datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ConceptEdit-12M
ConceptEdit: Unlocking the Potential of Image Editing via Concept Scaling and Dense Supervision
ConceptEdit-12M is a large-scale image editing dataset. Each sample is stored as a triplet:
a source image,
an edited image,
a JSON metadata file describing the edit instruction, edit category, relative image paths, and VQA-style quality checks.
The dataset is packaged as multiple .tar shards. All paths inside the tar files and JSON files are relative paths; no… See the full description on the dataset page: https://huggingface.co/datasets/inclusionAI/ConceptEdit-12M.conceptual_captions
Dataset Card for Conceptual Captions
Dataset Summary
Conceptual Captions is a dataset consisting of ~3.3M images annotated with captions. In contrast with the curated style of other image caption annotations, Conceptual Caption images and their raw descriptions are harvested from the web, and therefore represent a wider variety of styles. More precisely, the raw descriptions are harvested from the Alt-text HTML attribute associated with web images. To arrive at the… See the full description on the dataset page: https://huggingface.co/datasets/google-research-datasets/conceptual_captions.rule-ling-conceptsconceptual-captions-12m-webdatasetconceptual-captions-12m-webdataset-metadata
Conceptual Captions 12M — Webshart metadata indices
Per-shard webshart metadata indices for
laion/conceptual-captions-12m-webdataset:
1,100 JSON files under data/, one per source tar shard, mirroring the source's shard layout.
Each index records every tar member's byte offset and length (enabling ranged reads without
downloading whole shards), image geometry (width/height for aspect bucketing), and — as of
August 2026 — embedded captions for all 10,994,853 samples, coalesced… See the full description on the dataset page: https://huggingface.co/datasets/webshart/conceptual-captions-12m-webdataset-metadata.conceptnet5
Dataset Card for Conceptnet5
Dataset Summary
ConceptNet is a multilingual knowledge base, representing words and
phrases that people use and the common-sense relationships between
them. The knowledge in ConceptNet is collected from a variety of
resources, including crowd-sourced resources (such as Wiktionary and
Open Mind Common Sense), games with a purpose (such as Verbosity and
nadya.jp), and expert-created resources (such as WordNet and JMDict).
You can browse what… See the full description on the dataset page: https://huggingface.co/datasets/conceptnet5/conceptnet5.fruit-vegetable-conceptsconceptual-captions-12m-webdataset-bertsOLMo-2_Residual_Streams
OLMo-2_Residual_Streams
This dataset contains approximately 600 million residual streams derived from the FineWeb dataset. The residual streams were extracted using the allenai/OLMo-2-1124-7B-Instruct model and are stored in .parquet format.
Dataset Description
Usage
The dataset should work just fine with load_dataset:
>>> from datasets import load_dataset
# after loading the data, cast to bf16 using torch.view()
>>> dataset =… See the full description on the dataset page: https://huggingface.co/datasets/open-concept-steering/OLMo-2_Residual_Streams.conceptual-12M-224 I resized the conceptual-12M dataset to 224x224 to make it easier to download.
The dataset contains a total of 11 million data points. The resize method is shown as the code in resize.py.
conceptual_captions_jsonconceptnet_en_simpleconceptual-12m-mbart-50-multilingualConceptGraph
Homepage
Exploring and Verbalizing Academic Ideas by Concept Co-occurrence
https://github.com/xyjigsaw/Kiscovery
Evolving Concept Co-occurrence Graph
It is the official Evolving Concept Co-occurrence Graph dataset of paper Exploring and Verbalizing Academic Ideas by Concept Co-occurrence.
To train our model for temporal link prediction, we first collect 240 essential and common queries from 19 disciplines and one special topic (COVID-19). Then, we enter these queries into… See the full description on the dataset page: https://huggingface.co/datasets/Reacubeth/ConceptGraph.CONCEPTUAL_CAPTIONS_HU_FILTEREDmsmarco-conceptsaxbench-concept500Concept500 for Supervised Dictionary Learning (SDL)
Concept500 contains training and inference data for 500 concepts randomly sampled from the released GemmaScope concept list for Gemma-2-2B-it at layers 10 and 20, and Gemma-2-9B-it at layers 20 and 31.
This is our main evaluation dataset used in the AxBench paper.
Each subset comes with the following format:
input: Instructions sampled from publicly available instruction-tuning datasets for three genres: text, code, and math.
output: Model… See the full description on the dataset page: https://huggingface.co/datasets/pyvene/axbench-concept500.conceptual-captions-cc12m-llavanext
Dataset Card for conceptual-captions-cc12m-llavanext
Dataset Summary
This is a data of 21,930,344 synthetic captions for 10,965,172 images from conceptual_12m. In the interest of reproducibility, an archive found here on Huggingface was used (cc12m-wds). The captions were produced using llama3-llava-next-8b inferenced in float16, followed by cleanup and shortening with Meta-Llama-3-8B.
Languages
The captions are in English.
Data Instances
An… See the full description on the dataset page: https://huggingface.co/datasets/CaptionEmporium/conceptual-captions-cc12m-llavanext.vrm-color-concept-550k
VRM Color Concept 550K
Summary
This is a dataset to train anime-style text-to-image or any text and image multimodal models without copyright/licensing concerns.
All assets/materials utilized in this dataset are CC0 or properly licensed, and no pretrained models or any AI models are used to build this dataset.
Image, Metadata and Dataset License
All images, metadata in this dataset and the dataset itself are licensed under CC BY-NC 4.0 by ELAN MITSUA Project… See the full description on the dataset page: https://huggingface.co/datasets/Mitsua/vrm-color-concept-550k.conceptual-captions-12This file contains English captions from Conceptual 12M dataset by Google. Since we don't own the images, we have provided the link to images, name of downloaded file, and caption for that image in the TSV file.
We would like to thank Luke Melas for helping us get the cleaned CC-12M data on our TPU-VMs.
Concept_Targeted_Causal_Images
Dataset Card for Concept-Targeted Causal Images
Dataset Summary
Concept-Targeted Causal Images is a concept-centric image dataset designed for studying causal visual representations in the brain. For each concept, the dataset contains three complementary image types:
Positive images that clearly depict the target concept
Semantic negatives that are visually or semantically related to the concept, but do not satisfy it
Counterfactual edits created by editing… See the full description on the dataset page: https://huggingface.co/datasets/BrainCause/Concept_Targeted_Causal_Images.conceptual_captions_3m_en_tinyconceptual_captionsconcept-guard
Dataset Card for ConceptGuard
Dataset Details
Dataset Description
ConceptGuard is a benchmark dataset for evaluating concept-level unlearning in Large Language Models. It is built around dual-use concepts, where each concept appears in both harmful and benign contexts. The dataset is designed to assess whether models can suppress harmful behavior while preserving useful knowledge, enabling evaluation of contextual separation.
Curated by: Authors… See the full description on the dataset page: https://huggingface.co/datasets/sk0511/concept-guard.photo-concept-bucket-wds-v2
Photo Concept Bucket (WebDataset Format)
Converted from bghira/photo-concept-bucket to a layered WebDataset format.
Stats
Metric
Value
Total samples
565,552
Image shards
114
Caption coverage
100%
Structure
├── images/ # Image TAR shards
│ ├── shard_00000.tar
│ └── ... (114 shards, ~600MB each)
├── annotations/ # Annotations
│ └── captions_v1.parquet # CogVLM captions from source
└── manifest.parquet… See the full description on the dataset page: https://huggingface.co/datasets/zengxianyu/photo-concept-bucket-wds-v2.conceptual-12m-multilingual-marianThis dataset is created from subset of Conceptual Captions. The original dataset has 12M captions but this dataset has around 10M image, caption pairs in different languages with 2.5M unique images. This dataset has captions translated from English to Spanish, German, French using language specific English to Marian models. Data distribution is following:
train_file_marian_final.tsv: 10010625 captions (2502656 captions of English, German, Spanish, French each)
val_file_marian_final.tsv:… See the full description on the dataset page: https://huggingface.co/datasets/flax-community/conceptual-12m-multilingual-marian.sample-page-assets
Sample-page assets
Files the cards of the Fluid Concepts sample datasets on the Hub (ToolTalk, Multimodal Expert Instruction, Multimodal Peer Collaboration)
need to show to visitors who have not requested access yet, which the gated sample repositories cannot serve themselves:
the card banners (*.png);
a public copy of each sample repository's TECHNICAL.md and TERMS.md, under the repository's name, refreshed on every push of that repository.
Nothing else is published here. The… See the full description on the dataset page: https://huggingface.co/datasets/fluid-concepts/sample-page-assets.conceptnet_en_nomalizedThis is the English part of the ConceptNet and we have removed the useless information.
bg_photo_concepts_bucketed_1024
bghira_photo_concepts_bucketed_1024
Edit this section before uploading:
Title: bghira_photo_concepts_bucketed_1024
Description: (add a short description / intended use)
Technical details
This repository contains a bucketed-shards export (uncompressed TAR shards).
Format
Format: bucketed_shards_v2
Created: 2026-04-30T11:47:20.761643+00:00
Export ID: export-2026-04-30T11:47:20.761643+00:00
Manifest: manifest.json
Image mode: reencode_jpeg… See the full description on the dataset page: https://huggingface.co/datasets/data-archetype/bg_photo_concepts_bucketed_1024.conceptual_captions_3m_zh_tiny