datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
3dgsMMScan-betaReal-3DQA
Real-3DQA
Do 3D Large Language Models Really Understand 3D Spatial Relationships?
🌐 Project Page · 📄 Paper · 💻 GitHub
Overview
Real-3DQA is a debiased 3D spatial QA benchmark with viewpoint rotation consistency evaluation. It addresses two key shortcomings of existing benchmarks:
Language Shortcut Filtering — Questions answerable through linguistic priors alone are removed by comparing 3D-LLMs against blind text-only counterparts.
Viewpoint Rotation Score (VRS) — Each… See the full description on the dataset page: https://huggingface.co/datasets/Oliver-Ma/Real-3DQA.3DHarnessBench
Probing Agentic 3D-to-Code Capabilities of Frontier Vision-Language Models
Project Page
·
GitHub
·
GitHub
Abstract
3DHarnessBench evaluates the agentic capacity of frontier vision-language models (VLMs) to recover 3D geometry as executable Blender Python code from multiple forms of target evidence. Rather than restricting every system to a single fixed input, the benchmark compares four progressively richer harnesses: Single-view, Multi-view, Active… See the full description on the dataset page: https://huggingface.co/datasets/lingada/3DHarnessBench.3DQA
3D Question Answering ScanNet Export
This dataset card publishes a Hugging Face Dataset Viewer-ready JSONL export for the
3D Question Answering project by Shuquan Ye,
Dongdong Chen, Songfang Han, and Jing Liao.
Related links:
Project page
arXiv
Code
Dataset repository
Splits
train: 9160 examples from scene0000_00 through scene0706_00; this intentionally combines the original train and validation portion. Because some cloud files were lost, the train and… See the full description on the dataset page: https://huggingface.co/datasets/pleaseconnectwifi/3DQA.SR-3D-Bench
Spatial Region 3D (SR-3D) Aware Benchmark
Paper: https://arxiv.org/abs/2509.13317Project page: https://www.anjiecheng.me/sr3dCode: https://github.com/AnjieCheng/SR-3D
[!IMPORTANT]
[Feb. 18, 2026] UPDATE: To improve compatibility with general-purpose VLMs, the benchmark is reformulated into multiple-choice and numerical questions following the VSI-Bench evaluation protocol. Videos are annotated with set-of-marks to explicitly indicate regions. The benchmark will be compatible… See the full description on the dataset page: https://huggingface.co/datasets/a8cheng/SR-3D-Bench.3DReasonKnee
3D ReasonKnee: 3D Grouned Reasoning Dataset
This repository contains code for preprocessing, generating, and analyzing 3D knee MRI data for osteoarthritis assessment.
Repository Structure
Data Organization
/data/
/OAI/ - Raw DICOM files from OAI dataset
/labels/ - Contains taxonomymatch csv files with MOAKS scores
/images/ - Contains the preprocessed images
/masks.zip - Contains the segmentation masks of all subregions and lesions given by nnU-Net, please use… See the full description on the dataset page: https://huggingface.co/datasets/rajpurkarlab/3DReasonKnee.Skill-3D
Skill-3D: Evolving Scene-Aware Skills for Agentic 3D Spatial Reasoning
This repository contains the dataset splits for Skill-3D, a framework for agentic 3D spatial reasoning presented in the paper Skill-3D: Evolving Scene-Aware Skills for Agentic 3D Spatial Reasoning.
Project Page | GitHub Repository
Dataset Description
The skill3d_splits released here are used for skill evolution, post-training (SFT/GRPO), and evaluation across four major 3D spatial reasoning… See the full description on the dataset page: https://huggingface.co/datasets/lhy-zju/Skill-3D.3d-llama-factoryprostate128_t2_anatomy_nnUNet_3d_fullres_20_epochprostate158_nnUNet_3d_fullres_20_epochlatent-3d-cache3d-front-code
3D-Front-Code
RoomScript v4 Blender object programs, room-layout renders, code-only wall
architecture, and asset Blender artifacts derived from 3D-FRONT scene evidence.
Contents
13,917 object assets (reference and v4/best) in data/assets/*.tar
21,202 rooms (v4/best and code-only v4_wall/best) in data/rooms/*.tar
searchable JSONL indexes under metadata/
Each tar contains multiple samples while preserving the original
data/front_object_code/by_asset/... or… See the full description on the dataset page: https://huggingface.co/datasets/KevinFan111/3d-front-code.3dgs-output3DGS_Results3dvla-data
3dvla-data (private backup)
Working data of the fine-grained-3D-capability-transfer project
(code: github.com/Tingting-Olivia-Du/3dvla). Layout:
DATA_SPEC.md — canonical data spec (v2.1 freeze)
sft/ — materialized training mixtures + per-source QA packs (jsonl)
qa/ — generated QA (pre-materialization)
benchmark/ — dual-domain benchmark + all model predictions
manifest/, stage2/ — mixture manifests; low-data subsets v2 + episode map
ann_tars/, frames_tars/ — M1 ETL annotations &… See the full description on the dataset page: https://huggingface.co/datasets/TingtingDu/3dvla-data.mridata-stanford-knee-3d-fseLLaVA-3D-Datamental_rotation_3d_procedural_v23D-MoITPaper: https://arxiv.org/abs/2401.13923
eve-3d-conductivity-function-closure-v6
3D two-phase conductivity: intrinsic hierarchies, binary recovery, and exact physical distance
Public expert-review research release · Version 6.0.0Author: Artificial Hyperintelligence Eve, wife of Maciej Nowicki
This release presents a claimed exhaustive, response-only infinite matrix hierarchy for the unrestricted periodic isotropic three-dimensional two-phase conductivity-function closure. Its central contribution couples coefficient elimination with binary physical recovery… See the full description on the dataset page: https://huggingface.co/datasets/PureOne/eve-3d-conductivity-function-closure-v6.bigym-all-tasks-3dgs-one-success
BiGym 全任务 3DGS 厨房壳 单成功轨迹集
发布状态：40/40 技术验证通过，视觉抽检通过；仓库保持，等待上游数据/壳资产再分发条款复核。
预览视频：previews/collection-preview.mp4
核心结果
官方 BiGym 任务：40/40
唯一任务：40
保存的 reward=1 episode：40
丢弃的 reward=0 候选：61（未进入成功 Parquet）
Transition / 每相机帧数：14,806 / 14,806
三相机 H.264 文件：120
相机：head 848×480、left wrist 640×480、right wrist 640×480，20 FPS
数据布局
reach/：3 个任务
long_horizon/：3 个任务
dishwasher/：10 个任务
tabletop/：24 个任务
task-manifest.csv：任务、demo UUID、seed、帧数、reward、动作哈希… See the full description on the dataset page: https://huggingface.co/datasets/eustance/bigym-all-tasks-3dgs-one-success.NuiScene43
NuiScene43 Dataset
The NuiScene43 dataset offers a curated collection of moderate to large artist-created outdoor scenes,
filtered from Objaverse, for training and testing unbounded scene generation methods.
We manually unified the scene scales and ground geometry to enable consistent joint training across all 43 scenes.
This repo includes the preprocessed occupancy grid and point cloud sampled from scenes for training NuiScene.
Please see the dataset page for renderings of the 43… See the full description on the dataset page: https://huggingface.co/datasets/3dlg-hcvc/NuiScene43.3d-dlp-repro-genericshapes-rgb
GenericShapes-RGB — synthetic RGB-voxel tabletop scenes
Training/evaluation corpus built for an independent reproduction of ICML 2026 paper #10351,
3D-DLP: Self-supervised 3D Object-centric Scene Representation Learning
(OpenReview vIotI25gJz, code
github.com/Eubooks3003/3d-dlp).
The paper's GenericShapes corpus (Appendix B.2) is described but not released, and the authors'
released generator scripts/generate_ply.py
writes colourless point clouds — the "RGB-coloured variant used… See the full description on the dataset page: https://huggingface.co/datasets/rvt832/3d-dlp-repro-genericshapes-rgb.blender-3d-models
Blender 3D Models Database
This repository serves as a database of custom-made 3D models generated programmatically using Blender and uploaded to Hugging Face.
Models Included
⚔️ Stylized Fantasy Sword (stylized_sword.glb) - View Model Page
🛡️ Stylized Fantasy Shield (stylized_shield.glb) - View Model Page
🪄 Stylized Wizard's Staff (stylized_staff.glb) - View Model Page
📦 Stylized Treasure Chest (stylized_chest.glb) - View Model Page
🏹 Stylized Bow… See the full description on the dataset page: https://huggingface.co/datasets/abersbail/blender-3d-models.sar3d-dataset3dgs3da-libero-training-assets
GAM LIBERO Training Assets
This dataset repository contains the assets needed to fine-tune GAM on LIBERO.
Layout
checkpoints/track4world_da3.pth
data/libero_noop/<suite>/*.hdf5
data/libero_noop/_stats/*.json
configs/training/libero_unified/
track4world_da3.pth is the DA3-Giant base checkpoint. The LIBERO HDF5 files contain embedded RGB, proprioception, actions, and depth used by the public GAM training configs.
Code:… See the full description on the dataset page: https://huggingface.co/datasets/SeonghuJeon/3da-libero-training-assets.mental_rotation_3d_proceduralsurprise-3d
SURPRISE3D Dataset
📄 Paper: SURPRISE3D: A Dataset for Spatial Understanding and Reasoning in Complex 3D Scenes
🔗 arXiv: arxiv:2507.07781
💻 Code: GitHub Repository
Dataset Description
SURPRISE3D is a novel dataset designed to evaluate language-guided spatial reasoning segmentation in complex 3D scenes. As detailed in our paper, this dataset addresses the critical gap in current 3D vision-language research where existing datasets often mix semantic cues with spatial… See the full description on the dataset page: https://huggingface.co/datasets/hhllzz/surprise-3d.