datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
tcga-wsi-uni2h-features
TCGA WSI UNI2H Features
Dataset Summary
This dataset provides tile-level UNI2-h embeddings extracted from TCGA whole-slide images (WSIs) using a reproducible, auditable pipeline designed for computational pathology research.
Data is organized by project (for example TCGA-HNSC) and currently exposes:
features/ containing H5 feature files with tile-level embeddings
vis/ containing overlay images for quality inspection and pipeline verification
[!IMPORTANT]
Unlike the… See the full description on the dataset page: https://huggingface.co/datasets/W8Yi/tcga-wsi-uni2h-features.openwakeword_featuresThis dataset contains precomputed audio features designed for use with the openWakeWord library.
Specifically, they are intended to be used as general purpose negative data (that is, data that does not contain the target wake word/phrase) for training custom openWakeWord models.
The individual .npy files in this dataset are not original audio data, but rather are low dimensional audio features produced by a pre-trained speech embedding model from Google.
openWakeWord uses these features as… See the full description on the dataset page: https://huggingface.co/datasets/davidscripka/openwakeword_features.behaviour1k-Qwen3-features
BEHAVIOR-1K Qwen3 skill features
Per-frame conditioned features e_t = Phi(f_t, L_sub^(j), L), mean-pooled primitive skill latents S_j, aligned proprioception q_t, actions a_t, and subtask progress p_t.
These are the inputs and targets for a Primitive Skill Composer VLA Skill Predictor.
Ground-truth primitives come from BEHAVIOR-1K's hand-authored
primitive_annotation, so the segmentation is human-labelled rather than
predicted, and nothing here depends on a keyframe detector.… See the full description on the dataset page: https://huggingface.co/datasets/erl-hub/behaviour1k-Qwen3-features.FeatureBench
FeatureBench: Agent Coding Evaluation Benchmark
Dataset Description
FeatureBench is a comprehensive benchmark designed to evaluate AI agents' capabilities in end-to-end feature-level code generation. Unlike traditional benchmarks that focus on function-level or algorithm-specific tasks, FeatureBench challenges agents to implement complete features within real-world software projects.
Key Characteristics
Feature-Level Tasks: Each task requires… See the full description on the dataset page: https://huggingface.co/datasets/LiberCoders/FeatureBench.bulk-cc12m-features
bulk-cc12m-features — ten teacher towers over CC12M, plus their consensus
Precomputed image-tower features for 10,968,539 CC12M images (all 2,176
shards of
pixparse/cc12m-wds)
from ten independent teacher extractions — eight CLIP variants across
three pretraining corpora and two model scales, SigLIP, and DINOv3 — plus
one derived consensus target.
About 110 million feature vectors, roughly 130 GPU-hours of extraction,
so that a student can be distilled against any of these… See the full description on the dataset page: https://huggingface.co/datasets/AbstractPhil/bulk-cc12m-features.tinygiant-omni-featureslivekit_wakeword_featuresThis dataset contains precomputed audio features designed for use with the openWakeWord library.
Specifically, they are intended to be used as general purpose negative data (that is, data that does not contain the target wake word/phrase) for training custom openWakeWord models.
The individual .npy files in this dataset are not original audio data, but rather are low dimensional audio features produced by a pre-trained speech embedding model from Google.
openWakeWord uses these features as… See the full description on the dataset page: https://huggingface.co/datasets/binhpham/livekit_wakeword_features.SN-Features
SoccerNet Features
Pre-extracted per-game features for the SoccerNet benchmark, structured as <league>/<season>/<game>/<file>, one file per game half (1_.../2_...).
This main branch holds no data — each feature type lives on its own branch so you only download what you need:
Branch
Files
Description
baidu-soccer-embeddings
{1,2}_baidu_soccer_embeddings.npy
Frame embeddings from baidu-research/vidpress-sports, used by the Action Spotting and Dense Video Captioning 2023… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-Features.FeatureCoding-DINOv3
DINOv3-ViT7B16: https://github.com/facebookresearch/dinov3/tree/54694f7627fd815f62a5dcc82944ffa6153bbb76
@misc{pang2026largemodelfeaturecoding,
title={Towards Large Model Feature Coding},
author={Youwei Pang and Changsheng Gao and Dong Liu and Huchuan Lu and Weisi Lin},
year={2026},
eprint={2605.24025},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2605.24025},
}
UNI2-h-features
Dataset Card for UNI2-h Pretrained Features
This dataset card provides the UNI2-h features for TCGA, CPTAC, and PANDA datasets with patch size 256 x 256 pixels at 20x magnification.
Requesting Access
As mentioned in the gated prompt, you must agree to the outlined terms of use, with the primary email for your HuggingFace account matching your institutional email. If your primary email is a personal email (@gmail/@hotmail/@qq) your request will be denied. To fix this, you… See the full description on the dataset page: https://huggingface.co/datasets/MahmoodLab/UNI2-h-features.Agri-LLaVA_Agricultural_Pests_And_Diseases_Feature_Alignment_Dataset
Agri-LLaVA
Agri-LLaVA is a large multimodal instruction dataset for agriculture, pairing crop/leaf images with multi-turn diagnostic conversations about plant diseases, pests, and nutrient deficiencies. It is compiled from 16 public source datasets (see the license table below).
This dataset has been converted to Parquet format with image bytes embedded directly, standardized to the HF image_text_to_text format with a single conversational messages schema.
This dataset is… See the full description on the dataset page: https://huggingface.co/datasets/Project-AgML/Agri-LLaVA_Agricultural_Pests_And_Diseases_Feature_Alignment_Dataset.ztf-dr3-m31-featuresMintaka_Graph_Features_T5-xl-ssm
Dataset Card for "Mintaka_Graph_Features_T5-xl-ssm"
More Information needed
embodied_features_and_demos_liberoDataset for Embodied Chain-of-Thought Reasoning for LIBERO-90, as used by ECoT-Lite.
TFDS Demonstration Data
The TFDS dataset contains successful demonstration trajectories for LIBERO-90 (50 trajectories for each of 90 tasks). It was created by rolling out the actions provided in the original LIBERO release and filtering out all unsuccessful ones, leaving 3917 successful demo trajectories. This is done via a modified version of a script from the MiniVLA codebase. In addition to… See the full description on the dataset page: https://huggingface.co/datasets/Embodied-CoT/embodied_features_and_demos_libero.vitra-dinotxt-featuresprecomputed_audio_featuresOmni-iEEG-Channel-featurefeatures-dinov3-vith16plus-224-imagenet-22k-wdswan-feature-cache-512x256features_mavos_completeucf-crime-clip-features360x_dataset_featuresvtg-featuresucf-crime-processed-featuresTissueMNIST-224-full-gpt5nano-with-vlm-features
TissueMNIST 224 Full Train Val with GPT-5-nano VLM Features
The full TissueMNIST train and validation splits with categorical morphology features generated by GPT-5-nano. Test is included as the full TissueMNIST passthrough split with null vlm_model_name and placeholder vlm_feature values for schema consistency.
This dataset is derived from the official MedMNIST TissueMNIST 224px data.
The VLM feature labels are categorical privileged-information annotations
for CS231N VLM-LUPI… See the full description on the dataset page: https://huggingface.co/datasets/Koa-Chang/TissueMNIST-224-full-gpt5nano-with-vlm-features.FeatureBench_Litetallyqa_extracted_featuresLlama-slideQA-Sample-FeaturesAlgonautsDS-features
Saved Features for Algonauts '25 Dataset
This repository contains pre-extracted features for the Algonauts Challenge dataset using baseline models.
Features Overview
The developer_kit directory contains features extracted for the entire dataset using the following models:
Video Features
Model: SlowFast R50
Extracts spatiotemporal features from video frames
Captures motion and appearance information
Audio Features
Model: MFCC (Mel-frequency… See the full description on the dataset page: https://huggingface.co/datasets/medarc/AlgonautsDS-features.mind-games-features
Mind Games — Features
Pre-computed visual features for the mind-games project.
Lunar Lander (MobileNetV3-Small)
200 episodes of MobileNetV3-Small embeddings extracted from LunarLander-v3 gameplay frames.
Backbone: MobileNetV3-Small (ImageNet pretrained, frozen)
Embedding dim: 576 (global average pooled)
Precision: float16
Size: ~108MB
Structure: lunar_lander/episode_NNNN/ with embeddings.npy (N×576) and actions.npy (N,)
Index: lunar_lander/index.json with per-episode… See the full description on the dataset page: https://huggingface.co/datasets/mad-bot/mind-games-features.