datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
instrument-imagesrobotic-instrument-segmentation-miccai-2017instrumental-choices-agent-traces
Instrumental Choices Agent Traces
This dataset contains the 1,680 agent trajectories reported in Instrumental Choices: Measuring the Propensity of LLM Agents to Pursue Instrumental Behaviors. It covers ten models, seven terminal-based tasks, eight variants, and three repetitions per task-variant-model combination.
Paper: arXiv:2605.06490
Project website: instrumentalchoices.com
Contents
inspect_logs/: 210 sanitised native Inspect .eval logs. Open locally with… See the full description on the dataset page: https://huggingface.co/datasets/aisa-group/instrumental-choices-agent-traces.Instrumental-LAION-Audio-300MWe sorted through the LAION-Audio-300M dataset to find music, trying to get instrumental tracks. We did this by looking at the text written about each audio file (in transcription and caption of its metadata).
A lot of the text was wrong, so we ended up with many clips that aren't music. But, the dataset is much smaller now. This makes it possible to do another, more careful round of cleaning.
Crypto-L2-Orderbook-Depth-24-Instruments
license: cc-by-4.0
task_categories:
time-series-forecasting
reinforcement-learning
📈 Hyperliquid Crypto L2 Orderbook Depth - 24 Instruments (Free Sample)
⚠️ NOTE: This is a 7-day FREE SAMPLE dataset. > For the full institutional-grade dataset featuring 12+ months of continuous history (~2.3 Million rows) for 24 crypto assets, visit ImbalanceLabs.com.
Dataset Summary
Standard OHLCV candles are a "liquidity illusion." They hide the true market intent, spread, spoofing… See the full description on the dataset page: https://huggingface.co/datasets/AdamAtractor/Crypto-L2-Orderbook-Depth-24-Instruments.instrument_timbre
Dataset Card for Chinese Musical Instruments Timbre Evaluation Database
The original dataset is sourced from the National Musical Instruments Timbre Evaluation Dataset, which includes subjective timbre evaluation scores using 16 terms such as bright, dark, raspy, etc., evaluated across 37 Chinese instruments and 24 Western instruments by Chinese participants with musical backgrounds in a subjective evaluation experiment. Additionally, it contains 10 spectrogram analysis reports for… See the full description on the dataset page: https://huggingface.co/datasets/ccmusic-database/instrument_timbre.pdmx-multi-instrument-synthesizedSynthesized version of https://github.com/pnlong/PDMX/ (public domain) filtered to only songs with two or more instruments
pdmx-multi-instrument-synthesizedA synthesized subset of the PDMX dataset containing only MIDI files with multiple instruments.
Captions are auto-generated with Qwen2 Audio.
Audio is licensed under CC0 (from PDMX). Captions are licensed under CC-BY.
so100_instrumentThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so100",
"total_episodes": 85,
"total_frames": 20906,
"total_tasks": 2,
"total_videos": 170,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:85"
},
"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/aaronsu11/so100_instrument.musical-instrumentssurgical-instruments-pick-placeThis 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/sajibpra/surgical-instruments-pick-place.musical-instruments-sound-datasetseparate_instruments_from_tray_combo_3This 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/aboardman/separate_instruments_from_tray_combo_3.mongolian_instruments_dataset
Mongolian Traditional Instruments Dataset for Music Source Separation
Diploma project: Music Source Separation Using Deep Learning: A Study on Modern and Mongolian Traditional Instruments
Бүтэц
raw/ — YouTube-аас татсан түүхий аудио (4 зэмсэг)
morin_khuur/ — 719 wav (морин хуур)
yatga/ — 511 wav (ятга)
limbe/ — 643 wav (лимбэ)
tovshuur/ — 484 wav (товшуур)
clean_morin_khuur_v2/ — Audio QC v2 pipeline-аар цэвэр solo морин хуур сегмент (~1 GB)
synth_dataset/ — Synthetic… See the full description on the dataset page: https://huggingface.co/datasets/Tuugu/mongolian_instruments_dataset.Benchmark-Instrumentsaudio-diffusion-instrumental-hiphop-256256x256 mel spectrograms of 5 second samples of instrumental Hip Hop. The code to convert from audio to spectrogram and vice versa can be found in https://github.com/teticio/audio-diffusion along with scripts to train and run inference using De-noising Diffusion Probabilistic Models.
x_res = 256
y_res = 256
sample_rate = 22050
n_fft = 2048
hop_length = 512
InstrumentSoundDataset1aaronsu11_so100_instrument
SO100 Instrument TsFile
This dataset is an Apache TsFile conversion of
aaronsu11/so100_instrument, a LeRobot v2.1 SO100 robot-manipulation dataset
containing demonstrations for grabbing or picking up scissors and forceps.
Modalities: Time-series. The converted repository contains numeric robot state,
actions, frame timing, and episode/task tags. The two camera streams remain in
the original Hugging Face dataset and are linked below.
Source Dataset and Provenance… See the full description on the dataset page: https://huggingface.co/datasets/THULab/aaronsu11_so100_instrument.isc-gem-global-instrumental-earthquake-catalogue
Quick Links
Website
GitHub
Hugging Face
LinkedIn
X
ISC-GEM Global Instrumental Earthquake Catalogue, Modernized
74,158 earthquake events (1904-2021, magnitude 5.5+, plus continental
events down to magnitude 5.0) from the ISC-GEM Global Instrumental
Earthquake Catalogue, Version 12.1, compiled by the International
Seismological Centre. No new earthquake interpretation or machine-learning
modelling is introduced — locations, magnitudes, and… See the full description on the dataset page: https://huggingface.co/datasets/NoraResearchLab/isc-gem-global-instrumental-earthquake-catalogue.FTT_Musical_Instruments_trainInstrumentSoundDataset3InstrumentSoundDataset2cobe-firas-commanded-instrument-gains
COBE/FIRAS Actual Values of Commanded Instrument Gains
This table contains the actual commanded gains for the four COBE/FIRAS
bolometer channels. LAMBDA states that the values are used to normalize
interferograms and publishes the columns in the order RH, RL, LH, and
LL: right-high, right-low, left-high, and left-low.
Instrument
COBE Far Infrared Absolute Spectrophotometer (FIRAS)
Product
Actual Values of Commanded Instrument Gains
Structure
8 settings × 4… See the full description on the dataset page: https://huggingface.co/datasets/astro-legacy-archive/cobe-firas-commanded-instrument-gains.InstrumentSoundDataset4InstrumentSoundDataset5Instrumentinstrument_coversNS-Instruments
NS-Instruments
This is an audio classification dataset for Music Instruments Classification.
Classes = 10 , Split = Train-Test
Structure
audios folder contains audio files.
train.csv for training split and test.csv for the testing split.
Download
import os
import zipfile
import shutil
import huggingface_hub
audio_datasets_path = "DATASET_PATH/Audio-Datasets"
if not os.path.exists(audio_datasets_path): print(f"Given… See the full description on the dataset page: https://huggingface.co/datasets/MahiA/NS-Instruments.diffrhythm-instrument-cc0-oepngamearg-10x5-generated
What is this
A dataset of 50 instrumental music tracks generated with the DiffRhythm model, using 10 CC0-licensed instrument samples from OEPN Game Art.
Paper: https://huggingface.co/papers/2503.01183
Project page: https://nzqian.github.io/DiffRhythm/
Models
This dataset was created by running the DiffRhythm model on 2025 Mar 05 using a copy of the space available at https://huggingface.co/spaces/ASLP-lab/DiffRhythm.
The exact architecture (base or VAE) of the DiffRhythm… See the full description on the dataset page: https://huggingface.co/datasets/Akjava/diffrhythm-instrument-cc0-oepngamearg-10x5-generated.trilens-instrument-agreement
Tri-lens: do three interpretability instruments agree about the same activation?
Three 2025–26 instruments each claim to read what a model internally holds. Nobody had run
several on the same activations of the same model and asked whether they agree. This does,
with a mismatch null, and then makes it causal with steering as ground truth.
Model: google/gemma-3-27b-it, layer 41 — the only layer where a J-lens, an NLA
checkpoint, Gemma Scope 2 SAE weights, and Neuronpedia feature… See the full description on the dataset page: https://huggingface.co/datasets/mhough/trilens-instrument-agreement.