datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SLVMBench
SLVMBench: Skill Learning from Video Memory
[NeurIPS 2026 Datasets & Benchmarks Track]
Official Hugging Face Dataset repository for SLVMBench, a comprehensive benchmark designed to evaluate whether Video Large Language Models (video-LLMs) can acquire procedural skills from long video memory and apply them to real-time, ongoing tasks under heavy distractor noise.
📂 Repository Structure
To optimize download efficiency and bandwidth, this repository contains all… See the full description on the dataset page: https://huggingface.co/datasets/SLVMBench/SLVMBench.fineweb-2-slv-eduSLV-Set
SLV-Set
This repository releases the annotation portion of SLV-Set used in the SLVR paper.
What is included
slv_set: 387,039 region-grounded training examples derived from Visual-CoT.
slv_2q: 787,102 two-question training examples where each visual region is paired with two semantically different questions.
Data format
slv_set
Each row contains:
dataset: source dataset name.
split: split name.
question_id: example id.
image: relative image path… See the full description on the dataset page: https://huggingface.co/datasets/tinnel123/SLV-Set.deu_fin_slv_tur_Latn_traintenkgnad-clustering-p2pThis dataset can be used as a benchmark for clustering word embeddings for German.
The datasets contains news article titles and is based on the dataset of the One Million Posts Corpus and 10kGNAD. It contains 10'275 unique samples, 10 splits with 1'436 to 9'962 samples and 9 unique classes. Splits are built similarly to MTEB's TwentyNewsgroupsClustering.
Have a look at German Text Embedding Clustering Benchmark (Github, Paper) for more infos, datasets and evaluation results.
If you use this… See the full description on the dataset page: https://huggingface.co/datasets/slvnwhrl/tenkgnad-clustering-p2p.blurbs-clustering-s2sThis dataset can be used as a benchmark for clustering word embeddings for German.
The datasets contains book titles and is based on the dataset from the GermEval 2019 Shared Task on Hierarchical Classification of Blurbs. It contains 17'726 unqiue samples, 28 splits with 177 to 16'425 samples and 4 to 93 unique classes. Splits are built similarly to MTEB's ArxivClusteringS2S.
Have a look at German Text Embedding Clustering Benchmark (Github, Paper) for more infos, datasets and evaluation… See the full description on the dataset page: https://huggingface.co/datasets/slvnwhrl/blurbs-clustering-s2s.tenkgnad-clustering-s2sThis dataset can be used as a benchmark for clustering word embeddings for German.
The datasets contains news article titles and is based on the dataset of the One Million Posts Corpus and 10kGNAD. It contains 10'267 unique samples, 10 splits with 1'436 to 9'962 samples and 9 unique classes. Splits are built similarly to MTEB's TwentyNewsgroupsClustering.
Have a look at German Text Embedding Clustering Benchmark (Github, Paper) for more infos, datasets and evaluation results.
If you use this… See the full description on the dataset page: https://huggingface.co/datasets/slvnwhrl/tenkgnad-clustering-s2s.blurbs-clustering-p2pThis dataset can be used as a benchmark for clustering word embeddings for German.
The datasets contains book titles and is based on the dataset from the GermEval 2019 Shared Task on Hierarchical Classification of Blurbs. It contains 18'084 unqiue samples, 28 splits with 177 to 16'425 samples and 4 to 93 unique classes. Splits are built similarly to MTEB's ArxivClusteringP2P.
Have a look at German Text Embedding Clustering Benchmark (Github, Paper) for more infos, datasets and evaluation… See the full description on the dataset page: https://huggingface.co/datasets/slvnwhrl/blurbs-clustering-p2p.physicalaiThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/slvrfivo/physicalai.SLVMEval
Dataset Card for SLVMEval
Dataset Summary
SLVMEval (Synthetic Long-Video Meta-Evaluation Benchmark) is a benchmark for meta-evaluating automatic evaluation systems for text-to-long video (T2LV) generation.
The benchmark follows a pairwise comparison-based setup. It constructs controlled high-quality vs. low-quality long-video pairs by applying aspect-specific synthetic degradations to source videos.
The final benchmark data is built by retaining human-validated pairs… See the full description on the dataset page: https://huggingface.co/datasets/tohoku-nlp/SLVMEval.slvesnik-mk-sq
Службен весник MK–SQ Legal Parallel Corpus
293,612 Macedonian–Albanian sentence pairs from the Official Gazette of the
Republic of North Macedonia (Службен весник), 2001–2025.
from datasets import load_dataset
ds = load_dataset("EdonFetaji/slvesnik-mk-sq")
print(ds["test"][0]["mk_text"], ds["test"][0]["sq_text"])
Splits
The split is issue-disjoint: assignment happens at the level of the source
PDF, so no gazette issue contributes to more than one split. Gazette… See the full description on the dataset page: https://huggingface.co/datasets/EdonFetaji/slvesnik-mk-sq.slvhrreval_slvla_pick_cubesThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 1,
"total_frames": 872,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"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/hafeezjimoh/eval_slvla_pick_cubes.oscar_slv_Latn_trainspam.csvoscar_slv_Latn_deveval_slvaThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 0,
"total_frames": 0,
"total_tasks": 0,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {},
"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/zhangyuze999/eval_slva.oscar_slv_Latn_testslvqa
SLVQA — Streaming Long-Video QA, Perception-Test-anchored
10 × 24-hour streaming videos · 3,568 multiple-choice questions · 3 options each (chance = 33.3%).
SLVQA evaluates whether a model can (1) answer visual questions about a day-long video
that is revealed as a stream (the future is withheld), and (2) do so with O(1) query
latency — the time-to-first-token must not grow with how much video has already streamed.
Unlike LLM-generated video-QA sets, every question here is… See the full description on the dataset page: https://huggingface.co/datasets/treeleaves30760/slvqa.slv_videosuasSLvsi8EJ