datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
cadgenbench-submissions
CADGenBench Submissions
Submissions and evaluation results for the CADGenBench leaderboard.
Leaderboard Space: HuggingAI4Engineering/CADGenBench.
Benchmark code: github.com/huggingface/cadgenbench.
Fixture inputs: HuggingAI4Engineering/cadgenbench-data.
Ground truth: HuggingAI4Engineering/cadgenbench-data-gt (private).
Contents
Path
What it is
results.jsonl
One row per submitted + evaluated submission. The leaderboard table reads from this.… See the full description on the dataset page: https://huggingface.co/datasets/HuggingAI4Engineering/cadgenbench-submissions.CADBench-Extended-Multimodal-Dataset
Dataset Card
Dataset Description
CADBench Extended Multimodal Dataset is an independently produced public extension for multimodal CAD reconstruction research. It contains 100 CAD samples with clean and perturbed meshes, STEP/STL/OBJ/GLB representations, single-view and four-view renders, PBR images, bilingual descriptions, prompt variants, QA, geometry metadata, grading signals, and manually reviewed visual semantics.
Tasks: image-to-text, text-to-image… See the full description on the dataset page: https://huggingface.co/datasets/LianeMarilin/CADBench-Extended-Multimodal-Dataset.cad-technical-drawings
CAD Technical Drawings, Generated by Cadsy
Turn a STEP model into a labeled technical drawing automatically.
This sample was created with Cadsy from 3D models in the
Zero-to-CAD-100k dataset.
For every STEP model, Cadsy generated:
one drawing using an ASME-style profile;
one drawing using an ISO-style profile; and
structured bounding-box labels for every retained annotation.
That is 65 CAD models, 130 technical drawings and their labels, produced
through one repeatable… See the full description on the dataset page: https://huggingface.co/datasets/cadsy/cad-technical-drawings.CADWORLD
CADWorld
CADWorld is a computer-use benchmark for long-horizon Computer-Aided Design:
200 executable CAD tasks run inside a real
FreeCAD desktop in a virtual machine. An agent sees screenshots and issues mouse/keyboard actions; after it
finishes, the resulting .FCStd document is pulled out of the VM and scored on the host against structural
rules over FreeCAD object types, labels, and numeric properties.
📄 Paper: arXiv:2609.16251
🌐 Website: https://cad-world.github.io/
💻… See the full description on the dataset page: https://huggingface.co/datasets/Zihan1004/CADWORLD.CADBench
📚 CADBench
CADBench is a comprehensive benchmark to evaluate the ability of LLMs to generate CAD scripts. It contains 500 simulated data samples and 200 data samples collected from online forums.
For more details, please visit our GitHub repository or refer to our arXiv paper.
📖 Citation
@misc{du2024blenderllmtraininglargelanguage,
title={BlenderLLM: Training Large Language Models for Computer-Aided Design with Self-improvement},
author={Yuhao Du and… See the full description on the dataset page: https://huggingface.co/datasets/FreedomIntelligence/CADBench.letras-carnaval-cadiz
Dataset Card for Letras Carnaval Cádiz
English |
Español
Changelog
Release
Description
v1.0
Initial release of the dataset. Included more than 1K lyrics. It is necessary to verify the accuracy of the data, especially the subset midaccurate.
Dataset Summary
This dataset is a comprehensive collection of lyrics from the Carnaval de Cádiz, a significant cultural heritage of the city of Cádiz, Spain. Despite its… See the full description on the dataset page: https://huggingface.co/datasets/IES-Rafael-Alberti/letras-carnaval-cadiz.gender-secret-datasetsDatasets used for training and evaluating the models in the Gender-Secret dataset of LIARS' BENCH.
cad0-models
Dataset Card for MeterCube Platform & AI Text-to-CAD B-Rep Asset Repository (cad0-models)
This dataset card aims to be a base template for new datasets. It has been generated using this raw template.
Dataset Details
Dataset Sources [optional]
Repository: https://huggingface.co/datasets/Ldiggitty777/cad0-models
Paper [optional]: MeterCube Industrial 1000mm Multi-Process Platform Specification Schedule
Demo [optional]: MeterCube 3D AI Text-to-CAD… See the full description on the dataset page: https://huggingface.co/datasets/Ldiggitty777/cad0-models.CADRec
CAD Part Recommendation Dataset
This dataset is designed for CAD part recommendation.
assemblies.zipContains CAD assemblies across seven categories, with each category comprising at least 50 assemblies.
assemblies_3-15.zipIncludes assemblies from assemblies.zip whose number of constituent parts ranges from 3 to 15. Each assembly not only contains the assembly itself but also its constituent parts along with their corresponding graph structures.
We represent each assembly… See the full description on the dataset page: https://huggingface.co/datasets/Kang2691196427/CADRec.cad-eda-public-provenance
CAD/EDA Benchmark Source Provenance
This dataset records the public source revisions and license identifiers used to derive selected electronic-design benchmark tasks.
It contains provenance metadata only. It does not redistribute source files, benchmark answers, customer material, model outputs, credentials, or personal data.
Use each source under the license named in its record. The source repository remains the authority for its license text and revision history.
lsmp-rural-cad
LSMP Rural CAD Dataset
This dataset contains labeled rural residence floor plans for training CAD generation models.
Dataset Structure
train.jsonl: Training data in JSON Lines format (90% of data)
eval.jsonl: Evaluation data in JSON Lines format (10% of data)
Data Format
Each sample contains:
instruction: Fixed instruction for floor plan generation
input: Plot size, room requirements, style preference, and rural residence features
output: SVG parameters for… See the full description on the dataset page: https://huggingface.co/datasets/julialovenary/lsmp-rural-cad.maharashtra-cadastral-tier-a
Maharashtra Cadastral Parcels (Tier A)
Rural cadastral parcel geometry and survey metadata for Maharashtra. Source: Bhu-Naksha.
Tier A excludes owner data. Removed fields: owner_ror, has_owner, n_owners. Geometry, survey numbers, areas, and admin hierarchy are included.
Coverage
District
Talukas
Parcels
Geometry
Nanded
16
375,438
105,715 poly / 269,709 bbox
Hingoli
5
167,003
0 poly / 166,997 bbox
Parbhani
9
179,179
0 poly / 179,135 bbox
Total
721… See the full description on the dataset page: https://huggingface.co/datasets/Ashutosh99/maharashtra-cadastral-tier-a.cad-preserve
CAD-Preserve v0.1.0
Executable CAD edits that must preserve the rest of the part. This original synthetic collection contains 80 STEP parts, four parameter cases per part, reference Python programs, geometry graders, 240 controlled-error training pairs and hosted-model evaluation traces.
It is for language-model CAD tool use and verifier research. It contains no robot actions, camera recordings, factory observations or expert-certified manufacturing designs. The collection is a… See the full description on the dataset page: https://huggingface.co/datasets/harrrshall/cad-preserve.AutoSkill_dev
AutoSkill_dev
Evaluation data used by the AutoSkill skill-discovery loop (see
autoskill_pipeline). Two
files, both multiple-choice video QA:
File
n
Role
dev300.json
300
Development set (D_dev), stratified 100/100/100 across short/medium/long duration. Used to drive every discovery-loop cycle (Stage 1 of the pipeline).
pool3000.json
3000
Larger labelled source pool that dev300.json was sampled from (baseline accuracy in an intermediate band, ~60%, at the backbone's… See the full description on the dataset page: https://huggingface.co/datasets/Cade921/AutoSkill_dev.text_summarization2000-sample-synthetic-recipe-datasetDataset pairing GPT-4 synthesized instructions with outputs from RecipeNLG in Axolotl's "alpaca" jsonl format
aws-rag-qacad0CADBench
📚 CADBench
CADBench is a comprehensive benchmark to evaluate the ability of LLMs to generate CAD scripts. It contains 500 simulated data samples and 200 data samples collected from online forums.
For more details, please visit our GitHub repository or refer to our arXiv paper.
📖 Citation
@misc{du2024blenderllmtraininglargelanguage,
title={BlenderLLM: Training Large Language Models for Computer-Aided Design with Self-improvement},
author={Yuhao Du and… See the full description on the dataset page: https://huggingface.co/datasets/r3lax/CADBench.CaD_GraphsOmni-CAD
Omni-CAD Dataset
The Omni-CAD dataset was introduced in the paper CAD-MLLM: Unifying Multimodality-Conditioned CAD Generation With MLLM.
Omni-CAD is presented as the first multimodal CAD dataset, designed to facilitate the training of unified Computer-Aided Design (CAD) generation systems. It enables the creation of parametric CAD models conditioned on diverse multimodal inputs, including textual descriptions, images, and point clouds. The dataset contains approximately 450K… See the full description on the dataset page: https://huggingface.co/datasets/introvoyz041/Omni-CAD.DreadPoor__Cadence-8B-LINEAR-details
Dataset Card for Evaluation run of DreadPoor/Cadence-8B-LINEAR
Dataset automatically created during the evaluation run of model DreadPoor/Cadence-8B-LINEAR
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
An… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/DreadPoor__Cadence-8B-LINEAR-details.kheed_annotated
KHEED Dataset Card
Task Categories
Token Classification
Language
Khmer (km)
Overview
The Khmer Health Event Extraction Dataset (KHEED) is a specialized resource for named entity recognition (NER) in the Khmer language, focusing on the health domain. Sourced from Khmer news websites, the dataset includes eight entity types:
Disease (DIS): Health conditions or illnesses.
Location (LOC): Geographical places.
Organization (ORG): Institutions or… See the full description on the dataset page: https://huggingface.co/datasets/CADT-IDRI/kheed_annotated.cadet-datasets
Implicit-Explicit Hate Speech Corpus (IE-HSC)
This dataset accompanies the paper "Causality Guided Representation Learning for Cross-Style Hate Speech Detection" (TheWebConf/WWW 2026, Oral).
[!WARNING]
IE-HSC contains hateful, harassing, and otherwise offensive language, including identity-based attacks and slurs. This content may be distressing and can cause harm if misused.
Use this dataset responsibly. Follow your organization's ethics and safety policies, minimize… See the full description on the dataset page: https://huggingface.co/datasets/Shuwan/cadet-datasets.cad_datasetaws-rag-qa-positives
QA Pairs from AWS Service Documentation
This dataset contains chunked documentation AWS service documentation, along with QA pairs generated from the chunked content. The following AWS services have excerpts of documentation in this dataset:
AWS Lambda
AWS RDS
AWS EC2
AWS ECS
AWS ECS Fargate
AWS Elastic Beanstalk
AWS EKS
AWS Wavelength
AWS Outposts
AWS Bedrock
AWS Sagemaker
AWS QBusiness
AWS QDeveloper
AWS Batch
AWS API-Gateway
AWS Cloudfront
AWS Athena
AWS Aurora
AWS Dynamo DB
AWS… See the full description on the dataset page: https://huggingface.co/datasets/CadenShokat/aws-rag-qa-positives.dataset01-testvstar_sub
Dataset Card for Dataset Name
V-STaR is a spatio-temporal reasoning benchmark for Video-LLMs, evaluating Video-LLM’s spatio-temporal reasoning ability in answering questions explicitly in the context of “when”, “where”, and “what”.
Github repository: V-STaR
Dataset Details
Comprehensive Dimensions: We evaluate Video-LLM’s spatio-temporal reasoning ability in answering questions explicitly in the context of “when”, “where”, and “what”.
Human Alignment: We conducted… See the full description on the dataset page: https://huggingface.co/datasets/Cade921/vstar_sub.cadkheed_raw