datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
chapman-ecg-windows
Chapman-Shaoxing ECG Windows
Preprocessed from the Chapman-Shaoxing 12-lead ECG Database.
45,150 patients, 180,600 windows. Lead II, 128Hz, 4s windows (512 samples), 50% overlap.
Split
Windows
Patients
Train
126,420
31,605
Val
27,088
6,772
Test
27,092
6,773
Schema
Column
Type
Description
patient_id
string
Record ID (e.g. JS00001)
record_id
string
Record + window index
window
list[float32]
512-sample ECG (4s@128Hz, Lead II)… See the full description on the dataset page: https://huggingface.co/datasets/dheraingoud/chapman-ecg-windows.bnci-windows
EEG Dataset
This dataset was created using braindecode, a library for deep learning with EEG/MEG/ECoG signals.
Dataset Information
Number of recordings: 1
Number of channels: 26
Sampling frequency: 250.0 Hz
Data type: Windowed (from Epochs object)
Number of windows: 48
Total size: 0.04 MB
Storage format: zarr
Usage
To load this dataset:
from braindecode.datasets import BaseConcatDataset
# Load dataset from Hugging Face Hub
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Kkuntal990/bnci-windows.bnci-windows-test
EEG Dataset
This dataset was created using braindecode, a library for deep learning with EEG/MEG/ECoG signals.
Dataset Information
Number of recordings: 1
Number of channels: 26
Sampling frequency: 250.0 Hz
Data type: Windowed (from Epochs object)
Number of windows: 48
Total size: 0.04 MB
Storage format: zarr
Usage
To load this dataset:
from braindecode.datasets import BaseConcatDataset
# Load dataset from Hugging Face Hub
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Kkuntal990/bnci-windows-test.nvfp4-mtp-survey
Do Qwen3.8-27B NVFP4 repos actually ship a working MTP draft head?
A static survey of every NVFP4 quantization of Qwen3.8-27B and its finetunes that I could find
on the Hugging Face Hub, last run on 2026-08-24 (Rev 4) with
nvfp4_mtp_audit.py. Raw output: results.json.
I ran this to check a claim I had made in public, and the claim did not survive. The correction
is the first section, because it is the most important result here.
Revision history — read this, it is… See the full description on the dataset page: https://huggingface.co/datasets/windowsxp811203/nvfp4-mtp-survey.model-context-windows
LLM Context Windows — 206 models
Context-window sizes for 206 ready models served by the Qubax AI API (OpenAI-compatible), exported from the public /v1/models endpoint.
Columns
Column
Description
model_id
API model identifier
model_name
Display name
context_window_tokens
Max context window (tokens)
max_output_tokens
Max output (tokens, where published)
source
Provenance
Notes
License: CC0 1.0 (public domain) — use freely in… See the full description on the dataset page: https://huggingface.co/datasets/QubaxAI/model-context-windows.a2r-task-droid-bottle-to-windowsill
Schema v1 training admission: 0 / 1. All recordings are quarantined.
本仓库保留历史 LeRobot v2.1 演示供复核，数据字节与原读取版本保持一致。当前不满足完整原生 Schema，不进入正式训练索引。旧 10 Hz 策展时间轴的前后 20 帧不能直接作为原生两秒证明。逐条原因与原始来源。
把瓶子放到窗台
打开 LeRobot 查看器 · 场景对比页
1 条独立演示，10 Hz，共 133 帧/视角。本仓库只展示具体任务 droid_bottle_to_windowsill。原始英文指令逐字保留；同一任务的措辞可能不同。
机器人配置：Franka / DROID。多配置只为可视化组织，不代表动作空间已统一，不可据此直接跨机器人训练。
视频、原数值轨迹和裁剪边界沿用已有已发布版本；仅重编显示用的 episode/index/task_index。完整源录制仍保存在源数据集，来源定位见 annotations/sources.json；A2R 标段引用那些完整源录制。这里是… See the full description on the dataset page: https://huggingface.co/datasets/Travor278/a2r-task-droid-bottle-to-windowsill.test-windows
Dataset Card for "test-windows"
More Information needed
bi_so101_windows_15fps_mjpg_40ep_01This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_so_follower",
"total_episodes": 40,
"total_frames": 7953,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 15,
"splits": {
"train": "0:40"
},
"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/Moncif/bi_so101_windows_15fps_mjpg_40ep_01.a2r-task-droid-keys-windowsill-to-nightstand
Schema v1 training admission: 0 / 1. All recordings are quarantined.
本仓库保留历史 LeRobot v2.1 演示供复核，数据字节与原读取版本保持一致。当前不满足完整原生 Schema，不进入正式训练索引。旧 10 Hz 策展时间轴的前后 20 帧不能直接作为原生两秒证明。逐条原因与原始来源。
把钥匙从窗台移到床头柜
打开 LeRobot 查看器 · 场景对比页
1 条独立演示，10 Hz，共 79 帧/视角。本仓库只展示具体任务 droid_keys_windowsill_to_nightstand。原始英文指令逐字保留；同一任务的措辞可能不同。
机器人配置：Franka / DROID。多配置只为可视化组织，不代表动作空间已统一，不可据此直接跨机器人训练。
视频、原数值轨迹和裁剪边界沿用已有已发布版本；仅重编显示用的 episode/index/task_index。完整源录制仍保存在源数据集，来源定位见 annotations/sources.json；A2R… See the full description on the dataset page: https://huggingface.co/datasets/Travor278/a2r-task-droid-keys-windowsill-to-nightstand.bi_so101_windows_15fps_mjpg_20ep_01This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "bi_so_follower",
"total_episodes": 20,
"total_frames": 4009,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 15,
"splits": {
"train": "0:20"
},
"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/Moncif/bi_so101_windows_15fps_mjpg_20ep_01.thestack-v2-cpp-2011-windows-blamedaura-windows-pe-eval-v01
Traceix Mini Evaluation Dataset (Windows PE)
Traceix is a malware analysis platform that uses a neural network named AURA to classify files as safe or malicious. You can use Traceix at https://traceix.com.
This repository contains a mini evaluation dataset so that anyone can peer review AURA’s file-level classifications and recompute the basic metrics (accuracy, precision, recall, FPR, FNR) used in the Traceix model-quality page.
Each row includes:
sha256
true_label
predicted_label… See the full description on the dataset page: https://huggingface.co/datasets/PerkinsFund/aura-windows-pe-eval-v01.orbit2vec-windows
Orbit2Vec windows
Windows of daily orbital elements for 70,953 objects, the training data of jgalego/orbit2vec.
Configs
windows: 32-day windows of daily mean elements. a and b are the two windows of a pair (adjacent windows of one object for train; the earlier and later test windows for test), as flattened float32 arrays of 32 days by 7 elements. 302,056 train rows and 28,230 test rows.
objects: one row per object with its catalogue number, name, constellation or… See the full description on the dataset page: https://huggingface.co/datasets/jgalego/orbit2vec-windows.test-windows-refactored
EEG Dataset
This dataset was created using braindecode, a library for deep learning with EEG/MEG/ECoG signals.
Dataset Information
Number of recordings: 1
Number of channels: 26
Sampling frequency: 250.0 Hz
Data type: Windowed (from Epochs object)
Number of windows: 48
Total size: 0.04 MB
Storage format: zarr
Usage
To load this dataset:
from braindecode.datasets import BaseConcatDataset
# Load dataset from Hugging Face Hub
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Kkuntal990/test-windows-refactored.attacker-zero-windows-v1
Attacker Zero Windows v1
This dataset is a prescored local-window derivative of
OpAI-Bench1/OpAI-Bench for the
attacker-zero Verifiers environment.
Each row contains one human / AI-aided / human sentence window from OpAI-Bench:
[Previous]: human sentence
[TARGET]: AI-aided sentence
[Next]: human sentence
The dataset intentionally stores raw window fields and deterministic detector
scores, not prompts. The environment owns prompt rendering, action formatting,
turn logic, and… See the full description on the dataset page: https://huggingface.co/datasets/oliveirabruno01/attacker-zero-windows-v1.stack-v2-cpp-2011-windows-blamedstack-v2-cpp-2011-windowshuman-windows