasets/FinNLP-Multilingual-Understanding/finnlp2026-subtask1-greek-ner.wgo_bench_lerobot_subtask_bestThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 10,
"features": {
"observation.images.cam": {
"dtype": "video",
"shape": [
384,
384,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height":… See the full description on the dataset page: https://huggingface.co/datasets/pepijn223/wgo_bench_lerobot_subtask_best.s1_mix_plug_and_unplug_cable_subtask_round2_addNADI2026_subtask2_adi_testfinnlp2026-subtask2-japanese-icr
FinNLP 2026 Subtask 2 — Japanese Financial ICR
Files:
train.parquet: 253 public labeled examples.
test.parquet: 50 hidden-label participant examples.
SOURCE_DATASET_CARD.md: source JF-ICR dataset documentation.
The five labels are +2, +1, 0, -1, and -2. Test labels remain in the private competition repository.
strike_match_3_subtaskThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_nero_follower",
"total_episodes": 70,
"total_frames": 53794,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:70"
},
"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/ming326/strike_match_3_subtask.lerobot_ego_data_subtask