datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Cambrian-10M
Cambrian-10M Dataset
Please see paper & website for more information:
https://cambrian-mllm.github.io/
https://arxiv.org/abs/2406.16860
Overview
Cambrian-10M is a comprehensive dataset designed for instruction tuning, particularly in multimodal settings involving visual interaction data. The dataset is crafted to address the scarcity of high-quality multimodal instruction-tuning data and to maintain the language abilities of multimodal large language models (LLMs).… See the full description on the dataset page: https://huggingface.co/datasets/nyu-visionx/Cambrian-10M.Cambrian-Alignment
Cambrian-Alignment Dataset
Please see paper & website for more information:
https://cambrian-mllm.github.io/
https://arxiv.org/abs/2406.16860
Overview
Cambrian-Alignment is an question-answering alignment dataset comprised of alignment data from LLaVA, Mini-Gemini, Allava, and ShareGPT4V.
Getting Started with Cambrian Alignment Data
Before you start, ensure you have sufficient storage space to download and process the data.
Download the Data Repository… See the full description on the dataset page: https://huggingface.co/datasets/nyu-visionx/Cambrian-Alignment.tw-drug-labels-vision
Dataset Card for tw-drug-labels-vision
💊 tw-drug-labels-vision 是一份涵蓋臺灣食品藥物管理署（TFDA）核發之 44,663 筆藥品仿單／外盒 的繁體中文多模態資料集。每一筆紀錄同時包含 PDF 全部頁面的渲染圖（WebP 多頁）以及一份依統一 17 欄 JSON Schema 抽取自原始藥品標示文件的結構化資料，可直接用於語言模型微調、視覺語言模型訓練、文件問答、藥品知識檢索、繁體中文醫藥 NLP 任務之素材。
Dataset Details
Dataset Description
本資料集源自臺灣 TFDA 公開的藥品許可證查詢系統。每筆紀錄對應一份藥品文件（仿單或外盒），原始為 PDF 圖檔形式。處理流程分為三階段：
下載：依據 20251222政府開放資料集_仿單與藥品外盒_66032.xlsx 中的 PDF URL，下載原始檔。
頁面渲染：將 PDF 各頁渲染為 WebP 圖檔，封裝在 images 欄位中。
OCR +… See the full description on the dataset page: https://huggingface.co/datasets/twinkle-ai/tw-drug-labels-vision.agentic-llm-pretraining-1.7b
Agentic LLM Pretraining Dataset
A pretraining corpus for small language models (1-3B parameters) optimized for agentic tasks. The corpus emphasizes learning to comprehend language, reason, follow instructions, and use tools over memorizing factual knowledge — the assumption is that domain knowledge will be provided at runtime via RAG. The idea is that this could enable much smaller pretraining corpora by omitting the large volumes of text typically needed to memorize facts.… See the full description on the dataset page: https://huggingface.co/datasets/visionscaper/agentic-llm-pretraining-1.7b.vision-token-compression-bench
OPTIC-Bench
Optical Text In-Context Benchmark: how reliably do LLMs consume text
delivered as rendered images versus plain text tokens?
In summary, the evaluation reported here finds that optical text compression
is effective only within a narrow and specific envelope. Delivering content
as rendered images genuinely reduces input tokens, by thirteen to
fifty-four per cent depending on the model and the language, but only when
the document is long, the rendering is dense and the… See the full description on the dataset page: https://huggingface.co/datasets/translorentz/vision-token-compression-bench.Formosa-Vision
Dataset Card for Formosa-Vision
Formosa Vision 是一份以台灣在地文化為核心的開源視覺語言資料集，從國家文化記憶庫 2.0中精選兩千餘張資料，文字描述採用 OGDL 1.0 授權、及圖片為 CC By SA（及更開放的授權條款）授權的影像，內容涵蓋景點、建築、生活場景與歷史脈絡。資料集以模型生成與人工審核並行的方式建立，透過視覺語言模型產生影像對話，再由參與者逐一檢查與修訂，確保描述的正確性、文化脈絡的一致性與語句的自然性。專案由 Twinkle AI 社群發起，結合社群協作與開放文化精神，期待成為訓練繁體中文視覺語言模型的重要基礎，幫助研究者與開發者打造能真正理解台灣文化細節的 VLM 模型。
Dataset Details
Dataset Description
Formosa Vision（又稱 台灣視覺資料集）是一個以台灣在地視覺文化為核心、集結社群力量共創的開源資料集。這項專案源自近年視覺語言模型（Vision Language Model… See the full description on the dataset page: https://huggingface.co/datasets/twinkle-ai/Formosa-Vision.multimodal-vision-language-video-models-2026
👁️ Multimodal Vision-Language & Video Foundation Models Dataset (2026 Edition)
A structured research dataset featuring 1,000 domain-verified research papers and code repositories focused on Multimodal Vision-Language Models (VLM), Video Foundation Models, Diffusion Transformers (DiT), Visual Grounding, and World Simulators.
Built with Universal Scientific Engine V15.1 Gold, providing 47 schema attributes with verified repository attribution, modality capability matrix, vision… See the full description on the dataset page: https://huggingface.co/datasets/beatsprom/multimodal-vision-language-video-models-2026.Sujet-Finance-QA-Vision-100k
Dataset Description 📊🔍
The Sujet-Finance-QA-Vision-100k is a comprehensive dataset containing over 100,000 question-answer pairs derived from more than 9,800 financial document images. This dataset is designed to support research and development in the field of financial document analysis and visual question answering.
Key Features:
🖼️ 9,801 unique financial document images
❓ 107,050 question-answer pairs
🇬🇧 English language
📄 Diverse financial document types… See the full description on the dataset page: https://huggingface.co/datasets/sujet-ai/Sujet-Finance-QA-Vision-100k.unified-math-vision-dataset
Unified Math Vision Dataset Bundle
Generated at: 2025-09-19 17:12:44
This is a unified dataset bundle containing multiple math and vision reasoning datasets.
Dataset Statistics
Total samples: 15858
mathvision: 3344 samples
wemath: 500 samples
mmmu: 415 samples
mathvista: 6141 samples
logicvista: 448 samples
dynamath: 5010 samples
Contents
manifest.jsonl: Complete dataset in JSONL format (1 JSON per line)
manifest.csv: Summary in CSV format
images/: Directory… See the full description on the dataset page: https://huggingface.co/datasets/Haonian/unified-math-vision-dataset.HAERAE-VISION
HAERAE-VISION
A Korean visual QA benchmark featuring real-world, under-specified questions.
Dataset Description
This dataset includes two question types:
original: Under-specified, authentic user queries
explicit: Clarified queries with full context
Both share the same images and reference answers, allowing controlled evaluation of query under-specification.
Evaluation Code
See our GitHub repository for evaluation scripts.
Citation… See the full description on the dataset page: https://huggingface.co/datasets/HAERAE-HUB/HAERAE-VISION.GeoQA-train-Vision-R1-cot-rewrite
Dataset Card for GeoQA-train-Vision-R1-cot-rewrite
This dataset provides a rewritten version of the CoT (Chain-of-Thought) annotations for the GeoQA subset of the Vision-R1-cold dataset. It is designed to support efficient and structured multimodal reasoning with large language models.
Dataset Details
Dataset Description
The original Vision-R1 dataset, introduced in the paper Vision-R1: Reflective Multimodal Reasoning with Aha Moments, features detailed and… See the full description on the dataset page: https://huggingface.co/datasets/LoadingBFX/GeoQA-train-Vision-R1-cot-rewrite.dino-data-vision-tooling-preview
Dino Data Vision Tooling Preview
What This Dataset Is
This dataset is a focused vision-tooling preview built from two Dino Data capability slices:
image context understanding
image tooling
The goal is to train or inspect assistant behavior for image-related tasks where visual context, multimodal interpretation, or tool-aware image handling is relevant.
Included Capability Slices
Source lane
Public task name
What it teaches… See the full description on the dataset page: https://huggingface.co/datasets/DinoDS/dino-data-vision-tooling-preview.MATH-Vision-PT
MATH-Vision-PT
European Portuguese (pt-PT) machine translation of MATH-Vision, a benchmark of competition-level mathematics problems presented in visual contexts.
Translated from the original English test split using gemini-3.1-pro.
Original Dataset: https://huggingface.co/datasets/MathLLMs/MathVision
Note: This dataset is machine translated and may contain translation errors or artifacts.
This dataset is provided as part of the AMALIA project and is included in… See the full description on the dataset page: https://huggingface.co/datasets/amalia-llm/MATH-Vision-PT.GeoQA-PLUS-aug-train-Vision-R1-cot-rewrite
Dataset Card for GeoQA-PLUS-aug-train-Vision-R1-cot-rewrite
This dataset provides a rewritten version of the CoT (Chain-of-Thought) annotations for the GeoQA-PLUS subset of the Vision-R1-cold dataset. It is designed to support efficient and structured multimodal reasoning with large language models.
Dataset Details
Dataset Description
The original Vision-R1 dataset, introduced in the paper Vision-R1: Reflective Multimodal Reasoning with Aha Moments, features… See the full description on the dataset page: https://huggingface.co/datasets/LoadingBFX/GeoQA-PLUS-aug-train-Vision-R1-cot-rewrite.Sujet-Vision-QA
Dataset Description 📊🔍
The Sujet-Finance-QA-Vision-100k is a comprehensive dataset containing over 100,000 question-answer pairs derived from more than 9,800 financial document images. This dataset is designed to support research and development in the field of financial document analysis and visual question answering.
Key Features:
🖼️ 9,801 unique financial document images
❓ 107,050 question-answer pairs
🇬🇧 English language
📄 Diverse financial document types… See the full description on the dataset page: https://huggingface.co/datasets/1-800-SHARED-TASKS/Sujet-Vision-QA.CareQA-Vision
CareQA-Vision Dataset
Dataset Summary
CareQA-Vision is a vision-based healthcare QA dataset derived from the Spanish Specialized Healthcare Training (FSE) exams. All questions are curated by medical experts and cover the specialties of nursing and medicine. This dataset extends the original CareQA dataset by including image-based questions from exams conducted between… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/CareQA-Vision.gene-computer-vision-instruct
computer-vision-instruct v4
Gate-passed instruction data for computer-vision — published when 50 fresh examples cleared the quality bar
Kind: synthetic
Domain: computer-vision
Records: 177
Created: 2026-06-25T13:44:26+00:00
SHA-256: 60c3ff90c0668c54e691ed4f1ff2840584283f475e26358fa294a97882fb7767
Pipeline: v2.0.0
Filters: {"min_quality": 0.55, "limit": 1000, "source": null, "backend": "llama", "min_judge": 0.7}
Generated by: Qwen3-4B-Instruct-2507-Q4_K_M.gguf (backend:… See the full description on the dataset page: https://huggingface.co/datasets/Gene829/gene-computer-vision-instruct.Text-Before-VisionGitHub: https://github.com/MiliLab/Text-Before-Vision
brazilian-math-physics-qa-vision
Brazilian Math & Physics QA — Image Dependent
English | Português do Brasil
English
Summary
Brazilian Portuguese educational question-answer pairs whose problem statement or solution depends on one or more images.
Examples: 3,808
Referenced image URLs: 5,094 unique
Language: Brazilian Portuguese (pt-BR)
Schema
{"id":"vqa_...","subject":"matematica","category":"geometria","title":"...","messages":[{"role":"user","content":"...… See the full description on the dataset page: https://huggingface.co/datasets/artificialguybr/brazilian-math-physics-qa-vision.Text-Before-VisionGitHub: https://github.com/MiliLab/Text-Before-Vision
Legal_vision_finetuning_data
Sri Lankan Property Law Fine-Tuning Dataset
Dataset Summary
This dataset is a domain-specific legal instruction-tuning dataset designed for fine-tuning large language models for Sri Lankan property law reasoning and legal assistance.
It focuses on core areas of Sri Lankan property law, including:
Property transfer and conveyancing
Title registration (Bim Saviya)
Prescription and adverse possession
Partition of co-owned property
Mortgage and securities
Lease and tenancy… See the full description on the dataset page: https://huggingface.co/datasets/Sivanuja/Legal_vision_finetuning_data.OKReddit-Visionary
Dataset Summary
OKReddit Visionary is a collection of 50 GiB (~74K pairs) of image Question & Answers. This dataset has been prepared for research or archival purposes.
Curated by: KaraKaraWitch
Funded by: Recursal.ai
Shared by: KaraKaraWitch
Special Thanks: harrison (Suggestion)
Language(s) (NLP): Mainly English.
License: Refer to Licensing Information for data license.
Dataset Sources
Source Data: Academic Torrents by (stuck_in_the_matrix, Watchful1… See the full description on the dataset page: https://huggingface.co/datasets/recursal/OKReddit-Visionary.