datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
urgent26_track1_universal_seThe pre-simulated universal speech enhancement training and validation set of the ICASSP 2026 URGENT speech enhancement challenge, Track1.
Please check our GitHub Repo and webpage for more details.
How to use:
Use tar to decompress the dataset
cat ./urgent26_track2_se_dataset.tgz.* | tar xzv
The pre-simulated dataset can be loaded by the PreSimulatedDataset in the URGENT 2026 baseline code.
Directory structure:
.
├── data
│ ├── train_simulation # train set… See the full description on the dataset page: https://huggingface.co/datasets/lichenda/urgent26_track1_universal_se.mls_hq_urgent_track1ADD2023_track12_test_r1
ADD 2023 — Track 1.2 (FG-D Detection), Round 1 Test · labels only
Benchmark-ready packaging of the Round 1 (R1) evaluation partition of Track 1.2
(Fake Game Detection / FG-D) from the ADD 2023 challenge — the Second Audio
Deepfake Detection Challenge (arXiv 2305.13774).
Binary anti-spoofing over Mandarin audio: bonafide (genuine human speech) vs.
spoof (synthesized / fake speech). Track 1.2 is the defending side of an
attack-and-defense game; the fakes are produced by the… See the full description on the dataset page: https://huggingface.co/datasets/SpeechAntiSpoofingBenchmarks/ADD2023_track12_test_r1.ELSA1M_track1
ELSA - Multimedia use case
ELSA Multimedia is a large collection of Deep Fake images, generated using diffusion models
Dataset Summary
This dataset was developed as part of the EU project ELSA. Specifically for the Multimedia use-case.
Official webpage: https://benchmarks.elsa-ai.eu/
This dataset aims to develop effective solutions for detecting and mitigating the spread of deep fake images in multimedia content. Deep fake images, which are highly realistic and deceptive… See the full description on the dataset page: https://huggingface.co/datasets/elsaEU/ELSA1M_track1.AT-ADD-Track1
AT-ADD Track 1
This repository hosts Track 1 of the AT-ADD All-Type Audio Deepfake Detection Challenge. It contains the released audio splits and privacy-preserving sample-level metadata for non-commercial academic research and education.
Access
This is a gated dataset. Sign in to Hugging Face, review the access agreement, complete the short access form, and click Agree and access dataset. Access is granted automatically after acceptance.
Direct repository access… See the full description on the dataset page: https://huggingface.co/datasets/xieyuankun/AT-ADD-Track1.mls-hq-urgent-track1
Multilingual LibriSpeech HQ (MLS-HQ)
This is a mirror of the Multilingual LibriSpeech HQ (MLS-HQ) data used in URGENT 2025 Track 1.
The original files were converted from FLAC to Opus to reduce the size and accelerate streaming.
Sampling rate: 48 kHz (resampled from 44.1 kHz to support Opus format)
Channels: 1
Format: Opus
Splits:
spanish: 150 hours, 36031 utterances
german: 150 hours, 35890 utterances
french: 150 hours, 36078 utterances
License: CC0 1.0
Source:… See the full description on the dataset page: https://huggingface.co/datasets/philgzl/mls-hq-urgent-track1.parc2026-track1-physical-rgb-20260909FloodNet-Challenge-EARTHVISION2021-Track1worldarena-track1-hz-wamn2c2_2018_track1Track 1 of the 2018 National NLP Clinical Challenges shared tasks focused
on identifying which patients in a corpus of longitudinal medical records
meet and do not meet identified selection criteria.
This shared task aimed to determine whether NLP systems could be trained to identify if patients met or did not meet
a set of selection criteria taken from real clinical trials. The selected criteria required measurement detection (
“Any HbA1c value between 6.5 and 9.5%”), inference (“Use of aspirin to prevent myocardial infarction”),
temporal reasoning (“Diagnosis of ketoacidosis in the past year”), and expert judgment to assess (“Major
diabetes-related complication”). For the corpus, we used the dataset of American English, longitudinal clinical
narratives from the 2014 i2b2/UTHealth shared task 4.
The final selected 13 selection criteria are as follows:
1. DRUG-ABUSE: Drug abuse, current or past
2. ALCOHOL-ABUSE: Current alcohol use over weekly recommended limits
3. ENGLISH: Patient must speak English
4. MAKES-DECISIONS: Patient must make their own medical decisions
5. ABDOMINAL: History of intra-abdominal surgery, small or large intestine
resection, or small bowel obstruction.
6. MAJOR-DIABETES: Major diabetes-related complication. For the purposes of
this annotation, we define “major complication” (as opposed to “minor complication”)
as any of the following that are a result of (or strongly correlated with) uncontrolled diabetes:
a. Amputation
b. Kidney damage
c. Skin conditions
d. Retinopathy
e. nephropathy
f. neuropathy
7. ADVANCED-CAD: Advanced cardiovascular disease (CAD).
For the purposes of this annotation, we define “advanced” as having 2 or more of the following:
a. Taking 2 or more medications to treat CAD
b. History of myocardial infarction (MI)
c. Currently experiencing angina
d. Ischemia, past or present
8. MI-6MOS: MI in the past 6 months
9. KETO-1YR: Diagnosis of ketoacidosis in the past year
10. DIETSUPP-2MOS: Taken a dietary supplement (excluding vitamin D) in the past 2 months
11. ASP-FOR-MI: Use of aspirin to prevent MI
12. HBA1C: Any hemoglobin A1c (HbA1c) value between 6.5% and 9.5%
13. CREATININE: Serum creatinine > upper limit of normal
The training consists of 202 patient records with document-level annotations, 10 records
with textual spans indicating annotator’s evidence for their annotations while test set contains 86.
Note:
* The inter-annotator average agreement is 84.9%
* Whereabouts of 10 records with textual spans indicating annotator’s evidence are unknown.
However, author did a simple script based validation to check if any of the tags contained any text
in any of the training set and they do not, which confirms that atleast train and test do not
have any evidence tagged alongside corresponding tags.vmc2026-track1-metavmc2026-track1-dev
vmc2026-track1-dev
Development subset of the VMC 2026 Track 1 data.
The data is organized into two configs corresponding to two subjective evaluation paradigms: absolute rating (acr) and pairwise comparison (ccr). The sample_id values are namespaced strings such as vmc2026-track1-dev-acr_489 and vmc2026-track1-dev-ccr_7233.
acr -- Absolute Category Rating
1,008 samples. Each row pairs a sample_id with one speech audio file, its released Mean Opinion Score (MOS)… See the full description on the dataset page: https://huggingface.co/datasets/urgent-challenge/vmc2026-track1-dev.x-wm-open-track1BrainStorm2026-Track1STFModel-v1.0-WorldArena2-Track1vmc2026-track1-test
vmc2026-track1-test
Test subset of the VMC 2026 Track 1 data.
The data is organized into two configs corresponding to two subjective evaluation paradigms: absolute rating (acr) and pairwise comparison (ccr). The sample_id values are namespaced strings such as vmc2026-track1-test-acr_4588 and vmc2026-track1-test-ccr_3061.
acr -- Absolute Category Rating
4,032 samples. Each row pairs a sample_id with one speech audio file, its released Mean Opinion Score (MOS)… See the full description on the dataset page: https://huggingface.co/datasets/urgent-challenge/vmc2026-track1-test.track1FiveOneWorld-wa2-track1Stellar-Motion-WA2-Track1
Stellar-Motion
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Stellar-Motion
Version: v2-rank08
Inference seed: 23
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Stellar-Motion_WA2_Track1_submission.tar.gz: submission archive
Stellar-Motion_WA2_Track1_submission.tar.gz.sha256: archive checksum… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Stellar-Motion-WA2-Track1.Zenith-Vision-WA2-Track1
Zenith-Vision
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Zenith-Vision
Version: v2-rank07
Inference seed: 17
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Zenith-Vision_WA2_Track1_submission.tar.gz: submission archive
Zenith-Vision_WA2_Track1_submission.tar.gz.sha256: archive checksum… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Zenith-Vision-WA2-Track1.Vertex-Orbit-WA2-Track1
Vertex-Orbit
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Vertex-Orbit
Version: v2-rank04
Inference seed: 211
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Vertex-Orbit_WA2_Track1_submission.tar.gz: submission archive
Vertex-Orbit_WA2_Track1_submission.tar.gz.sha256: archive checksum… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Vertex-Orbit-WA2-Track1.Xperience-0-open-track1Aurora-KineWorld-Prime-WA2-Track1
Aurora-KineWorld-Prime
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Aurora-KineWorld-Prime
Base model: frozen KineWorld W0
Inference variant: seed 73
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 × 480
Frames per video: 121
Frame rate: 24 FPS
Files
Aurora-KineWorld-Prime_WA2_Track1_submission.tar.gz: submission archive
Aurora-KineWorld-Prime_WA2_Track1_submission.tar.gz.sha256:… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Aurora-KineWorld-Prime-WA2-Track1.Prism-Flux-WA2-Track1
Prism-Flux
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Prism-Flux
Version: v2-rank05
Inference seed: 1
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Prism-Flux_WA2_Track1_submission.tar.gz: submission archive
Prism-Flux_WA2_Track1_submission.tar.gz.sha256: archive checksum
Prism-Flux_submission_media_manifest.json: per-video… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Prism-Flux-WA2-Track1.Fusion-Dynamics-WA2-Track1
Fusion-Dynamics
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Fusion-Dynamics
Version: v2-rank09
Inference seed: 7
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Fusion-Dynamics_WA2_Track1_submission.tar.gz: submission archive
Fusion-Dynamics_WA2_Track1_submission.tar.gz.sha256: archive checksum… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Fusion-Dynamics-WA2-Track1.Atlas-Nova-WA2-Track1
Atlas-Nova
WorldArena 2.0 Track 1 package submitted by Lei Yang, Tsinghua University.
Model name: Atlas-Nova
Version: v2-rank02
Inference seed: 3407
Contact: yanglei20@mails.tsinghua.edu.cn
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
This package is a stochastic inference variant of the frozen W0 checkpoint.
It is one of multiple transparently disclosed inference-seed variants of the
same base checkpoint, not an independently trained… See the full description on the dataset page: https://huggingface.co/datasets/yanglei18/Atlas-Nova-WA2-Track1.Quantum-Atlas-WA2-Track1
Quantum-Atlas
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Quantum-Atlas
Version: v2-rank10
Inference seed: 101
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Quantum-Atlas_WA2_Track1_submission.tar.gz: submission archive
Quantum-Atlas_WA2_Track1_submission.tar.gz.sha256: archive checksum… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Quantum-Atlas-WA2-Track1.Nexus-Pulse-WA2-Track1
Nexus-Pulse
WorldArena 2.0 Track 1 package submitted by Lei Yang, Tsinghua University.
Model name: Nexus-Pulse
Version: v2-rank03
Inference seed: 42
Contact: yanglei20@mails.tsinghua.edu.cn
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
This package is a stochastic inference variant of the frozen W0 checkpoint.
It is one of multiple transparently disclosed inference-seed variants of the
same base checkpoint, not an independently trained… See the full description on the dataset page: https://huggingface.co/datasets/yanglei18/Nexus-Pulse-WA2-Track1.Horizon-Core-WA2-Track1
Horizon-Core
WorldArena 2.0 Track 1 submission package from Nanyang Technological
University, Singapore.
Model name: Horizon-Core
Version: v2-rank06
Inference seed: 11
Contact: ziying.song@ntu.edu.sg
Number of videos: 1000
Resolution: 640 x 480
Frames per video: 121
Frame rate: 24 FPS
Files
Horizon-Core_WA2_Track1_submission.tar.gz: submission archive
Horizon-Core_WA2_Track1_submission.tar.gz.sha256: archive checksum
Horizon-Core_submission_media_manifest.json:… See the full description on the dataset page: https://huggingface.co/datasets/ZI-YING/Horizon-Core-WA2-Track1.JEWorld-3.1-WorldArena2-Track1