datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MMScan-betaMMSearch-Plusmms-fa-alignmentsOmni_MMSI
Omni-MMSI: Toward Identity-attributed Social Interaction Understanding
Webpage |
arXiv |
YouTube |
Github
Xinpeng Li,
Bolin Lai,
Hardy Chen,
Shijian Deng,
Cihang Xie,
Yuyin Zhou,
James Matthew Rehg,
Yapeng Tian
Introduction
Omni-MMSI is an annotation dataset for identity-attributed social interaction understanding in multi-party social deduction game videos. It provides instruction-formatted metadata for clips, including labels, oracle cues, extracted cues… See the full description on the dataset page: https://huggingface.co/datasets/Xinpeng-Li/Omni_MMSI.MMSearch_Trainmmscifact-dev
MMSciFact
Multimodal scientific fact-checking benchmark: model-generated
question-answer pairs over scientific PDFs, with every answer sentence
human-annotated for role (Observation/Interpretation), dependency
structure (depends_on), and grounding label
(supported/contradiction/NEI/not_a_claim) against the source document.
40 QA pairs across 6 papers, 440 annotated sentences.
Every included QA pair has at least one contradiction/nei sentence
(see… See the full description on the dataset page: https://huggingface.co/datasets/alecocc/mmscifact-dev.MM-SY-BenchmarkMMSciFact
MMSciFact
Multimodal scientific fact-checking benchmark: model-generated
question-answer pairs over scientific PDFs, with every answer sentence
human-annotated for role, dependency structure, and grounding label
against the source document.
257 QA pairs across 45 papers, 2737 annotated sentences.
Files
mmscifact.jsonl — one row per QA pair.
pdfs.zip — one PDF per paper (pdfs/<paper_id>.pdf), referenced by
paper_pdf_path on every row.
More documentation (row… See the full description on the dataset page: https://huggingface.co/datasets/alecocc/MMSciFact.mmstar-qwen-sidecar-eval
MMStar Qwen Sidecar Eval
Local packaging of Lin-Chen/MMStar used for Qwen3-VL sidecar reproduction.
Standard splits:
validation.jsonl: full 1,500-sample split.
test.jsonl: first 1,000 samples used by the sidecar MMStar 55-point reproduction.
Compatibility files:
mmstar_val.jsonl: same content as validation.jsonl.
mmstar_eval_1k.jsonl: same content as test.jsonl.
images/: image files referenced by the JSONL image field.
The JSONL image paths are relative to the repository… See the full description on the dataset page: https://huggingface.co/datasets/huaXiaKyrie/mmstar-qwen-sidecar-eval.MMSQL
MMSQL
This repository contains the DATASET:MMSQL in paper "Evaluating and Enhancing LLMs for Multi-turn Text-to-SQL with Multiple Question Types" [Arxiv] [IJCNN2025]. The repository is structured to ensure the reproducibility of the experiments and includes scripts, notebooks, test suits, and data outputs. You can get an overview of this project and the paper through the page.
Construction
We used 🐦QDA-SQL. to generate a training set with 4 question types:… See the full description on the dataset page: https://huggingface.co/datasets/GuoZiming/MMSQL.MMSVG-Illustration-40k
SVG-Gen-70k
original dataset form OmniSVG/MMSVG-Illustration
Data format:
{
"id": "1234567890",
"svg": "<svg>...</svg>",
"description": "A beautiful illustration of a cat",
"description_length": 10, # word length of description
"element_count": 10, # number of elements in svg
"element_complexity": 10, # element complexity, scored by gemini-2.0-flash
"structural_complexity": 10, # structure complexity, scored by… See the full description on the dataset page: https://huggingface.co/datasets/SphereLab/MMSVG-Illustration-40k.multilinguality-m3-eval