datasetId large_stringlengths 6 121 | card_raw large_stringlengths 10 25.3M | card_text large_stringlengths 0 25.3M | downloads int64 0 2.26M | likes int64 0 9.39k | tags large listlengths 1 7.92k | created_at large_stringdate 2022-03-02 23:29:22 2025-11-12 17:47:45 | last_modified large_stringdate 2021-02-16 03:58:06 2025-11-12 17:57:42 | trending_score float32 0 90 |
|---|---|---|---|---|---|---|---|---|
chenglongy/env_codebase | <div align="center">
# SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Models (RSS 2025)
A spatial-enhanced vision-language-action model trained on 1.1 Million real robot episodes. 🤗
purely huggingFace-based, concise code with efficient performance.
> [Delin Qu*](https://github.com/DelinQu)<sup>1,2</sup>, [HaomingSong*](https://github.com/HaomingSong)<sup>1,3</sup>, [Qizhi Chen*](https://github.com/Tavish9)<sup>1,4</sup>, [Dong Wang†](https://scholar.google.com/citations?user=dasL9V4AAAAJ&hl=en)<sup>1</sup>, [Yuanqi Yao](https://scholar.google.com/citations?user=s482QHoAAAAJ&hl=zh-CN)<sup>1</sup>, [X. Ye](https://scholar.google.com/citations?user=GlYeyfoAAAAJ&hl=zh-CN)<sup>1</sup>, [Y. Ding](https://yding25.com)<sup>1</sup>, [Z. Wang](https://scholar.google.com/citations?user=cw3EaAYAAAAJ&hl=zh-CN)<sup>1</sup>, [Jiayuan Gu](https://cseweb.ucsd.edu/~jigu/)<sup>5</sup>, [Bin Zhao†](https://scholar.google.com/citations?hl=zh-CN&user=DQB0hqwAAAAJ)<sup>1</sup>, [Xuelong Li](https://scholar.google.com/citations?user=ahUibskAAAAJ)<sup>1,6</sup>
> Shanghai AI Laboratory<sup>1</sup>, Fudan University<sup>2</sup>, Shanghai Jiao Tong University<sup>3</sup>, Zhejiang University<sup>4</sup>, ShanghaiTech<sup>5</sup>, TeleAI<sup>6</sup>
[\[📄Paper\]](https://arxiv.org/pdf/2501.15830) [\[🔥Project Page\]](https://spatialvla.github.io/) [\[📖 Document\]](#documents) [\[🚀 Quick Start\]](#-quick-start) [\[🤗 Model Zoo\]](https://huggingface.co/collections/IPEC-COMMUNITY/foundation-vision-language-action-model-6795eb96a9c661f90236acbb) [\[✅ Performance\]](#-performance-in-simulation-and-real-world) [\[🙋 FAQs\]](#-faqs)
[\[🔥Pre-train\]](#-pre-train-from-scratch) [\[🚀 Fine-tune\]](#-fine-tune-from-spatialvla) [\[🎄Custom Dataset\]](#-use-custom-datasets)
</div>
## News 🚀🚀🚀
- `2025/01/29`: We release the [SpatialVLA 1.0](https://huggingface.co/collections/IPEC-COMMUNITY/foundation-vision-language-action-model-6795eb96a9c661f90236acbb). SpatialVLA achieves state-of-the-art performance across a diverse range of evaluations and shows significantly faster inference speed with fewer tokens per action.
- `2025/02/06`: We release the SimplerEnv evaluation code for SpatialVLA. Please refer to [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/), and make sure `transformers >= 4.47.0`.
- `2025/03/16`: Simplify the code structure and fix the dependencies conflict in issue [#19](https://github.com/SpatialVLA/SpatialVLA/issues/19).
> [!NOTE]
> 🔥 **An advanced version of SpatialVLA is under development! It leverages [lerobot](https://github.com/huggingface/lerobot) to simplify and accelerate data loading, supports multi-view and state inputs, and features a more streamlined code structure with enhanced performance! Please check out the [lerobot-branch](https://github.com/SpatialVLA/SpatialVLA/tree/lerobot)**
## Documents
### 🚀 Quick Start
> [!TIP]
> During the runtime process, a large amount of data is cached in the CPU content. To better manage and allocate content, we have replaced the memory management tool library with `tcmalloc`.
>
> For users with sudo privileges, you can install tcmalloc using `sudo apt-get install google-perftools` and find the `libtcmalloc.so.4` library in `/usr/lib/x86_64-linux-gnu` or `/usr/lib`.
>
> For users without sudo privileges, you can download the suitable version for your operating system from [official repo](https://rpmfind.net/linux/rpm2html/search.php?query=libtcmalloc.so.4()(64bit)) and install it manually.
>
> This step is **not** necessary and can be skipped based on your individual memory requirements.
SpatialVLA relies solely on HuggingFace Transformers 🤗, making deployment extremely easy. If your environment supports `transformers >= 4.47.0`, you can directly use the following code to load the model and perform inference. (requires 8.5GB of GPU memory).
```python
import torch
from PIL import Image
from transformers import AutoModel, AutoProcessor
model_name_or_path="IPEC-COMMUNITY/spatialvla-4b-224-pt"
processor = AutoProcessor.from_pretrained(model_name_or_path, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name_or_path, trust_remote_code=True, torch_dtype=torch.bfloat16).eval().cuda()
image = Image.open("example.png").convert("RGB")
prompt = "What action should the robot take to pick the cup?"
inputs = processor(images=[image], text=prompt, return_tensors="pt")
generation_outputs = model.predict_action(inputs)
actions = processor.decode_actions(generation_outputs, unnorm_key="bridge_orig/1.0.0")
print(actions)
```
If you want to use the model for fine-tuning or pre-training, you need to install the required packages and download the model from the Hugging Face model hub. The VLM backbone of SpatialVLA is PaLiGemma2, which requires transformers >= 4.47.0. Hence, create a Python environment with Python >= 3.10.
```bash
git clone git@github.com:SpatialVLA/SpatialVLA.git --depth 1
conda create -n spatialvla python=3.10
conda activate spatialvla
```
Install packages from `requirements.txt` file. Note that we use a customised `dlimp` to support seed setting for reproducibility. If you catch any problems, please manually install the dlimp form the [dlimp_custom](https://github.com/SpatialVLA/dlimp_custom).
```bash
pip install -r requirements.txt
```
### 🌟 **Pre-train from Scratch**
SpatialVLA is pre-trained with 1.1 Million real-robot demonstrations from the OXE and RH20T dataset on a cluster of 64 A100 GPUs for abut 10 days, using a batch size of 2048. You can pre-train the model from scratch using the following command. Before running the script, please download the [Open X-Embodiment](https://robotics-transformer-x.github.io) dataset and [RH20T](https://rh20t.github.io/#download) dataset (optional). Please also filter the dataset by following the instructions in the [moojink/rlds_dataset_builder](https://github.com/moojink/rlds_dataset_builder) and [spatialvla/rh20t](https://github.com/SpatialVLA/rh20t) to filter the dataset or convert it to the RLDS format.
```bash
# download paligemma2 and zoe depth
bash scripts/hf_download.sh
# torchrun
bash scripts/spatialvla_4b_pretrain/torchrun_pretrain.sh
# or in a slurm cluster
bash scripts/spatialvla_4b_pretrain/slurm_pretrain.sh
```
### 🌟 **Fine-tune from SpatialVLA**
Most of our fine-tuning experiments are conducted using LoRA on 4 or 8 A100 GPUs. You can use the following scripts for full-parameter or LoRA fine-tuning. For real-world experiments with small datasets, we prefer using LoRA for fine-tuning.
```bash
# full fine-tuning
bash scripts/spatialvla_4b_finetune/finetune_full.sh
# LoRA fine-tuning
bash scripts/spatialvla_4b_finetune/finetune_lora.sh
```
### 🌟 **SimplerEnv Benchmark**
We release the SimplerEnv evaluation code for SpatialVLA based on [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/). Please install the simpler_env environment by following [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/) and make sure `transformers >= 4.47.0`. Please refer to the Please refer to the [Model Zoo](#-model-zoo) for the model and dataset settings. After install all the dependencies, you can perform the evaluation by:
```bash
# under the project dir of SimplerEnv-OpenVLA/
bash scripts/run_spatialvla.sh
```
Note: Similar to most papers, e.g., HPT and TraceVLA, we omitted the `Open Top Drawer and Place Apple` from our evaluation, since the vast majority of policies achieved scores approaching 0 on this task.
### 🎄 Use Custom Datasets
To train on a custom dataset that is not part of OXE, we recommend converting it into the [RLDS](https://github.com/google-research/rlds) format, as this format directly aligns with our framework.
Once the dataset is converted, you’ll need to modify the following files:
- [data/oxe/mixtures.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/mixtures.py): Define a new mixture for your dataset in the OXE_NAMED_MIXTURES dictionary.
- [data/oxe/configs.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/configs.py): Add a new configuration specifying your dataset’s observation and action spaces to the OXE_DATASET_CONFIGS dictionary.
- [data/oxe/transforms.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/transforms.py): Define a new dataset transform function for your dataset, and add it to the OXE_STANDARDIZATION_TRANSFORMS registry at the bottom of the file.
## 🤗 Model Zoo
<table>
<tr>
<th>Model Name</th>
<th>Backbone</th>
<th>VLA Model</th>
<th>Note</th>
</tr>
<tr>
<td>SpatialVLA-4B-224-pt</td>
<td><a href="https://huggingface.co/google/paligemma2-3b-pt-224">google/paligemma2-3b-pt-224</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td>pretrained on openx and rh20t, TABLE I and II zero-shot, Fig.5 and 7</td>
</tr>
<tr>
<td>SpatialVLA-4B-mix-224-pt</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-mix-224-pt">spatialvla-4b-mix-224-pt</a></td>
<td>fine-tuning on the fractal and bridge mixture dataset, Fig.5 and 7</a></td>
</tr>
<tr>
<td>spatialvla-4b-224-sft-bridge</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-sft-bridge">spatialvla-4b-224-sft-bridge</a></td>
<td>fine-tuning on the bridge dataset, testing on simple-env widowx-robot, TABLE I fine-tuning</a></td>
</tr>
<tr>
<td>spatialvla-4b-224-sft-bridge</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-sft-fractal">spatialvla-4b-224-sft-fractal</a></td>
<td>fine-tuning on the fractal dataset, testing on simple-env google-robot, TABLE II ine-tuning</a></td>
</tr>
</table>
## ✅ Performance in Simulation and Real-world
> [!NOTE]
> SimplerEnv evaluation on Google Robot tasks.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="4">Visual Matching</th>
<th colspan="4">Variant Aggregation</th>
</tr>
<tr style="text-align: center;">
<th>Pick Coke Can</th>
<th>Move Near</th>
<th>Open/Close Drawer</th>
<th>#Average</th>
<th>Pick Coke Can</th>
<th>Move Near</th>
<th>Open/Close Drawer</th>
<th>#Average</th>
</tr>
</thead>
<tbody>
<tr>
<td>RT-1 (Begin)</td>
<td>2.7%</td>
<td>5.0%</td>
<td>13.9%</td>
<td>6.8%</td>
<td>2.2%</td>
<td>4.0%</td>
<td>6.9%</td>
<td>4.2%</td>
</tr>
<tr>
<td>RT-1 (15%)</td>
<td>71.0%</td>
<td>35.4%</td>
<td>56.5%</td>
<td>60.2%</td>
<td>81.3%</td>
<td>44.6%</td>
<td>26.7%</td>
<td>56.2%</td>
</tr>
<tr>
<td>RT-1 (Converged)</td>
<td>85.7%</td>
<td>44.2%</td>
<td>73.0%</td>
<td>74.6%</td>
<td>89.8%</td>
<td>50.0%</td>
<td>32.3%</td>
<td>63.3%</td>
</tr>
<tr>
<td>HPT</td>
<td>56.0%</td>
<td>60.0%</td>
<td>24.0%</td>
<td>46.0%</td>
<td>--</td>
<td>--</td>
<td>31.0%</td>
<td>45.0%</td>
</tr>
<tr>
<td>TraceVLA</td>
<td>28.0%</td>
<td>53.7%</td>
<td>57.0%</td>
<td>42.0%</td>
<td>60.0%</td>
<td>56.4%</td>
<td>29.4%</td>
<td>39.6%</td>
</tr>
<tr>
<td>RT-1-X</td>
<td>56.7%</td>
<td>31.7%</td>
<td>59.7%</td>
<td>53.4%</td>
<td>49.0%</td>
<td>32.3%</td>
<td>35.3%</td>
<td>64.3%</td>
</tr>
<tr>
<td>RT-2-X</td>
<td>78.7%</td>
<td>77.9%</td>
<td>25.0%</td>
<td>60.7%</td>
<td>82.3%</td>
<td>79.2%</td>
<td>--</td>
<td>--</td>
</tr>
<tr>
<td>Octo-Base</td>
<td>17.0%</td>
<td>4.2%</td>
<td>22.7%</td>
<td>16.8%</td>
<td>0.6%</td>
<td>3.1%</td>
<td>1.1%</td>
<td>1.1%</td>
</tr>
<tr>
<td>OpenVLA</td>
<td>16.3%</td>
<td>46.2%</td>
<td>35.6%</td>
<td>27.7%</td>
<td>54.5%</td>
<td>47.7%</td>
<td>17.7%</td>
<td>39.8%</td>
</tr>
<tr>
<td>RoboVLM (zero-shot)</td>
<td>72.7%</td>
<td>66.3%</td>
<td>26.8%</td>
<td>56.3%</td>
<td>68.3%</td>
<td>56.0%</td>
<td>8.5%</td>
<td>46.3%</td>
</tr>
<tr>
<td>RoboVLM (fine-tuning)</td>
<td>77.3%</td>
<td>61.7%</td>
<td>43.5%</td>
<td>63.4%</td>
<td>75.6%</td>
<td>60.0%</td>
<td>10.6%</td>
<td>51.3%</td>
</tr>
<tr>
<td>SpatialVLA (zero-shot)</td>
<td><b>81.0%</b></td>
<td><b>69.6%</b></td>
<td><b>59.3%</b></td>
<td><b>71.9%</b></td>
<td><b>89.5%</b></td>
<td><b>71.7%</b></td>
<td>36.2%</td>
<td><b>68.8%</b></td>
</tr>
<tr>
<td>SpatialVLA (fine-tuning)</td>
<td><b>86.0%</b></td>
<td><b>77.9%</b></td>
<td>57.4%</td>
<td><b>75.1%</b></td>
<td>88.0%</td>
<td>72.7%</td>
<td>41.8%</td>
<td><b>70.7%</b></td>
</tr>
</tbody>
</table>
> [!NOTE]
> SimplerEnv evaluation on WidowX Robot tasks.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="2">Put Spoon on Towel</th>
<th colspan="2">Put Carrot on Plate</th>
<th colspan="2">Stack Green Block on Yellow Block</th>
<th colspan="2">Put Eggplant in Yellow Basket</th>
<th rowspan="2">#Overall Average</th>
</tr>
<tr style="text-align: center;">
<th>Grasp Spoon</th>
<th>Success</th>
<th>Grasp Carrot</th>
<th>Success</th>
<th>Grasp Green Block</th>
<th>Success</th>
<th>Grasp Eggplant</th>
<th>Success</th>
</tr>
</thead>
<tbody>
<tr>
<td>RT-1-X</td>
<td>16.7%</td>
<td>0.0%</td>
<td>20.8%</td>
<td>4.2%</td>
<td>8.3%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>1.1%</td>
</tr>
<tr>
<td>Octo-Base</td>
<td>34.7%</td>
<td>12.5%</td>
<td>52.8%</td>
<td>8.3%</td>
<td>31.9%</td>
<td>0.0%</td>
<td>66.7%</td>
<td>43.1%</td>
<td>16.0%</td>
</tr>
<tr>
<td>Octo-Small</td>
<td>77.8%</td>
<td>47.2%</td>
<td>27.8%</td>
<td>9.7%</td>
<td>40.3%</td>
<td>4.2%</td>
<td>87.5%</td>
<td>56.9%</td>
<td>30.0%</td>
</tr>
<tr>
<td>OpenVLA</td>
<td>4.1%</td>
<td>0.0%</td>
<td>33.3%</td>
<td>0.0%</td>
<td>12.5%</td>
<td>0.0%</td>
<td>8.3%</td>
<td>4.1%</td>
<td>1.0%</td>
</tr>
<tr>
<td>RoboVLM (zero-shot)</td>
<td>37.5%</td>
<td>20.8%</td>
<td>33.3%</td>
<td>25.0%</td>
<td>8.3%</td>
<td>8.3%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>13.5%</td>
</tr>
<tr>
<td>RoboVLM (fine-tuning)</td>
<td>54.2%</td>
<td>29.2%</td>
<td>25.0%</td>
<td>25.0%</td>
<td>45.8%</td>
<td>12.5%</td>
<td>58.3%</td>
<td>58.3%</td>
<td>31.3%</td>
</tr>
<tr>
<td>SpatialVLA (zero-shot)</td>
<td><b>25.0%</b></td>
<td><b>20.8%</b></td>
<td><b>41.7%</b></td>
<td>20.8%</td>
<td><b>58.3%</b></td>
<td>25.0%</td>
<td><b>79.2%</b></td>
<td>70.8%</td>
<td><b>34.4%</b></td>
</tr>
<tr>
<td>SpatialVLA (fine-tuning)</td>
<td><b>20.8%</b></td>
<td>16.7%</td>
<td>29.2%</td>
<td>25.0%</td>
<td><b>62.5%</b></td>
<td>29.2%</td>
<td><b>100.0%</b></td>
<td><b>100.0%</b></td>
<td><b>42.7%</b></td>
</tr>
</tbody>
</table>
> [!NOTE]
> LIBERO Simulation Benchmark Results.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="2">LIBERO-Spatial</th>
<th colspan="2">LIBERO-Object</th>
<th colspan="2">LIBERO-Goal</th>
<th colspan="2">LIBERO-Long</th>
<th colspan="2">Average</th>
</tr>
<tr style="text-align: center;">
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
</tr>
</thead>
<tbody>
<tr>
<td>Diffusion Policy from scratch</td>
<td>78.3 ± 1.1%</td>
<td>5</td>
<td><b>92.5 ± 0.7%</b></td>
<td>1</td>
<td>68.3 ± 1.2%</td>
<td>5</td>
<td>50.5 ± 1.3%</td>
<td>5</td>
<td>72.4 ± 0.7%</td>
<td>5</td>
</tr>
<tr>
<td>Octo fine-tuned</td>
<td>78.9 ± 1.0%</td>
<td>4</td>
<td>85.7 ± 0.9%</td>
<td>4</td>
<td><b>84.6 ± 0.9%</b></td>
<td>1</td>
<td>51.1 ± 1.3%</td>
<td>4</td>
<td>75.1 ± 0.6%</td>
<td>3</td>
</tr>
<tr>
<td>OpenVLA fine-tuned</td>
<td>84.7 ± 0.9%</td>
<td>2</td>
<td>88.4 ± 0.8%</td>
<td>3</td>
<td>79.2 ± 1.0%</td>
<td>2</td>
<td>53.7 ± 1.3%</td>
<td>3</td>
<td>76.5 ± 0.6%</td>
<td>2</td>
</tr>
<tr>
<td>TraceVLA fine-tuned</td>
<td>84.6 ± 0.2%</td>
<td>3</td>
<td>85.2 ± 0.4%</td>
<td>5</td>
<td>75.1 ± 0.3%</td>
<td>4</td>
<td>54.1 ± 1.0%</td>
<td>2</td>
<td>74.8 ± 0.5%</td>
<td>4</td>
</tr>
<tr>
<td>SpatialVLA fine-tuned</td>
<td><b>88.2 ± 0.5%</b></td>
<td>1</td>
<td>89.9 ± 0.7%</td>
<td>2</td>
<td>78.6 ± 0.6%</td>
<td>3</td>
<td><b>55.5 ± 1.0%</b></td>
<td>1</td>
<td><b>78.1 ± 0.7%</b></td>
<td>1</td>
</tr>
</tbody>
</table>
> [!NOTE]
> Zero-shot Robot Control Evaluation on real-world WidowX Robot.
<img src=".assets/widowX_zeroshot.png" alt="perform">
> [!NOTE]
> Spatial Understanding Capability Evaluation.
<img src=".assets/spatial_setup.png" alt="perform">
> [!NOTE]
> Adapting to New Robot Setups on Franka Robot.
<img src=".assets/franka_sft.png" alt="perform">
## TODO List
- [x] Release pre-training / fine-tuning code for SpatialVLA series.
- [x] Release the code, model, and custom data of SpatialVLA.
- [x] Release the SimplerENV evaluation code for SpatialVLA series
- [ ] Release SpatialVLA2
## 🤗 FAQs
If you encounter any issues, feel free to open an issue on GitHub or reach out through discussions. We appreciate your feedback and contributions! 🚀
## License
This project is released under the [MIT license](LICENSE). Parts of this project contain code and models from other sources, which are subject to their respective licenses.
## Citation
If you find this project useful in your research, please consider cite:
```BibTeX
@article{qu2025spatialvla,
title={SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model},
author={Qu, Delin and Song, Haoming and Chen, Qizhi and Yao, Yuanqi and Ye, Xinyi and Ding, Yan and Wang, Zhigang and Gu, JiaYuan and Zhao, Bin and Wang, Dong and others},
journal={arXiv preprint arXiv:2501.15830},
year={2025}
}
```
## Acknowledgement
SpatialVLA is built with reference to the code of the following projects: [InternVL](https://github.com/OpenGVLab/InternVL), [Google Paligemma2](https://huggingface.co/google/paligemma2-3b-pt-224), [Transformers](https://github.com/huggingface/transformers), [OpenVLA](https://github.com/openvla/openvla) and [ZoeDepth](https://huggingface.co/spaces/shariqfarooq/ZoeDepth). Thanks for their awesome work!
| <div align="center">
# SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Models (RSS 2025)
A spatial-enhanced vision-language-action model trained on 1.1 Million real robot episodes. 🤗
purely huggingFace-based, concise code with efficient performance.
> [Delin Qu*](https://github.com/DelinQu)<sup>1,2</sup>, [HaomingSong*](https://github.com/HaomingSong)<sup>1,3</sup>, [Qizhi Chen*](https://github.com/Tavish9)<sup>1,4</sup>, [Dong Wang†](https://scholar.google.com/citations?user=dasL9V4AAAAJ&hl=en)<sup>1</sup>, [Yuanqi Yao](https://scholar.google.com/citations?user=s482QHoAAAAJ&hl=zh-CN)<sup>1</sup>, [X. Ye](https://scholar.google.com/citations?user=GlYeyfoAAAAJ&hl=zh-CN)<sup>1</sup>, [Y. Ding](https://yding25.com)<sup>1</sup>, [Z. Wang](https://scholar.google.com/citations?user=cw3EaAYAAAAJ&hl=zh-CN)<sup>1</sup>, [Jiayuan Gu](https://cseweb.ucsd.edu/~jigu/)<sup>5</sup>, [Bin Zhao†](https://scholar.google.com/citations?hl=zh-CN&user=DQB0hqwAAAAJ)<sup>1</sup>, [Xuelong Li](https://scholar.google.com/citations?user=ahUibskAAAAJ)<sup>1,6</sup>
> Shanghai AI Laboratory<sup>1</sup>, Fudan University<sup>2</sup>, Shanghai Jiao Tong University<sup>3</sup>, Zhejiang University<sup>4</sup>, ShanghaiTech<sup>5</sup>, TeleAI<sup>6</sup>
[\[📄Paper\]](https://arxiv.org/pdf/2501.15830) [\[🔥Project Page\]](https://spatialvla.github.io/) [\[📖 Document\]](#documents) [\[🚀 Quick Start\]](#-quick-start) [\[🤗 Model Zoo\]](https://huggingface.co/collections/IPEC-COMMUNITY/foundation-vision-language-action-model-6795eb96a9c661f90236acbb) [\[✅ Performance\]](#-performance-in-simulation-and-real-world) [\[🙋 FAQs\]](#-faqs)
[\[🔥Pre-train\]](#-pre-train-from-scratch) [\[🚀 Fine-tune\]](#-fine-tune-from-spatialvla) [\[🎄Custom Dataset\]](#-use-custom-datasets)
</div>
## News 🚀🚀🚀
- `2025/01/29`: We release the [SpatialVLA 1.0](https://huggingface.co/collections/IPEC-COMMUNITY/foundation-vision-language-action-model-6795eb96a9c661f90236acbb). SpatialVLA achieves state-of-the-art performance across a diverse range of evaluations and shows significantly faster inference speed with fewer tokens per action.
- `2025/02/06`: We release the SimplerEnv evaluation code for SpatialVLA. Please refer to [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/), and make sure `transformers >= 4.47.0`.
- `2025/03/16`: Simplify the code structure and fix the dependencies conflict in issue [#19](https://github.com/SpatialVLA/SpatialVLA/issues/19).
> [!NOTE]
> 🔥 **An advanced version of SpatialVLA is under development! It leverages [lerobot](https://github.com/huggingface/lerobot) to simplify and accelerate data loading, supports multi-view and state inputs, and features a more streamlined code structure with enhanced performance! Please check out the [lerobot-branch](https://github.com/SpatialVLA/SpatialVLA/tree/lerobot)**
## Documents
### 🚀 Quick Start
> [!TIP]
> During the runtime process, a large amount of data is cached in the CPU content. To better manage and allocate content, we have replaced the memory management tool library with `tcmalloc`.
>
> For users with sudo privileges, you can install tcmalloc using `sudo apt-get install google-perftools` and find the `libtcmalloc.so.4` library in `/usr/lib/x86_64-linux-gnu` or `/usr/lib`.
>
> For users without sudo privileges, you can download the suitable version for your operating system from [official repo](https://rpmfind.net/linux/rpm2html/search.php?query=libtcmalloc.so.4()(64bit)) and install it manually.
>
> This step is **not** necessary and can be skipped based on your individual memory requirements.
SpatialVLA relies solely on HuggingFace Transformers 🤗, making deployment extremely easy. If your environment supports `transformers >= 4.47.0`, you can directly use the following code to load the model and perform inference. (requires 8.5GB of GPU memory).
```python
import torch
from PIL import Image
from transformers import AutoModel, AutoProcessor
model_name_or_path="IPEC-COMMUNITY/spatialvla-4b-224-pt"
processor = AutoProcessor.from_pretrained(model_name_or_path, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name_or_path, trust_remote_code=True, torch_dtype=torch.bfloat16).eval().cuda()
image = Image.open("example.png").convert("RGB")
prompt = "What action should the robot take to pick the cup?"
inputs = processor(images=[image], text=prompt, return_tensors="pt")
generation_outputs = model.predict_action(inputs)
actions = processor.decode_actions(generation_outputs, unnorm_key="bridge_orig/1.0.0")
print(actions)
```
If you want to use the model for fine-tuning or pre-training, you need to install the required packages and download the model from the Hugging Face model hub. The VLM backbone of SpatialVLA is PaLiGemma2, which requires transformers >= 4.47.0. Hence, create a Python environment with Python >= 3.10.
```bash
git clone git@github.com:SpatialVLA/SpatialVLA.git --depth 1
conda create -n spatialvla python=3.10
conda activate spatialvla
```
Install packages from `requirements.txt` file. Note that we use a customised `dlimp` to support seed setting for reproducibility. If you catch any problems, please manually install the dlimp form the [dlimp_custom](https://github.com/SpatialVLA/dlimp_custom).
```bash
pip install -r requirements.txt
```
### 🌟 **Pre-train from Scratch**
SpatialVLA is pre-trained with 1.1 Million real-robot demonstrations from the OXE and RH20T dataset on a cluster of 64 A100 GPUs for abut 10 days, using a batch size of 2048. You can pre-train the model from scratch using the following command. Before running the script, please download the [Open X-Embodiment](https://robotics-transformer-x.github.io) dataset and [RH20T](https://rh20t.github.io/#download) dataset (optional). Please also filter the dataset by following the instructions in the [moojink/rlds_dataset_builder](https://github.com/moojink/rlds_dataset_builder) and [spatialvla/rh20t](https://github.com/SpatialVLA/rh20t) to filter the dataset or convert it to the RLDS format.
```bash
# download paligemma2 and zoe depth
bash scripts/hf_download.sh
# torchrun
bash scripts/spatialvla_4b_pretrain/torchrun_pretrain.sh
# or in a slurm cluster
bash scripts/spatialvla_4b_pretrain/slurm_pretrain.sh
```
### 🌟 **Fine-tune from SpatialVLA**
Most of our fine-tuning experiments are conducted using LoRA on 4 or 8 A100 GPUs. You can use the following scripts for full-parameter or LoRA fine-tuning. For real-world experiments with small datasets, we prefer using LoRA for fine-tuning.
```bash
# full fine-tuning
bash scripts/spatialvla_4b_finetune/finetune_full.sh
# LoRA fine-tuning
bash scripts/spatialvla_4b_finetune/finetune_lora.sh
```
### 🌟 **SimplerEnv Benchmark**
We release the SimplerEnv evaluation code for SpatialVLA based on [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/). Please install the simpler_env environment by following [DelinQu/SimplerEnv-OpenVLA](https://github.com/DelinQu/SimplerEnv-OpenVLA/) and make sure `transformers >= 4.47.0`. Please refer to the Please refer to the [Model Zoo](#-model-zoo) for the model and dataset settings. After install all the dependencies, you can perform the evaluation by:
```bash
# under the project dir of SimplerEnv-OpenVLA/
bash scripts/run_spatialvla.sh
```
Note: Similar to most papers, e.g., HPT and TraceVLA, we omitted the `Open Top Drawer and Place Apple` from our evaluation, since the vast majority of policies achieved scores approaching 0 on this task.
### 🎄 Use Custom Datasets
To train on a custom dataset that is not part of OXE, we recommend converting it into the [RLDS](https://github.com/google-research/rlds) format, as this format directly aligns with our framework.
Once the dataset is converted, you’ll need to modify the following files:
- [data/oxe/mixtures.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/mixtures.py): Define a new mixture for your dataset in the OXE_NAMED_MIXTURES dictionary.
- [data/oxe/configs.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/configs.py): Add a new configuration specifying your dataset’s observation and action spaces to the OXE_DATASET_CONFIGS dictionary.
- [data/oxe/transforms.py](https://github.com/SpatialVLA/SpatialVLA/blob/main/data/oxe/transforms.py): Define a new dataset transform function for your dataset, and add it to the OXE_STANDARDIZATION_TRANSFORMS registry at the bottom of the file.
## 🤗 Model Zoo
<table>
<tr>
<th>Model Name</th>
<th>Backbone</th>
<th>VLA Model</th>
<th>Note</th>
</tr>
<tr>
<td>SpatialVLA-4B-224-pt</td>
<td><a href="https://huggingface.co/google/paligemma2-3b-pt-224">google/paligemma2-3b-pt-224</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td>pretrained on openx and rh20t, TABLE I and II zero-shot, Fig.5 and 7</td>
</tr>
<tr>
<td>SpatialVLA-4B-mix-224-pt</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-mix-224-pt">spatialvla-4b-mix-224-pt</a></td>
<td>fine-tuning on the fractal and bridge mixture dataset, Fig.5 and 7</a></td>
</tr>
<tr>
<td>spatialvla-4b-224-sft-bridge</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-sft-bridge">spatialvla-4b-224-sft-bridge</a></td>
<td>fine-tuning on the bridge dataset, testing on simple-env widowx-robot, TABLE I fine-tuning</a></td>
</tr>
<tr>
<td>spatialvla-4b-224-sft-bridge</td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-pt">spatialvla-4b-224-pt</a></td>
<td><a href="https://huggingface.co/IPEC-COMMUNITY/spatialvla-4b-224-sft-fractal">spatialvla-4b-224-sft-fractal</a></td>
<td>fine-tuning on the fractal dataset, testing on simple-env google-robot, TABLE II ine-tuning</a></td>
</tr>
</table>
## ✅ Performance in Simulation and Real-world
> [!NOTE]
> SimplerEnv evaluation on Google Robot tasks.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="4">Visual Matching</th>
<th colspan="4">Variant Aggregation</th>
</tr>
<tr style="text-align: center;">
<th>Pick Coke Can</th>
<th>Move Near</th>
<th>Open/Close Drawer</th>
<th>#Average</th>
<th>Pick Coke Can</th>
<th>Move Near</th>
<th>Open/Close Drawer</th>
<th>#Average</th>
</tr>
</thead>
<tbody>
<tr>
<td>RT-1 (Begin)</td>
<td>2.7%</td>
<td>5.0%</td>
<td>13.9%</td>
<td>6.8%</td>
<td>2.2%</td>
<td>4.0%</td>
<td>6.9%</td>
<td>4.2%</td>
</tr>
<tr>
<td>RT-1 (15%)</td>
<td>71.0%</td>
<td>35.4%</td>
<td>56.5%</td>
<td>60.2%</td>
<td>81.3%</td>
<td>44.6%</td>
<td>26.7%</td>
<td>56.2%</td>
</tr>
<tr>
<td>RT-1 (Converged)</td>
<td>85.7%</td>
<td>44.2%</td>
<td>73.0%</td>
<td>74.6%</td>
<td>89.8%</td>
<td>50.0%</td>
<td>32.3%</td>
<td>63.3%</td>
</tr>
<tr>
<td>HPT</td>
<td>56.0%</td>
<td>60.0%</td>
<td>24.0%</td>
<td>46.0%</td>
<td>--</td>
<td>--</td>
<td>31.0%</td>
<td>45.0%</td>
</tr>
<tr>
<td>TraceVLA</td>
<td>28.0%</td>
<td>53.7%</td>
<td>57.0%</td>
<td>42.0%</td>
<td>60.0%</td>
<td>56.4%</td>
<td>29.4%</td>
<td>39.6%</td>
</tr>
<tr>
<td>RT-1-X</td>
<td>56.7%</td>
<td>31.7%</td>
<td>59.7%</td>
<td>53.4%</td>
<td>49.0%</td>
<td>32.3%</td>
<td>35.3%</td>
<td>64.3%</td>
</tr>
<tr>
<td>RT-2-X</td>
<td>78.7%</td>
<td>77.9%</td>
<td>25.0%</td>
<td>60.7%</td>
<td>82.3%</td>
<td>79.2%</td>
<td>--</td>
<td>--</td>
</tr>
<tr>
<td>Octo-Base</td>
<td>17.0%</td>
<td>4.2%</td>
<td>22.7%</td>
<td>16.8%</td>
<td>0.6%</td>
<td>3.1%</td>
<td>1.1%</td>
<td>1.1%</td>
</tr>
<tr>
<td>OpenVLA</td>
<td>16.3%</td>
<td>46.2%</td>
<td>35.6%</td>
<td>27.7%</td>
<td>54.5%</td>
<td>47.7%</td>
<td>17.7%</td>
<td>39.8%</td>
</tr>
<tr>
<td>RoboVLM (zero-shot)</td>
<td>72.7%</td>
<td>66.3%</td>
<td>26.8%</td>
<td>56.3%</td>
<td>68.3%</td>
<td>56.0%</td>
<td>8.5%</td>
<td>46.3%</td>
</tr>
<tr>
<td>RoboVLM (fine-tuning)</td>
<td>77.3%</td>
<td>61.7%</td>
<td>43.5%</td>
<td>63.4%</td>
<td>75.6%</td>
<td>60.0%</td>
<td>10.6%</td>
<td>51.3%</td>
</tr>
<tr>
<td>SpatialVLA (zero-shot)</td>
<td><b>81.0%</b></td>
<td><b>69.6%</b></td>
<td><b>59.3%</b></td>
<td><b>71.9%</b></td>
<td><b>89.5%</b></td>
<td><b>71.7%</b></td>
<td>36.2%</td>
<td><b>68.8%</b></td>
</tr>
<tr>
<td>SpatialVLA (fine-tuning)</td>
<td><b>86.0%</b></td>
<td><b>77.9%</b></td>
<td>57.4%</td>
<td><b>75.1%</b></td>
<td>88.0%</td>
<td>72.7%</td>
<td>41.8%</td>
<td><b>70.7%</b></td>
</tr>
</tbody>
</table>
> [!NOTE]
> SimplerEnv evaluation on WidowX Robot tasks.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="2">Put Spoon on Towel</th>
<th colspan="2">Put Carrot on Plate</th>
<th colspan="2">Stack Green Block on Yellow Block</th>
<th colspan="2">Put Eggplant in Yellow Basket</th>
<th rowspan="2">#Overall Average</th>
</tr>
<tr style="text-align: center;">
<th>Grasp Spoon</th>
<th>Success</th>
<th>Grasp Carrot</th>
<th>Success</th>
<th>Grasp Green Block</th>
<th>Success</th>
<th>Grasp Eggplant</th>
<th>Success</th>
</tr>
</thead>
<tbody>
<tr>
<td>RT-1-X</td>
<td>16.7%</td>
<td>0.0%</td>
<td>20.8%</td>
<td>4.2%</td>
<td>8.3%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>1.1%</td>
</tr>
<tr>
<td>Octo-Base</td>
<td>34.7%</td>
<td>12.5%</td>
<td>52.8%</td>
<td>8.3%</td>
<td>31.9%</td>
<td>0.0%</td>
<td>66.7%</td>
<td>43.1%</td>
<td>16.0%</td>
</tr>
<tr>
<td>Octo-Small</td>
<td>77.8%</td>
<td>47.2%</td>
<td>27.8%</td>
<td>9.7%</td>
<td>40.3%</td>
<td>4.2%</td>
<td>87.5%</td>
<td>56.9%</td>
<td>30.0%</td>
</tr>
<tr>
<td>OpenVLA</td>
<td>4.1%</td>
<td>0.0%</td>
<td>33.3%</td>
<td>0.0%</td>
<td>12.5%</td>
<td>0.0%</td>
<td>8.3%</td>
<td>4.1%</td>
<td>1.0%</td>
</tr>
<tr>
<td>RoboVLM (zero-shot)</td>
<td>37.5%</td>
<td>20.8%</td>
<td>33.3%</td>
<td>25.0%</td>
<td>8.3%</td>
<td>8.3%</td>
<td>0.0%</td>
<td>0.0%</td>
<td>13.5%</td>
</tr>
<tr>
<td>RoboVLM (fine-tuning)</td>
<td>54.2%</td>
<td>29.2%</td>
<td>25.0%</td>
<td>25.0%</td>
<td>45.8%</td>
<td>12.5%</td>
<td>58.3%</td>
<td>58.3%</td>
<td>31.3%</td>
</tr>
<tr>
<td>SpatialVLA (zero-shot)</td>
<td><b>25.0%</b></td>
<td><b>20.8%</b></td>
<td><b>41.7%</b></td>
<td>20.8%</td>
<td><b>58.3%</b></td>
<td>25.0%</td>
<td><b>79.2%</b></td>
<td>70.8%</td>
<td><b>34.4%</b></td>
</tr>
<tr>
<td>SpatialVLA (fine-tuning)</td>
<td><b>20.8%</b></td>
<td>16.7%</td>
<td>29.2%</td>
<td>25.0%</td>
<td><b>62.5%</b></td>
<td>29.2%</td>
<td><b>100.0%</b></td>
<td><b>100.0%</b></td>
<td><b>42.7%</b></td>
</tr>
</tbody>
</table>
> [!NOTE]
> LIBERO Simulation Benchmark Results.
<table border="1" class="dataframe">
<thead>
<tr style="text-align: center;">
<th rowspan="2">Model</th>
<th colspan="2">LIBERO-Spatial</th>
<th colspan="2">LIBERO-Object</th>
<th colspan="2">LIBERO-Goal</th>
<th colspan="2">LIBERO-Long</th>
<th colspan="2">Average</th>
</tr>
<tr style="text-align: center;">
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
<th>SR (↑)</th>
<th>Rank (↓)</th>
</tr>
</thead>
<tbody>
<tr>
<td>Diffusion Policy from scratch</td>
<td>78.3 ± 1.1%</td>
<td>5</td>
<td><b>92.5 ± 0.7%</b></td>
<td>1</td>
<td>68.3 ± 1.2%</td>
<td>5</td>
<td>50.5 ± 1.3%</td>
<td>5</td>
<td>72.4 ± 0.7%</td>
<td>5</td>
</tr>
<tr>
<td>Octo fine-tuned</td>
<td>78.9 ± 1.0%</td>
<td>4</td>
<td>85.7 ± 0.9%</td>
<td>4</td>
<td><b>84.6 ± 0.9%</b></td>
<td>1</td>
<td>51.1 ± 1.3%</td>
<td>4</td>
<td>75.1 ± 0.6%</td>
<td>3</td>
</tr>
<tr>
<td>OpenVLA fine-tuned</td>
<td>84.7 ± 0.9%</td>
<td>2</td>
<td>88.4 ± 0.8%</td>
<td>3</td>
<td>79.2 ± 1.0%</td>
<td>2</td>
<td>53.7 ± 1.3%</td>
<td>3</td>
<td>76.5 ± 0.6%</td>
<td>2</td>
</tr>
<tr>
<td>TraceVLA fine-tuned</td>
<td>84.6 ± 0.2%</td>
<td>3</td>
<td>85.2 ± 0.4%</td>
<td>5</td>
<td>75.1 ± 0.3%</td>
<td>4</td>
<td>54.1 ± 1.0%</td>
<td>2</td>
<td>74.8 ± 0.5%</td>
<td>4</td>
</tr>
<tr>
<td>SpatialVLA fine-tuned</td>
<td><b>88.2 ± 0.5%</b></td>
<td>1</td>
<td>89.9 ± 0.7%</td>
<td>2</td>
<td>78.6 ± 0.6%</td>
<td>3</td>
<td><b>55.5 ± 1.0%</b></td>
<td>1</td>
<td><b>78.1 ± 0.7%</b></td>
<td>1</td>
</tr>
</tbody>
</table>
> [!NOTE]
> Zero-shot Robot Control Evaluation on real-world WidowX Robot.
<img src=".assets/widowX_zeroshot.png" alt="perform">
> [!NOTE]
> Spatial Understanding Capability Evaluation.
<img src=".assets/spatial_setup.png" alt="perform">
> [!NOTE]
> Adapting to New Robot Setups on Franka Robot.
<img src=".assets/franka_sft.png" alt="perform">
## TODO List
- [x] Release pre-training / fine-tuning code for SpatialVLA series.
- [x] Release the code, model, and custom data of SpatialVLA.
- [x] Release the SimplerENV evaluation code for SpatialVLA series
- [ ] Release SpatialVLA2
## 🤗 FAQs
If you encounter any issues, feel free to open an issue on GitHub or reach out through discussions. We appreciate your feedback and contributions! 🚀
## License
This project is released under the [MIT license](LICENSE). Parts of this project contain code and models from other sources, which are subject to their respective licenses.
## Citation
If you find this project useful in your research, please consider cite:
```BibTeX
@article{qu2025spatialvla,
title={SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model},
author={Qu, Delin and Song, Haoming and Chen, Qizhi and Yao, Yuanqi and Ye, Xinyi and Ding, Yan and Wang, Zhigang and Gu, JiaYuan and Zhao, Bin and Wang, Dong and others},
journal={arXiv preprint arXiv:2501.15830},
year={2025}
}
```
## Acknowledgement
SpatialVLA is built with reference to the code of the following projects: [InternVL](https://github.com/OpenGVLab/InternVL), [Google Paligemma2](https://huggingface.co/google/paligemma2-3b-pt-224), [Transformers](https://github.com/huggingface/transformers), [OpenVLA](https://github.com/openvla/openvla) and [ZoeDepth](https://huggingface.co/spaces/shariqfarooq/ZoeDepth). Thanks for their awesome work!
| 116 | 0 | [
"arxiv:2501.15830",
"region:us"
] | 2025-11-11T13:49:26+00:00 | 2025-11-11T14:56:52+00:00 | 0 |
Lyric1010/numina-drop-merge-0.02B |
# Dataset: numina-drop-merge-0.02B
This dataset was uploaded from `/mnt/yulan_pretrain/mount/data_final_train_qwen3/numina-drop-merge-0.02B/stage_1`. |
# Dataset: numina-drop-merge-0.02B
This dataset was uploaded from `/mnt/yulan_pretrain/mount/data_final_train_qwen3/numina-drop-merge-0.02B/stage_1`. | 11 | 0 | [
"modality:text",
"region:us",
"text",
"pretraining"
] | 2025-11-11T14:18:18+00:00 | 2025-11-11T14:52:20+00:00 | 0 |
nsimonato25/rubber_duck_dataset_panda_256x256_smoothed |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "Panda",
"total_episodes": 1,
"total_frames": 178,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.images.image2": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
8
],
"names": [
"states"
],
"fps": 30
},
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"actions"
],
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "Panda",
"total_episodes": 1,
"total_frames": 178,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.images.image2": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
8
],
"names": [
"states"
],
"fps": 30
},
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"actions"
],
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 19 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:n<1K",
"format:parquet",
"modality:image",
"modality:timeseries",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T14:56:44+00:00 | 2025-11-11T14:56:49+00:00 | 0 |
mosquito-alert/ai-mosquito-alert-challenge-2023 | # AI Mosquito Alert Challenge Dataset 2023
> A curated dataset for automated mosquito species identification from citizen-science images.
The AI Mosquito Alert 2023 Challenge Dataset is a curated dataset used in the AI Mosquito Alert 2023 Challenge (https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023), aimed at improving mosquito species identification through AI and deep learning models.
## Dataset Details
### Dataset Description
This dataset was prepared for the AI Mosquito Alert 2023 Challenge (https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023),
an initiative focused on improving mosquito species identification through AI and deep learning models. An example implementation and baseline model are available on the
[AIcrowd Showcase: Mosquito Alert YOLOv5 Baseline Submission](https://www.aicrowd.com/showcase/mosquitoalert-yolov5-baseline-submission).
The dataset for this challenge is derived from a citizen science project focused on mosquito identification. It comprises over 10000 real-world images of mosquitoes captured by participants using mobile phones. These images offer a diverse representation of mosquitoes in various scenarios and locations.
Each image is labelled with bounding box coordinates and mosquito class information.
The dataset is distributed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0): https://creativecommons.org/licenses/by-nc-sa/4.0/.
The dataset was created through the efforts of the Mosquito Alert team, collaborators and thousands of citizen scientists. Please credit the Mosquito Alert Community (www.mosquitoalert.com) if you use this dataset (e.g., 'Mosquito Alert dataset, downloaded from [link], CC BY-NC-SA 4.0').
The intellectual property (IP) rights of this dataset belong to the Mosquito Alert team.
The license is included in the file license.txt within the dataset zip file, along with the images, labels and dataset description.
The dataset consists of 10357 labelled images (approximately 9.8 GB in total). Images are accompanied by a designated CSV file called: annotations.csv. The CSV files include bounding box coordinates in the format: top left and bottom right notation ("bbx_xtl", "bbx_ytl", "bbx_xbr", "bbx_ybr").
**Classes distribution:**
The dataset consists of six distinct classes, including species and genus levels, as well as a species
complex.
A summary of the mosquito classes, their descriptions, and corresponding class names used in the
dataset:
- Aedes aegypti (species level) - class name: "aegypti"
- Aedes albopictus (species level) - class name: "albopictus"
- Anopheles (genus level) - class name: "anopheles"
- Culex (genus level) - class name: "culex" (species classification is challenging, so it is given
at the genus level)
- Culiseta (genus level) - class name: "culiseta"
- Aedes japonicus/Aedes koreicus (species complex - difficult to differentiate between the two
species) - class name: "japonicus-koreicus"
| Class name | Taxonomic level / description | Image count |
|-------------|-------------------------------|--------------:|
| **aegypti** | *Aedes aegypti* (species level) | 47 |
| **albopictus** | *Aedes albopictus* (species level) | 4612 |
| **anopheles** | *Anopheles* (genus level) | 84 |
| **culex** | *Culex* (genus level) | 4563 |
| **culiseta** | *Culiseta* (genus level) | 622 |
| **japonicus-koreicus** | *Aedes japonicus / Aedes koreicus* species complex | 429 |
| **Total** | | **10357** |
**Label file:**
The dataset includes a single CSV file: annotations.csv, which contains all the annotations for the
images. Each row in the file provides the following information:
- img_fName: image file name
- img_w: image width
- img_h: image height
- bbx_xtl: bounding box top-left x-coordinate
- bbx_ytl: bounding box top-left y-coordinate
- bbx_xbr: bounding box bottom-right x-coordinate
- bbx_ybr: bounding box bottom-right y-coordinate
- class_label: class label (e.g., 'albopictus').
**Additional notes:**
- a broader description of the dataset and classes will be provided in the
https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023#dataset and
https://www.youtube.com/watch?v=qSWJZUY-5DM challenge video.
- exif information has been removed from the images for privacy protection.
- most images contain a single mosquito with its corresponding bounding box and class label. However, in rare cases with multiple mosquitoes, only one mosquito is assigned a bounding
box and label for consistency and compatibility.
**Curated by:** Mosquito Alert Team, collaborators, and thousands of citizen scientists.
**Shared by [optional]:** [Monika Falk](https://www.linkedin.com/in/falk-monika/), AI Research Technician (Mosquito Alert / CEAB-CSIC)
**License:** [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/)
### Dataset Sources [optional]
<!-- Provide the basic links for the dataset. -->
- **Repository:** [Zenodo Record – AI Mosquito Alert Challenge Dataset 2023](https://doi.org/10.5281/zenodo.15063886) DOI: https://doi.org/10.5281/zenodo.15063886
- **Paper [optional]:** In preparation, dataset description and methodological details will be included in the upcoming Mosquito Alert AI Challenge 2023 paper.
- **Demo [optional]:** [AIcrowd Challenge Page](https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023)
## Uses
<!-- Address questions around how the dataset is intended to be used. -->
### Direct Use
The dataset is intended for research and educational purposes related to computer vision, biodiversity monitoring, and vector surveillance.
It can be used to train, validate, and benchmark object detection or image classification models for automatic mosquito species identification.
It is suitable for developing AI tools to support citizen-science initiatives and public health research on mosquito distribution.
### Out-of-Scope Use
The dataset must not be used for commercial purposes, as it is distributed under a non-commercial licence (CC BY-NC-SA 4.0).
## Dataset Structure
The dataset folder structure is organized as follows:
```text
mosquito_dataset_ai_v1/
├── images/
│ ├── image1.jpg
│ ├── image2.jpg
│ └── ...
├── labels/
│ └── annotations.csv
└── license.txt
```
## Dataset Creation
### Curation Rationale
This dataset was assembled to support research on automatic mosquito species identification within the Mosquito Alert project.
Images were selected from citizen-science submissions and reviewed by entomologists to ensure correct taxonomic labels.
The aim was to create a realistic, high-quality research image collection that reflects the variety and quality of data typically received through the Mosquito Alert app.
### Source Data
#### Data Collection and Processing
Images were collected through the Mosquito Alert citizen-science platform, where volunteers submit photographs of mosquitoes using a mobile application.
Each submission is reviewed by a group of volunteer entomologists who validate the species identity of the visible specimens.
Low-quality, duplicate, or uncertain images are excluded. All metadata that could reveal personal or location information is removed before dataset release.
#### Who are the source data producers?
The original images were contributed by citizen scientists participating in the Mosquito Alert project.
Species validations and annotations were performed by a volunteer network of entomologists collaborating with the Mosquito Alert project.
### Annotations [optional]
#### Annotation process
Bounding boxes were created using an automated detection model trained on Mosquito Alert data.
A subset of annotations was manually reviewed to improve label accuracy and bounding-box precision.
Images with multiple mosquitoes were excluded, and bounding boxes were checked for consistency.
Minor imperfections may still exist, but the dataset reflects an improved and verified version prepared for the AI Mosquito Alert Challenge 2023.
#### Who are the annotators?
Bounding boxes were generated automatically using a neural network model,
while mosquito taxa were identified by volunteer entomologists collaborating with Mosquito Alert.
#### Personal and Sensitive Information
The dataset contains no personal or sensitive information. Images depict mosquitoes photographed in natural or household environments and do not include identifiable persons or private objects. All EXIF metadata was removed to eliminate any residual location or device information.
## Bias, Risks, and Limitations
- **Geographical bias:** image submissions depend on the geographic distribution of Mosquito Alert users, so some regions are underrepresented.
- **Class imbalance:** certain mosquito species, such as *Aedes albopictus*, are overrepresented compared to others.
- **Variable image quality:** photos differ in lighting, focus, and scale, which may affect model performance.
- **Automated annotation noise:** as bounding boxes were partly generated automatically, minor inaccuracies may remain.
These factors should be considered when using the dataset for model training or evaluation.
### Recommendations
Users should consider the dataset’s biases and limitations when designing experiments and interpreting results.
It is recommended to report class-wise performance metrics and use data augmentation or balancing techniques to mitigate class imbalance.
Proper attribution to the Mosquito Alert Community is required when using or publishing results based on this dataset.
## Citation [optional]
The dataset is introduced as part of the *AI Mosquito Alert Challenge 2023*.
A formal paper describing the dataset and methodology is in preparation.
**BibTeX:**
```bibtex
@dataset{mosquito_alert_2023,
title = {AI Mosquito Alert Challenge Dataset 2023},
author = {Mosquito Alert Community},
year = {2025},
publisher = {Zenodo},
version = {1.0},
doi = {10.5281/zenodo.15063886},
url = {https://doi.org/10.5281/zenodo.15063886}
}
```
**APA:**
Bartumeus, F., Garriga, J., Falk, M., & Mosquito Alert Expert Community. (2025). AI Mosquito Alert Challenge Dataset 2023 (Version v1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.15063886
## Glossary [optional]
- **Bounding box:** A rectangular region that encloses the mosquito in an image, defined by its top-left and bottom-right coordinates.
- **Class label:** The mosquito species or genus assigned to an image (e.g., *albopictus*, *culex*).
- **Citizen science:** A form of scientific research that relies on public participation for data collection.
- **IoU (Intersection over Union):** A metric used to evaluate the overlap between predicted and ground-truth bounding boxes.
- **F1 score:** A measure that balances precision and recall, used to evaluate detection model performance.
## More Information [optional]
For more information about the Mosquito Alert project, visit [www.mosquitoalert.com](https://www.mosquitoalert.com).
Details about the AI Mosquito Alert Challenge can be found on [AIcrowd](https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023).
Questions regarding the dataset or collaboration opportunities can be directed to the Mosquito Alert team.
## Dataset Card Authors [optional]
**Frederic Bartumeus** - Supervisor (1, 2, 3, 4)
**Monika Falk** - Data Manager (4, 5)
**Joan Garriga** - Data Manager (4)
**Mosquito Alert Expert Community** - Annotators
*(1) CEAB-CSIC - Centre for Advanced Studies of Blanes*
*(2) ICREA - Catalan Institution for Research and Advanced Studies*
*(3) CREAF - Centre de Recerca Ecològica i Aplicacions Forestals*
*(4) Mosquito Alert Project*
*(5) VŠB - Technical University of Ostrava*
## Dataset Card Contact
For questions about the dataset or collaboration inquiries, please contact the Mosquito Alert team through the official form:
[https://www.mosquitoalert.com/en/about-us/contact/](https://www.mosquitoalert.com/en/about-us/contact/) | # AI Mosquito Alert Challenge Dataset 2023
> A curated dataset for automated mosquito species identification from citizen-science images.
The AI Mosquito Alert 2023 Challenge Dataset is a curated dataset used in the AI Mosquito Alert 2023 Challenge (https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023), aimed at improving mosquito species identification through AI and deep learning models.
## Dataset Details
### Dataset Description
This dataset was prepared for the AI Mosquito Alert 2023 Challenge (https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023),
an initiative focused on improving mosquito species identification through AI and deep learning models. An example implementation and baseline model are available on the
[AIcrowd Showcase: Mosquito Alert YOLOv5 Baseline Submission](https://www.aicrowd.com/showcase/mosquitoalert-yolov5-baseline-submission).
The dataset for this challenge is derived from a citizen science project focused on mosquito identification. It comprises over 10000 real-world images of mosquitoes captured by participants using mobile phones. These images offer a diverse representation of mosquitoes in various scenarios and locations.
Each image is labelled with bounding box coordinates and mosquito class information.
The dataset is distributed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0): https://creativecommons.org/licenses/by-nc-sa/4.0/.
The dataset was created through the efforts of the Mosquito Alert team, collaborators and thousands of citizen scientists. Please credit the Mosquito Alert Community (www.mosquitoalert.com) if you use this dataset (e.g., 'Mosquito Alert dataset, downloaded from [link], CC BY-NC-SA 4.0').
The intellectual property (IP) rights of this dataset belong to the Mosquito Alert team.
The license is included in the file license.txt within the dataset zip file, along with the images, labels and dataset description.
The dataset consists of 10357 labelled images (approximately 9.8 GB in total). Images are accompanied by a designated CSV file called: annotations.csv. The CSV files include bounding box coordinates in the format: top left and bottom right notation ("bbx_xtl", "bbx_ytl", "bbx_xbr", "bbx_ybr").
**Classes distribution:**
The dataset consists of six distinct classes, including species and genus levels, as well as a species
complex.
A summary of the mosquito classes, their descriptions, and corresponding class names used in the
dataset:
- Aedes aegypti (species level) - class name: "aegypti"
- Aedes albopictus (species level) - class name: "albopictus"
- Anopheles (genus level) - class name: "anopheles"
- Culex (genus level) - class name: "culex" (species classification is challenging, so it is given
at the genus level)
- Culiseta (genus level) - class name: "culiseta"
- Aedes japonicus/Aedes koreicus (species complex - difficult to differentiate between the two
species) - class name: "japonicus-koreicus"
| Class name | Taxonomic level / description | Image count |
|-------------|-------------------------------|--------------:|
| **aegypti** | *Aedes aegypti* (species level) | 47 |
| **albopictus** | *Aedes albopictus* (species level) | 4612 |
| **anopheles** | *Anopheles* (genus level) | 84 |
| **culex** | *Culex* (genus level) | 4563 |
| **culiseta** | *Culiseta* (genus level) | 622 |
| **japonicus-koreicus** | *Aedes japonicus / Aedes koreicus* species complex | 429 |
| **Total** | | **10357** |
**Label file:**
The dataset includes a single CSV file: annotations.csv, which contains all the annotations for the
images. Each row in the file provides the following information:
- img_fName: image file name
- img_w: image width
- img_h: image height
- bbx_xtl: bounding box top-left x-coordinate
- bbx_ytl: bounding box top-left y-coordinate
- bbx_xbr: bounding box bottom-right x-coordinate
- bbx_ybr: bounding box bottom-right y-coordinate
- class_label: class label (e.g., 'albopictus').
**Additional notes:**
- a broader description of the dataset and classes will be provided in the
https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023#dataset and
https://www.youtube.com/watch?v=qSWJZUY-5DM challenge video.
- exif information has been removed from the images for privacy protection.
- most images contain a single mosquito with its corresponding bounding box and class label. However, in rare cases with multiple mosquitoes, only one mosquito is assigned a bounding
box and label for consistency and compatibility.
**Curated by:** Mosquito Alert Team, collaborators, and thousands of citizen scientists.
**Shared by [optional]:** [Monika Falk](https://www.linkedin.com/in/falk-monika/), AI Research Technician (Mosquito Alert / CEAB-CSIC)
**License:** [Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0)](https://creativecommons.org/licenses/by-nc-sa/4.0/)
### Dataset Sources [optional]
<!-- Provide the basic links for the dataset. -->
- **Repository:** [Zenodo Record – AI Mosquito Alert Challenge Dataset 2023](https://doi.org/10.5281/zenodo.15063886) DOI: https://doi.org/10.5281/zenodo.15063886
- **Paper [optional]:** In preparation, dataset description and methodological details will be included in the upcoming Mosquito Alert AI Challenge 2023 paper.
- **Demo [optional]:** [AIcrowd Challenge Page](https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023)
## Uses
<!-- Address questions around how the dataset is intended to be used. -->
### Direct Use
The dataset is intended for research and educational purposes related to computer vision, biodiversity monitoring, and vector surveillance.
It can be used to train, validate, and benchmark object detection or image classification models for automatic mosquito species identification.
It is suitable for developing AI tools to support citizen-science initiatives and public health research on mosquito distribution.
### Out-of-Scope Use
The dataset must not be used for commercial purposes, as it is distributed under a non-commercial licence (CC BY-NC-SA 4.0).
## Dataset Structure
The dataset folder structure is organized as follows:
```text
mosquito_dataset_ai_v1/
├── images/
│ ├── image1.jpg
│ ├── image2.jpg
│ └── ...
├── labels/
│ └── annotations.csv
└── license.txt
```
## Dataset Creation
### Curation Rationale
This dataset was assembled to support research on automatic mosquito species identification within the Mosquito Alert project.
Images were selected from citizen-science submissions and reviewed by entomologists to ensure correct taxonomic labels.
The aim was to create a realistic, high-quality research image collection that reflects the variety and quality of data typically received through the Mosquito Alert app.
### Source Data
#### Data Collection and Processing
Images were collected through the Mosquito Alert citizen-science platform, where volunteers submit photographs of mosquitoes using a mobile application.
Each submission is reviewed by a group of volunteer entomologists who validate the species identity of the visible specimens.
Low-quality, duplicate, or uncertain images are excluded. All metadata that could reveal personal or location information is removed before dataset release.
#### Who are the source data producers?
The original images were contributed by citizen scientists participating in the Mosquito Alert project.
Species validations and annotations were performed by a volunteer network of entomologists collaborating with the Mosquito Alert project.
### Annotations [optional]
#### Annotation process
Bounding boxes were created using an automated detection model trained on Mosquito Alert data.
A subset of annotations was manually reviewed to improve label accuracy and bounding-box precision.
Images with multiple mosquitoes were excluded, and bounding boxes were checked for consistency.
Minor imperfections may still exist, but the dataset reflects an improved and verified version prepared for the AI Mosquito Alert Challenge 2023.
#### Who are the annotators?
Bounding boxes were generated automatically using a neural network model,
while mosquito taxa were identified by volunteer entomologists collaborating with Mosquito Alert.
#### Personal and Sensitive Information
The dataset contains no personal or sensitive information. Images depict mosquitoes photographed in natural or household environments and do not include identifiable persons or private objects. All EXIF metadata was removed to eliminate any residual location or device information.
## Bias, Risks, and Limitations
- **Geographical bias:** image submissions depend on the geographic distribution of Mosquito Alert users, so some regions are underrepresented.
- **Class imbalance:** certain mosquito species, such as *Aedes albopictus*, are overrepresented compared to others.
- **Variable image quality:** photos differ in lighting, focus, and scale, which may affect model performance.
- **Automated annotation noise:** as bounding boxes were partly generated automatically, minor inaccuracies may remain.
These factors should be considered when using the dataset for model training or evaluation.
### Recommendations
Users should consider the dataset’s biases and limitations when designing experiments and interpreting results.
It is recommended to report class-wise performance metrics and use data augmentation or balancing techniques to mitigate class imbalance.
Proper attribution to the Mosquito Alert Community is required when using or publishing results based on this dataset.
## Citation [optional]
The dataset is introduced as part of the *AI Mosquito Alert Challenge 2023*.
A formal paper describing the dataset and methodology is in preparation.
**BibTeX:**
```bibtex
@dataset{mosquito_alert_2023,
title = {AI Mosquito Alert Challenge Dataset 2023},
author = {Mosquito Alert Community},
year = {2025},
publisher = {Zenodo},
version = {1.0},
doi = {10.5281/zenodo.15063886},
url = {https://doi.org/10.5281/zenodo.15063886}
}
```
**APA:**
Bartumeus, F., Garriga, J., Falk, M., & Mosquito Alert Expert Community. (2025). AI Mosquito Alert Challenge Dataset 2023 (Version v1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.15063886
## Glossary [optional]
- **Bounding box:** A rectangular region that encloses the mosquito in an image, defined by its top-left and bottom-right coordinates.
- **Class label:** The mosquito species or genus assigned to an image (e.g., *albopictus*, *culex*).
- **Citizen science:** A form of scientific research that relies on public participation for data collection.
- **IoU (Intersection over Union):** A metric used to evaluate the overlap between predicted and ground-truth bounding boxes.
- **F1 score:** A measure that balances precision and recall, used to evaluate detection model performance.
## More Information [optional]
For more information about the Mosquito Alert project, visit [www.mosquitoalert.com](https://www.mosquitoalert.com).
Details about the AI Mosquito Alert Challenge can be found on [AIcrowd](https://www.aicrowd.com/challenges/mosquitoalert-challenge-2023).
Questions regarding the dataset or collaboration opportunities can be directed to the Mosquito Alert team.
## Dataset Card Authors [optional]
**Frederic Bartumeus** - Supervisor (1, 2, 3, 4)
**Monika Falk** - Data Manager (4, 5)
**Joan Garriga** - Data Manager (4)
**Mosquito Alert Expert Community** - Annotators
*(1) CEAB-CSIC - Centre for Advanced Studies of Blanes*
*(2) ICREA - Catalan Institution for Research and Advanced Studies*
*(3) CREAF - Centre de Recerca Ecològica i Aplicacions Forestals*
*(4) Mosquito Alert Project*
*(5) VŠB - Technical University of Ostrava*
## Dataset Card Contact
For questions about the dataset or collaboration inquiries, please contact the Mosquito Alert team through the official form:
[https://www.mosquitoalert.com/en/about-us/contact/](https://www.mosquitoalert.com/en/about-us/contact/) | 74 | 0 | [
"task_categories:object-detection",
"task_categories:image-classification",
"language:en",
"license:cc-by-nc-sa-4.0",
"size_categories:10K<n<100K",
"modality:image",
"region:us",
"biology"
] | 2025-11-04T12:26:59+00:00 | 2025-11-11T14:52:16+00:00 | 0 |
GoAGI-AI/STEM_arabic_qa_1k |
A high quality STEM question and answer dataset in Arabic. Subjects include: mathematics, physics, biology, chemistry.
Data provider: <a href="https://goagi.ai">goagi.ai</a>.
**Total rows** in the dataset - 1001. Explanation of columns:
**"id"** - id of the row
**"subject"** - indicates which subject does the question belongs to: eg. biology
**"grade"** - indicates which year of school is the question likely to belong to, e.g. "12" - the final year of school or "university" for a university level question.
**"question"** - actual question. See example below.
اختر الإجابة الصحيحة مما يلي : تحدث الحركة في الإنسان بتآزر مجموعة من الأجهزة وهي :
Translation:
Choose the correct answer from the following:
Movement in humans occurs through the coordination of a group of systems, which are:
**"options"** - answer options if it is a multiple choice question. Example of options below.
١) الجهاز العضلي والهيكلي والدوري .
٢) الجهاز التنفسي والعصبي والهيكلي .
٣) الجهاز الهيكلي والعصبي والعضلي .
٤) الجهاز الهيكلي والتنفسي والدوري .
Translation:
The muscular, skeletal, and circulatory systems.
The respiratory, nervous, and skeletal systems.
The skeletal, nervous, and muscular systems.
The skeletal, respiratory, and circulatory systems.
**"answer"** - answer to the question. Example below.
الجهاز الهيكلي والعصبي والعضلي.
Translation:
The skeletal, nervous and muscular systems.
**"solution"** - detailed step by step solution of the question if it requires a solution.
**"hint"** - hint required to solve a question.
**"location"** - which country is this question likely to be asked in. "eg" is for Egypt.
**"language"** - language in which this question is written. "arb" is standard modern arabic.
|
A high quality STEM question and answer dataset in Arabic. Subjects include: mathematics, physics, biology, chemistry.
Data provider: <a href="https://goagi.ai">goagi.ai</a>.
**Total rows** in the dataset - 1001. Explanation of columns:
**"id"** - id of the row
**"subject"** - indicates which subject does the question belongs to: eg. biology
**"grade"** - indicates which year of school is the question likely to belong to, e.g. "12" - the final year of school or "university" for a university level question.
**"question"** - actual question. See example below.
اختر الإجابة الصحيحة مما يلي : تحدث الحركة في الإنسان بتآزر مجموعة من الأجهزة وهي :
Translation:
Choose the correct answer from the following:
Movement in humans occurs through the coordination of a group of systems, which are:
**"options"** - answer options if it is a multiple choice question. Example of options below.
١) الجهاز العضلي والهيكلي والدوري .
٢) الجهاز التنفسي والعصبي والهيكلي .
٣) الجهاز الهيكلي والعصبي والعضلي .
٤) الجهاز الهيكلي والتنفسي والدوري .
Translation:
The muscular, skeletal, and circulatory systems.
The respiratory, nervous, and skeletal systems.
The skeletal, nervous, and muscular systems.
The skeletal, respiratory, and circulatory systems.
**"answer"** - answer to the question. Example below.
الجهاز الهيكلي والعصبي والعضلي.
Translation:
The skeletal, nervous and muscular systems.
**"solution"** - detailed step by step solution of the question if it requires a solution.
**"hint"** - hint required to solve a question.
**"location"** - which country is this question likely to be asked in. "eg" is for Egypt.
**"language"** - language in which this question is written. "arb" is standard modern arabic.
| 65 | 0 | [
"task_categories:question-answering",
"language:ar",
"license:mit",
"size_categories:1K<n<10K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"chemistry",
"biology",
"mathematics",
"physics"
] | 2025-10-26T12:29:45+00:00 | 2025-11-11T14:54:10+00:00 | 0 |
nsimonato25/rubber_duck_dataset_panda_256x256 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "Panda",
"total_episodes": 1,
"total_frames": 484,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.images.image2": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
8
],
"names": [
"states"
],
"fps": 30
},
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"actions"
],
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "Panda",
"total_episodes": 1,
"total_frames": 484,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.images.image2": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channels"
],
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
8
],
"names": [
"states"
],
"fps": 30
},
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"actions"
],
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 12 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:n<1K",
"format:parquet",
"modality:image",
"modality:timeseries",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T14:49:00+00:00 | 2025-11-11T14:49:07+00:00 | 0 |
PleIAs/SYNTH |
# SYNTH
<div align="center">
<img src="figures/pleias.png" width="60%" alt="Pleias" />
</div>
<p align="center">
<a href="https://pleias.fr/blog/blogsynth-the-new-data-frontier"><b>Blog announcement</b></a>
</p>
**SYNTH** is the first open generalist synthetic dataset for training small reasoning model end-to-end, jointly released by Pleias and the AI Alliance.
SYNTH includes 79,648,272 individual text samples, comprising over 41 billion words (about 75 billion tokens with Pleias tokenizer). It is based on the amplification of 58,698 articles from Wikipedia and made possible thanks to the *Structured Wikipedia* dataset from Wikimedia Enterprise.
SYNTH differs from existing open synthetic dataset in being:
* **fully open** based on seed text under open license (CC-By-SA) and generated with models allowing for output reuse. This means that SYNTH can be universally release and serve as a basis for further reproducible synthetic pipelines.
* **state of the art** for small models below 350 million parameters. We release two models train on SYNTH achieving current best results for size range on MMLU and other standard evaluation metrics.
* **data efficient** with best results attained with only 100-200 billions tokens trained on SYNTH.
* **reasoning by design** with all generated answers being accompanied with intermediary reasoning traces in an entirely new syntax.
* **diverse** comprising a wide range of exercises that cover many use cases of small models: retrieval-augmented generation, creative writing, arithmetics, information extraction, etc.
* **multilingual** with about 20% of all texts in other languages than English, for now limited on European languages (German, French, Spanish, Italian, Polish, Dutch, Latin).
SYNTH is not only the name of a dataset but an initiative for open synthetic data and open environment led by AI Alliance and Pleias that aims to address the critical gap in open-source AI development by creating a cutting-edge, open-source data corpus for training sovereign AI models and advanced AI agents.
## Dataset Design
## Amplified knowledge
At its core, SYNTH is a fully synthetic and engineered corpus derived from a sample of 50,000 pages curated by the Wikipedia community. Throughout the past two decades, thousands of contributors selected a collection of core topics that every encyclopedia should have, Wikipedia:Vital articles. It’s a concentric selection starting at level 1 (10 articles) up to level 5 (50,000 articles). SYNTH includes as its starting point all articles featured in level 5.
SYNTH further expands on this core nucleus with three additional seed collections:
* **specialized articles**: following on intermediary evaluation, we added 8,698 articles to reinforce coverage of specific fields like law, medicine, chemistry. Selection was based on category tree search analysis and aimed to fill remaining holes in knowledge coverage from Wikipedia:Vital articles.
* **textbooks**: wikipedia articles are focused on encyclopedic knowledge but lag on *practical* knowledge and *how to*, which happens to be the focus of another Wikimedia project, Wikibooks. For now we included 3,727 pages on cooking from Wikibooks but looking forward to expand on additional forms of experential knowledge (gardening, language acquisition, etc.)
* **recent/self knowledge**: we incorporated a small sample of 130 texts hand-crafted internally to expand model familiarity with recent events, self-awareness about training condition and general research information on AI. This collection has been highly amplified.
This content act as the SYNTH memory base and has been amplified at a minimum 100 times (about 10000 times for recent/self knowledge). Our amplification strategy relies on a new synthetic pipeline, partly inspired by RAG applications:
* Selection of individual consistent **sections** from the original articles (about 250,000 for the core sample of 50,000 pages).
* Generation of queries with randomized constraints for style variation, query outcomes. It proved especially determining to have enough negative queries to reinforce world knowledge and limit hallucinations.
## Synthetic exercises
The approach has been originally explored by Pleias for retrieval-augmented generation. It has been extended to virtually most of the expected use case of small reasoning models:
* **arithmetics**
* **creative writing** We injected randomized constraints
## Dataset Details
### Dataset Description
- **Curated by:** Wikipedia community (Wikipedia:Vital Articles) and Pleias.
- **Funded by [optional]:** Pleias
- **Shared by [optional]:** Pleias
- **Language(s) (NLP):** English (80%), French, German, Italian, Spanish, Polish, Dutch and Latin.
- **License:**
### Dataset Sources [optional]
While the final training data is fully synthetic, it relied on seeds collected from three data sources:
- **[Structured Wikipedia](https://huggingface.co/datasets/wikimedia/structured-wikipedia):** We used directly the dumps made available by the Wikimedia Foundation.
- **Wikibooks:** extracted through the official Wikimedia API.
- **Internal documents from Pleias:** mostly model-self documentation and few updated information.
## Uses
The dataset aims to support data efficient training of small reasoning model. It provide a generalist, self-sufficient collection of multilingual amplified encyclopedic texts along with synthetic reasoning traces, as well as synthetic tasks that reinforce most of the expected capacities of small model.
In contrast with organic pretraining dataset, SYNTH allows for fast convergence to the existing SOTA (about 100 billion tokens). Furthermore, SYNTH is fully releasable, only use sourced text under free license.
Overall, SYNTH aims to support an emerging ecosystem of small training model by providing a reusable generalist foundational dataset.
### Direct Use
Direct use include:
- **Pretraining of small reasoning models**: the dataset is sufficient to elicit most expected capacities in small models.
- **Mid-training/fine-tuning of existing models**: we already led successful experiments with Pleias-350m.
- **Research/explainability experiment**: with its openness and data efficiency, SYNTH should be an ideal resource for research on model memorization or skill acquisition.
### Out-of-Scope Use
Current out-of-scope use include:
- **Code generation**: we intently excluded code data from SYNTH as this would require the development of specific synthetic pipeline.
- **Global multilingual support**: SYNTH only claims support from our current list of eight languages.
- **Training of large models**: the difficulty of synthetic exercises has been calibrated for models smaller than a few billion parameters.
Yet, SYNTH is a live resources and we intend to cover some of these use cases in future releases.
## Dataset Structure
| Field | Type | Description |
| ----------------------- | -------- | ------------------------------------------------------------------------------------------------------------------- |
| **synth_id** | `string` | Unique synthetic identifier for each generated sample. |
| **language** | `string` | Language of the text sample (e.g., `"en"`, `"fr"`, `"it"`, `"es"`, `"de"`, `"pl"`, `"nl"`, `"la"`). |
| **exercise** | `string` | Type of synthetic exercise (e.g., reasoning, writing, retrieval, arithmetic). Describes the synthetic task context. |
| **model** | `string` | Finetuned model used to generate the synthetic sample |
| **query** | `string` | Backtranslated query. |
| **query_seed_url** | `string` | URL of the Wikipedia or Wikibooks section that served as the seed for query generation. |
| **query_seed_text** | `string` | Extend text used as seed for query generation. |
| **additional_seed_url** | `string` | Optional additional URL(s) used as supplementary seed |
| **seed_license** | `string` | License of the seed text (most of the time `"CC-BY-SA 4.0"`). |
| **constraints** | `string` | Generation constraints applied to answer generation. Varies depending on the exercise |
| **script** | `string` | Internal template or script identifier defining the structure of the synthetic exercise. |
| **synthetic_reasoning** | `string` | Generated reasoning draft. |
| **synthetic_answer** | `string` | Final generated answer or output corresponding to the query. |
| **words** | `int64` | Word count of the full generated text sample (query + draft + answer) |
## Dataset Creation
### Curation Rationale
SYNTH is structured around a “memory core”, the Wikipedia vital articles.. Throughout the past two decades, thousands of contributors selected a collection of core topics that every encyclopedia should have: it’s a concentric selection starting at level 1 (10 articles) up to level 5 (50,000 articles). SYNTH includes as its starting point all articles featured in level 5. It further expands on this selection by increasing coverage of more specialized domains (physics, chemistry, law…) through targeted expansion of wikidata knowledge graphs.
### Source Data
The 58,698 Wikipedia articles were collected thanks to ''Structured Wikipedia'', a project from Wikimedia Enterprise that parsed directly rendered Wikipedia articles in html. Structured Wikipedia fixed most of the formatting issues linked with the mediawiki syntax and provides a clean, section-based version of all Wikipedia pages.
We additionally extracted 3,000 cooking recipes from Wikibooks using the standard API method from Wikimedia.
#### Data Collection and Processing
#### Who are the source data producers?
The main sourced dataset used for synthetic amplification was curated by the English Wikipedia communities throughout nearly 2 decades. Rationale for selection are available on the relevant talk pages of Wikipedia:Vital articles.
The selection reflect similar bias for "canon" general knowledge in English-speaking countries than major LLM benchmarks like MMLU (drawn from high school exams).
#### Personal and Sensitive Information
The dataset only contain encyclopedic information on highly well-known historical people. No PII curation was needed.
## Bias, Risks, and Limitations
The dataset was created from a collection of 50,000 Wikipedia articles curated by the community (Wikipedia:Vital Articles).
On top of the well documented structural bias in Wikipedia contribution and editing, the selection has been intently made from the perspective of western US/European culture.
Due to systematic Wikipedia grounding, the data presents a very low risk of toxic or problematic content, as well as poor or highly hallucinated information.
|
# SYNTH
<div align="center">
<img src="figures/pleias.png" width="60%" alt="Pleias" />
</div>
<p align="center">
<a href="https://pleias.fr/blog/blogsynth-the-new-data-frontier"><b>Blog announcement</b></a>
</p>
**SYNTH** is the first open generalist synthetic dataset for training small reasoning model end-to-end, jointly released by Pleias and the AI Alliance.
SYNTH includes 79,648,272 individual text samples, comprising over 41 billion words (about 75 billion tokens with Pleias tokenizer). It is based on the amplification of 58,698 articles from Wikipedia and made possible thanks to the *Structured Wikipedia* dataset from Wikimedia Enterprise.
SYNTH differs from existing open synthetic dataset in being:
* **fully open** based on seed text under open license (CC-By-SA) and generated with models allowing for output reuse. This means that SYNTH can be universally release and serve as a basis for further reproducible synthetic pipelines.
* **state of the art** for small models below 350 million parameters. We release two models train on SYNTH achieving current best results for size range on MMLU and other standard evaluation metrics.
* **data efficient** with best results attained with only 100-200 billions tokens trained on SYNTH.
* **reasoning by design** with all generated answers being accompanied with intermediary reasoning traces in an entirely new syntax.
* **diverse** comprising a wide range of exercises that cover many use cases of small models: retrieval-augmented generation, creative writing, arithmetics, information extraction, etc.
* **multilingual** with about 20% of all texts in other languages than English, for now limited on European languages (German, French, Spanish, Italian, Polish, Dutch, Latin).
SYNTH is not only the name of a dataset but an initiative for open synthetic data and open environment led by AI Alliance and Pleias that aims to address the critical gap in open-source AI development by creating a cutting-edge, open-source data corpus for training sovereign AI models and advanced AI agents.
## Dataset Design
## Amplified knowledge
At its core, SYNTH is a fully synthetic and engineered corpus derived from a sample of 50,000 pages curated by the Wikipedia community. Throughout the past two decades, thousands of contributors selected a collection of core topics that every encyclopedia should have, Wikipedia:Vital articles. It’s a concentric selection starting at level 1 (10 articles) up to level 5 (50,000 articles). SYNTH includes as its starting point all articles featured in level 5.
SYNTH further expands on this core nucleus with three additional seed collections:
* **specialized articles**: following on intermediary evaluation, we added 8,698 articles to reinforce coverage of specific fields like law, medicine, chemistry. Selection was based on category tree search analysis and aimed to fill remaining holes in knowledge coverage from Wikipedia:Vital articles.
* **textbooks**: wikipedia articles are focused on encyclopedic knowledge but lag on *practical* knowledge and *how to*, which happens to be the focus of another Wikimedia project, Wikibooks. For now we included 3,727 pages on cooking from Wikibooks but looking forward to expand on additional forms of experential knowledge (gardening, language acquisition, etc.)
* **recent/self knowledge**: we incorporated a small sample of 130 texts hand-crafted internally to expand model familiarity with recent events, self-awareness about training condition and general research information on AI. This collection has been highly amplified.
This content act as the SYNTH memory base and has been amplified at a minimum 100 times (about 10000 times for recent/self knowledge). Our amplification strategy relies on a new synthetic pipeline, partly inspired by RAG applications:
* Selection of individual consistent **sections** from the original articles (about 250,000 for the core sample of 50,000 pages).
* Generation of queries with randomized constraints for style variation, query outcomes. It proved especially determining to have enough negative queries to reinforce world knowledge and limit hallucinations.
## Synthetic exercises
The approach has been originally explored by Pleias for retrieval-augmented generation. It has been extended to virtually most of the expected use case of small reasoning models:
* **arithmetics**
* **creative writing** We injected randomized constraints
## Dataset Details
### Dataset Description
- **Curated by:** Wikipedia community (Wikipedia:Vital Articles) and Pleias.
- **Funded by [optional]:** Pleias
- **Shared by [optional]:** Pleias
- **Language(s) (NLP):** English (80%), French, German, Italian, Spanish, Polish, Dutch and Latin.
- **License:**
### Dataset Sources [optional]
While the final training data is fully synthetic, it relied on seeds collected from three data sources:
- **[Structured Wikipedia](https://huggingface.co/datasets/wikimedia/structured-wikipedia):** We used directly the dumps made available by the Wikimedia Foundation.
- **Wikibooks:** extracted through the official Wikimedia API.
- **Internal documents from Pleias:** mostly model-self documentation and few updated information.
## Uses
The dataset aims to support data efficient training of small reasoning model. It provide a generalist, self-sufficient collection of multilingual amplified encyclopedic texts along with synthetic reasoning traces, as well as synthetic tasks that reinforce most of the expected capacities of small model.
In contrast with organic pretraining dataset, SYNTH allows for fast convergence to the existing SOTA (about 100 billion tokens). Furthermore, SYNTH is fully releasable, only use sourced text under free license.
Overall, SYNTH aims to support an emerging ecosystem of small training model by providing a reusable generalist foundational dataset.
### Direct Use
Direct use include:
- **Pretraining of small reasoning models**: the dataset is sufficient to elicit most expected capacities in small models.
- **Mid-training/fine-tuning of existing models**: we already led successful experiments with Pleias-350m.
- **Research/explainability experiment**: with its openness and data efficiency, SYNTH should be an ideal resource for research on model memorization or skill acquisition.
### Out-of-Scope Use
Current out-of-scope use include:
- **Code generation**: we intently excluded code data from SYNTH as this would require the development of specific synthetic pipeline.
- **Global multilingual support**: SYNTH only claims support from our current list of eight languages.
- **Training of large models**: the difficulty of synthetic exercises has been calibrated for models smaller than a few billion parameters.
Yet, SYNTH is a live resources and we intend to cover some of these use cases in future releases.
## Dataset Structure
| Field | Type | Description |
| ----------------------- | -------- | ------------------------------------------------------------------------------------------------------------------- |
| **synth_id** | `string` | Unique synthetic identifier for each generated sample. |
| **language** | `string` | Language of the text sample (e.g., `"en"`, `"fr"`, `"it"`, `"es"`, `"de"`, `"pl"`, `"nl"`, `"la"`). |
| **exercise** | `string` | Type of synthetic exercise (e.g., reasoning, writing, retrieval, arithmetic). Describes the synthetic task context. |
| **model** | `string` | Finetuned model used to generate the synthetic sample |
| **query** | `string` | Backtranslated query. |
| **query_seed_url** | `string` | URL of the Wikipedia or Wikibooks section that served as the seed for query generation. |
| **query_seed_text** | `string` | Extend text used as seed for query generation. |
| **additional_seed_url** | `string` | Optional additional URL(s) used as supplementary seed |
| **seed_license** | `string` | License of the seed text (most of the time `"CC-BY-SA 4.0"`). |
| **constraints** | `string` | Generation constraints applied to answer generation. Varies depending on the exercise |
| **script** | `string` | Internal template or script identifier defining the structure of the synthetic exercise. |
| **synthetic_reasoning** | `string` | Generated reasoning draft. |
| **synthetic_answer** | `string` | Final generated answer or output corresponding to the query. |
| **words** | `int64` | Word count of the full generated text sample (query + draft + answer) |
## Dataset Creation
### Curation Rationale
SYNTH is structured around a “memory core”, the Wikipedia vital articles.. Throughout the past two decades, thousands of contributors selected a collection of core topics that every encyclopedia should have: it’s a concentric selection starting at level 1 (10 articles) up to level 5 (50,000 articles). SYNTH includes as its starting point all articles featured in level 5. It further expands on this selection by increasing coverage of more specialized domains (physics, chemistry, law…) through targeted expansion of wikidata knowledge graphs.
### Source Data
The 58,698 Wikipedia articles were collected thanks to ''Structured Wikipedia'', a project from Wikimedia Enterprise that parsed directly rendered Wikipedia articles in html. Structured Wikipedia fixed most of the formatting issues linked with the mediawiki syntax and provides a clean, section-based version of all Wikipedia pages.
We additionally extracted 3,000 cooking recipes from Wikibooks using the standard API method from Wikimedia.
#### Data Collection and Processing
#### Who are the source data producers?
The main sourced dataset used for synthetic amplification was curated by the English Wikipedia communities throughout nearly 2 decades. Rationale for selection are available on the relevant talk pages of Wikipedia:Vital articles.
The selection reflect similar bias for "canon" general knowledge in English-speaking countries than major LLM benchmarks like MMLU (drawn from high school exams).
#### Personal and Sensitive Information
The dataset only contain encyclopedic information on highly well-known historical people. No PII curation was needed.
## Bias, Risks, and Limitations
The dataset was created from a collection of 50,000 Wikipedia articles curated by the community (Wikipedia:Vital Articles).
On top of the well documented structural bias in Wikipedia contribution and editing, the selection has been intently made from the perspective of western US/European culture.
Due to systematic Wikipedia grounding, the data presents a very low risk of toxic or problematic content, as well as poor or highly hallucinated information.
| 2,768 | 76 | [
"task_categories:text-generation",
"task_categories:zero-shot-classification",
"task_categories:summarization",
"language:en",
"language:fr",
"language:it",
"language:es",
"language:de",
"language:pl",
"language:nl",
"language:la",
"license:cdla-permissive-2.0",
"size_categories:10M<n<100M",
"format:parquet",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"wikipedia",
"art",
"math",
"writing"
] | 2025-11-10T14:08:26+00:00 | 2025-11-11T14:38:51+00:00 | 76 |
AIbnuHibban/e-commerce-sentiment-bahasa-indonesia |
# E-Commerce Sentiment Analysis Dataset (Indonesian)
Dataset komentar dan ulasan produk e-commerce dalam Bahasa Indonesia untuk analisis sentiment.
## Dataset Summary
Dataset ini berisi 21,840 komentar e-commerce dalam Bahasa Indonesia yang telah dilabeli dengan sentiment (positif, netral, negatif). Dataset mencakup berbagai jenis komentar termasuk sarkasme dan ironi yang umum ditemukan dalam ulasan online.
## Dataset Structure
### Data Fields
- `comment` (string): Komentar atau ulasan produk
- `rating` (int): Rating produk (1-5)
- `sentiment` (string): Label sentiment - "positive", "neutral", atau "negative"
### Data Splits
Dataset terdiri dari:
- **simple.json**: 17,000 samples (data utama)
- **challange.json**: 4,840 samples (data challenging termasuk sarkasme)
- **Total**: 21,840 samples
### Data Distribution
- Positive: 7,480 (34.3%)
- Negative: 7,470 (34.2%)
- Neutral: 6,890 (31.5%)
## Usage
```python
from datasets import load_dataset
# Load dataset
dataset = load_dataset("AIbnuHibban/e-commerce-sentiment-bahasa-indonesia")
# Or load from JSON files directly
import json
import pandas as pd
with open('simple.json', 'r') as f:
data = json.load(f)
df = pd.DataFrame(data)
```
## Example Data
```json
{
"comment": "Barang mantap sekali! Worth it banget dengan harganya",
"rating": 5,
"sentiment": "positive"
}
```
```json
{
"comment": "Wow, senang banget ditipu! Barang jelek tapi harganya mahal",
"rating": 1,
"sentiment": "negative"
}
```
## Characteristics
- **Language**: Bahasa Indonesia (Indonesian)
- **Domain**: E-commerce product reviews
- **Special Features**: Includes sarcastic and ironic comments
- **Use Cases**:
- Sentiment analysis
- Opinion mining
- Customer feedback analysis
- Sarcasm detection
## Citation
If you use this dataset, please cite:
```
@dataset{e_commerce_sentiment_bahasa_indonesia,
author = {AIbnuHibban},
title = {E-Commerce Sentiment Analysis Dataset (Indonesian)},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/AIbnuHibban/e-commerce-sentiment-bahasa-indonesia}
}
```
## License
MIT License
|
# E-Commerce Sentiment Analysis Dataset (Indonesian)
Dataset komentar dan ulasan produk e-commerce dalam Bahasa Indonesia untuk analisis sentiment.
## Dataset Summary
Dataset ini berisi 21,840 komentar e-commerce dalam Bahasa Indonesia yang telah dilabeli dengan sentiment (positif, netral, negatif). Dataset mencakup berbagai jenis komentar termasuk sarkasme dan ironi yang umum ditemukan dalam ulasan online.
## Dataset Structure
### Data Fields
- `comment` (string): Komentar atau ulasan produk
- `rating` (int): Rating produk (1-5)
- `sentiment` (string): Label sentiment - "positive", "neutral", atau "negative"
### Data Splits
Dataset terdiri dari:
- **simple.json**: 17,000 samples (data utama)
- **challange.json**: 4,840 samples (data challenging termasuk sarkasme)
- **Total**: 21,840 samples
### Data Distribution
- Positive: 7,480 (34.3%)
- Negative: 7,470 (34.2%)
- Neutral: 6,890 (31.5%)
## Usage
```python
from datasets import load_dataset
# Load dataset
dataset = load_dataset("AIbnuHibban/e-commerce-sentiment-bahasa-indonesia")
# Or load from JSON files directly
import json
import pandas as pd
with open('simple.json', 'r') as f:
data = json.load(f)
df = pd.DataFrame(data)
```
## Example Data
```json
{
"comment": "Barang mantap sekali! Worth it banget dengan harganya",
"rating": 5,
"sentiment": "positive"
}
```
```json
{
"comment": "Wow, senang banget ditipu! Barang jelek tapi harganya mahal",
"rating": 1,
"sentiment": "negative"
}
```
## Characteristics
- **Language**: Bahasa Indonesia (Indonesian)
- **Domain**: E-commerce product reviews
- **Special Features**: Includes sarcastic and ironic comments
- **Use Cases**:
- Sentiment analysis
- Opinion mining
- Customer feedback analysis
- Sarcasm detection
## Citation
If you use this dataset, please cite:
```
@dataset{e_commerce_sentiment_bahasa_indonesia,
author = {AIbnuHibban},
title = {E-Commerce Sentiment Analysis Dataset (Indonesian)},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/AIbnuHibban/e-commerce-sentiment-bahasa-indonesia}
}
```
## License
MIT License
| 1 | 0 | [
"task_categories:text-classification",
"language:id",
"license:mit",
"size_categories:10K<n<100K",
"format:json",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"region:us",
"sentiment-analysis",
"indonesian",
"e-commerce",
"reviews"
] | 2025-11-11T14:32:10+00:00 | 2025-11-11T14:43:04+00:00 | 0 |
open-law-data-thailand/soc-ratchakitcha |
# ชุดข้อมูลราชกิจจานุเบกษาไทย (Thai Royal Gazette Dataset)
**คลังข้อมูลนี้เป็นส่วนหนึ่งของ [โครงการ Open Law Data](https://github.com/DGA-Thailand/open-law-data)**
## Dataset Description
ชุดข้อมูลนี้คือคลังเอกสาร **"ราชกิจจานุเบกษา"** ของประเทศไทยฉบับดิจิทัลในรูปแบบ PDF และในรูปแบบที่ผ่านการประมวลผลด้วยเทคโนโลยี OCR (Optical Character Recognition) เพื่อแปลงเอกสาร PDF ให้อยู่ในรูปแบบข้อความ (Text) ที่คอมพิวเตอร์สามารถอ่านและนำไปวิเคราะห์ต่อได้
ข้อมูลครอบคลุมประกาศในราชกิจจานุเบกษาตั้งแต่ปี พ.ศ. 2427 - ปัจจุบัน โดยมีเป้าหมายเพื่อเป็นข้อมูลตั้งต้นสำหรับนักพัฒนา, นักวิจัย, นักกฎหมาย, และประชาชนทั่วไป ในการสร้างสรรค์นวัตกรรมด้านกฎหมาย (Legal Tech) และส่งเสริมความโปร่งใสในภาครัฐ
|
# ชุดข้อมูลราชกิจจานุเบกษาไทย (Thai Royal Gazette Dataset)
**คลังข้อมูลนี้เป็นส่วนหนึ่งของ [โครงการ Open Law Data](https://github.com/DGA-Thailand/open-law-data)**
## Dataset Description
ชุดข้อมูลนี้คือคลังเอกสาร **"ราชกิจจานุเบกษา"** ของประเทศไทยฉบับดิจิทัลในรูปแบบ PDF และในรูปแบบที่ผ่านการประมวลผลด้วยเทคโนโลยี OCR (Optical Character Recognition) เพื่อแปลงเอกสาร PDF ให้อยู่ในรูปแบบข้อความ (Text) ที่คอมพิวเตอร์สามารถอ่านและนำไปวิเคราะห์ต่อได้
ข้อมูลครอบคลุมประกาศในราชกิจจานุเบกษาตั้งแต่ปี พ.ศ. 2427 - ปัจจุบัน โดยมีเป้าหมายเพื่อเป็นข้อมูลตั้งต้นสำหรับนักพัฒนา, นักวิจัย, นักกฎหมาย, และประชาชนทั่วไป ในการสร้างสรรค์นวัตกรรมด้านกฎหมาย (Legal Tech) และส่งเสริมความโปร่งใสในภาครัฐ
| 11,172 | 0 | [
"language:th",
"license:cc-by-sa-4.0",
"size_categories:1M<n<10M",
"region:us",
"law",
"legal",
"thai",
"government",
"open-data"
] | 2025-10-29T16:43:15+00:00 | 2025-11-11T14:41:13+00:00 | 0 |
ESmike/med_qa_offline |
# Dataset Card for MedQA Offline
This is an Adaptation of the MedQA Dataset without the need for [scripts which are no longer supported by huggingface](https://discuss.huggingface.co/t/dataset-scripts-are-no-longer-supported/163891).
Below is the original dataset description:
## Dataset Description
- **Homepage:** https://github.com/jind11/MedQA
- **Original HF:** https://huggingface.co/datasets/bigbio/med_qa
- **Pubmed:** False
- **Public:** True
- **Tasks:** QA
In this work, we present the first free-form multiple-choice OpenQA dataset for solving medical problems, MedQA,
collected from the professional medical board exams. It covers three languages: English, simplified Chinese, and
traditional Chinese, and contains 12,723, 34,251, and 14,123 questions for the three languages, respectively. Together
with the question data, we also collect and release a large-scale corpus from medical textbooks from which the reading
comprehension models can obtain necessary knowledge for answering the questions.
## Citation Information
```
@article{jin2021disease,
title={What disease does this patient have? a large-scale open domain question answering dataset from medical exams},
author={Jin, Di and Pan, Eileen and Oufattole, Nassim and Weng, Wei-Hung and Fang, Hanyi and Szolovits, Peter},
journal={Applied Sciences},
volume={11},
number={14},
pages={6421},
year={2021},
publisher={MDPI}
}
``` |
# Dataset Card for MedQA Offline
This is an Adaptation of the MedQA Dataset without the need for [scripts which are no longer supported by huggingface](https://discuss.huggingface.co/t/dataset-scripts-are-no-longer-supported/163891).
Below is the original dataset description:
## Dataset Description
- **Homepage:** https://github.com/jind11/MedQA
- **Original HF:** https://huggingface.co/datasets/bigbio/med_qa
- **Pubmed:** False
- **Public:** True
- **Tasks:** QA
In this work, we present the first free-form multiple-choice OpenQA dataset for solving medical problems, MedQA,
collected from the professional medical board exams. It covers three languages: English, simplified Chinese, and
traditional Chinese, and contains 12,723, 34,251, and 14,123 questions for the three languages, respectively. Together
with the question data, we also collect and release a large-scale corpus from medical textbooks from which the reading
comprehension models can obtain necessary knowledge for answering the questions.
## Citation Information
```
@article{jin2021disease,
title={What disease does this patient have? a large-scale open domain question answering dataset from medical exams},
author={Jin, Di and Pan, Eileen and Oufattole, Nassim and Weng, Wei-Hung and Fang, Hanyi and Szolovits, Peter},
journal={Applied Sciences},
volume={11},
number={14},
pages={6421},
year={2021},
publisher={MDPI}
}
``` | 36 | 0 | [
"multilinguality:multilingual",
"language:en",
"language:zh",
"license:unknown",
"size_categories:100K<n<1M",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T10:45:23+00:00 | 2025-11-11T14:36:15+00:00 | 0 |
TheFactoryX/edition_0312_tatsu-lab-alpaca-readymade |
# edition_0312_tatsu-lab-alpaca-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[tatsu-lab/alpaca](https://huggingface.co/datasets/tatsu-lab/alpaca)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0312_tatsu-lab-alpaca-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[tatsu-lab/alpaca](https://huggingface.co/datasets/tatsu-lab/alpaca)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 5 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T14:36:09+00:00 | 2025-11-11T14:36:10+00:00 | 0 |
hiepp2/tvp4 |
<img src="mot-thumbnail.png" alt="Centered Image" style="display: block; margin: 0 auto;" width="500">
# Dataset summary
Mixture-of-Thoughts is a curated dataset of 350k verified reasoning traces distilled from [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1). The dataset spans tasks in mathematics, coding, and science, and is designed to teach language models to reason step-by-step. It was used in the Open R1 project to train [OpenR1-Distill-7B](https://huggingface.co/open-r1/OpenR1-Distill-7B), an SFT model that replicates the reasoning capabilities of [deepseek-ai/DeepSeek-R1-Distill-Qwen-7B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B) from the same base model.
To load the dataset, run:
```python
from datasets import load_dataset
dataset = load_dataset("open-r1/Mixture-of-Thoughts", "all", split="train")
# Load a specific domain
dataset_math = load_dataset("open-r1/Mixture-of-Thoughts", "math", split="train")
```
## Dataset composition
Mixture-of-Thoughts is composed of three domains: math, code, and science. Each domain contains reasoning traces that are designed to teach language models to reason step-by-step. The dataset is structured as follows:
- **math**: 93.7k reasoning traces for mathematical problems, sourced from the `default` subset of [open-r1/OpenR1-Math-220k](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k)
- **code**: 83.1k reasoning traces for competitive programming problems in Python and C++, sourced from the `solutions` and `solutions_w_editorials` subsets of [open-r1/codeforces-cots](https://huggingface.co/datasets/open-r1/codeforces-cots)
- **science**: 173k reasoning traces for scientific problems, sourced from the `science` subset of [nvidia/Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset)
- **all**: Contains all reasoning traces from the three domains, for a total of 350k traces.
## Curation methodology
To optimise the data mixture, we followed the same methodology described in the [Phi-4-reasoning tech report](https://huggingface.co/papers/2504.21318), namely that mixtures can be optimised independently per domain, and then combined into a single dataset. For each ablation, we evaluate on AIME 2024, GPQA Diamond, and LiveCodeBench v4 every epoch and take the best performing model checkpoint. The figure below shows the results from post-training [open-r1/Qwen2.5-Math-7B-RoPE-300k](https://huggingface.co/open-r1/Qwen2.5-Math-7B-RoPE-300k) on each individual domain compared to the final mixture:
<img src="data_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
Overall, we find that training on all domains simultaneously yields the best results. See the subsections below for more details on optimising the data mixture per domain.
> [!NOTE]
> We use LiveCodeBench v4 to accelerate evaluation during our ablations as it contains around half the problems of v5, yet is still representative of the full benchmark.
### Code
During the development of [open-r1/OlympicCoder-7B](https://huggingface.co/open-r1/OlympicCoder-7B), we observed that generating R1 reasoning traces in C++ produced better results on the challenging [IOI 2024 benchmark](https://github.com/huggingface/ioi), while Python traces produced better results on LiveCodeBench (a Python-only benchmark). To optimise the data mixture, we therefore used a mix of C++ and Python traces sourced from the following subsets of [open-r1/codeforces-cots](https://huggingface.co/datasets/open-r1/codeforces-cots):
- `solutions`: we prompt R1 to solve the problem and produce code in C++.
- `solutions_py`: same as `solutions`, but with R1 prompted to produce code in Python.
- `solutions_w_editorials`: we prompt R1 to solve the problem and produce code, but also provide it with a human-written solution.
- `solutions_w_editorials_py`: same as `solutions_w_editorials`, but with R1 prompted to produce code in Python.
The figure below shows the evolution of our ablations on these subsets, using [Qwen/Qwen2.5-Coder-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct) as the base model:
<img src="code_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
The individual experiments correspond to the following:
* **exp1 - exp3:** scaling the learning rate on the `solutions` subset from 1e-5 to 2e-5, and 4e-5 respectively.
* **exp4 - exp5:** measuring the impact of training on the `solutions_w_editorials` subset vs the combined `solutions` and `solutions_w_editorials` subsets.
* **exp6 - exp9:** measuring the impact of blending in Python traces from the `solutions_py` and `solutions_w_editorials_py` subsets. exp6 combines the `solutions_w_editorials` and `solutions_w_editorials_py` subsets, while exp7 combines the `solutions` and `solutions_py` subsets. Finally, exp8 combines all four subsets.
We found that combining all subsets of C++ and Python traces yielded the best results on LiveCodeBench. We also found that using this data mixture to fine-tune [open-r1/Qwen2.5-Coder-7B-RoPE-300k](https://huggingface.co/open-r1/Qwen2.5-Coder-7B-RoPE-300k) led to comparable performance improvements, which shows the effectiveness of our curation strategy.
### Math
For the math domain, we mostly focused on comparing the `default` and `extended` subsets of [open-r1/OpenR1-Math-220k](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k). The `default` subset contains 93.7k reasoning traces, while the `extended` subset contains an additional 131k traces, containing simpler problems than the `default` subset. The figure below shows performance on each subset, using [Qwen/Qwen2.5-Math-7B-RoPE-300k](https://huggingface.co/Qwen/Qwen2.5-Math-7B-RoPE-300k) as the base model:
<img src="math_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
Overall, we found that training on the `default` subset yielded better results than training on the `extended` subset, and that training on both subsets together yielded the best results. Nevertheless, we opted to use the `default` subset only for the final mixture, as including both would have led to a significant increase in the size of the dataset, for a modest improvement in performance.
### Science
For the science domain, we used the `science` subset of [nvidia/Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset/viewer/SFT/science), which contains 483k reasoning traces. However, we found that the subset was too large to be used in its entirety, as it would have led to a significant increase in the size of the dataset. Instead, we selected the subset of traces where no Qwen models were used for prompt pre-processing--see this [discussion](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset/discussions/6) for more details. The result was 173k reasoning traces, which we used in the final mixture after ablating on the learning rate.
## Citation
If you find this dataset is useful in your own work, please consider citing it as follows, together with the source of the specific domain you are using:
```bibtex
@misc{openr1,
title = {Open R1: A fully open reproduction of DeepSeek-R1},
url = {https://github.com/huggingface/open-r1},
author = {Hugging Face},
month = {January},
year = {2025}
}
```
**open-r1/codeforces-cots**
```bibtex
@misc{penedo2025codeforces,
title={CodeForces CoTs},
author={Guilherme Penedo and Anton Lozhkov and Hynek Kydlíček and Loubna Ben Allal and Edward Beeching and Agustín Piqueres Lajarín and Quentin Gallouédec and Nathan Habib and Lewis Tunstall and Leandro von Werra},
year={2025},
publisher = {Hugging Face},
journal = {Hugging Face repository},
howpublished = {\url{https://huggingface.co/datasets/open-r1/codeforces-cots}}
}
```
**open-r1/OpenR1-Math-220k**
```bibtex
@misc{lozhkov2025openr1math220k,
title={OpenR1-Math-220k},
author={Anton Lozhkov and Hynek Kydlíček and Loubna Ben Allal and Guilherme Penedo and Edward Beeching and Quentin Gallouédec and Nathan Habib and Lewis Tunstall and Leandro von Werra},
year={2025},
publisher = {Hugging Face},
journal = {Hugging Face repository},
howpublished = {\url{https://huggingface.co/datasets/open-r1/OpenR1-Math-220k}}
}
```
**nvidia/Llama-Nemotron-Post-Training-Dataset**
```bibtex
@misc{bercovich2025llamanemotronefficientreasoningmodels,
title={Llama-Nemotron: Efficient Reasoning Models},
author={Akhiad Bercovich and Itay Levy and Izik Golan and Mohammad Dabbah and Ran El-Yaniv and Omri Puny and Ido Galil and Zach Moshe and Tomer Ronen and Najeeb Nabwani and Ido Shahaf and Oren Tropp and Ehud Karpas and Ran Zilberstein and Jiaqi Zeng and Soumye Singhal and Alexander Bukharin and Yian Zhang and Tugrul Konuk and Gerald Shen and Ameya Sunil Mahabaleshwarkar and Bilal Kartal and Yoshi Suhara and Olivier Delalleau and Zijia Chen and Zhilin Wang and David Mosallanezhad and Adi Renduchintala and Haifeng Qian and Dima Rekesh and Fei Jia and Somshubra Majumdar and Vahid Noroozi and Wasi Uddin Ahmad and Sean Narenthiran and Aleksander Ficek and Mehrzad Samadi and Jocelyn Huang and Siddhartha Jain and Igor Gitman and Ivan Moshkov and Wei Du and Shubham Toshniwal and George Armstrong and Branislav Kisacanin and Matvei Novikov and Daria Gitman and Evelina Bakhturina and Jane Polak Scowcroft and John Kamalu and Dan Su and Kezhi Kong and Markus Kliegl and Rabeeh Karimi and Ying Lin and Sanjeev Satheesh and Jupinder Parmar and Pritam Gundecha and Brandon Norick and Joseph Jennings and Shrimai Prabhumoye and Syeda Nahida Akter and Mostofa Patwary and Abhinav Khattar and Deepak Narayanan and Roger Waleffe and Jimmy Zhang and Bor-Yiing Su and Guyue Huang and Terry Kong and Parth Chadha and Sahil Jain and Christine Harvey and Elad Segal and Jining Huang and Sergey Kashirsky and Robert McQueen and Izzy Putterman and George Lam and Arun Venkatesan and Sherry Wu and Vinh Nguyen and Manoj Kilaru and Andrew Wang and Anna Warno and Abhilash Somasamudramath and Sandip Bhaskar and Maka Dong and Nave Assaf and Shahar Mor and Omer Ullman Argov and Scot Junkin and Oleksandr Romanenko and Pedro Larroy and Monika Katariya and Marco Rovinelli and Viji Balas and Nicholas Edelman and Anahita Bhiwandiwalla and Muthu Subramaniam and Smita Ithape and Karthik Ramamoorthy and Yuting Wu and Suguna Varshini Velury and Omri Almog and Joyjit Daw and Denys Fridman and Erick Galinkin and Michael Evans and Katherine Luna and Leon Derczynski and Nikki Pope and Eileen Long and Seth Schneider and Guillermo Siman and Tomasz Grzegorzek and Pablo Ribalta and Monika Katariya and Joey Conway and Trisha Saar and Ann Guan and Krzysztof Pawelec and Shyamala Prayaga and Oleksii Kuchaiev and Boris Ginsburg and Oluwatobi Olabiyi and Kari Briski and Jonathan Cohen and Bryan Catanzaro and Jonah Alben and Yonatan Geifman and Eric Chung and Chris Alexiuk},
year={2025},
eprint={2505.00949},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2505.00949},
}
``` |
<img src="mot-thumbnail.png" alt="Centered Image" style="display: block; margin: 0 auto;" width="500">
# Dataset summary
Mixture-of-Thoughts is a curated dataset of 350k verified reasoning traces distilled from [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1). The dataset spans tasks in mathematics, coding, and science, and is designed to teach language models to reason step-by-step. It was used in the Open R1 project to train [OpenR1-Distill-7B](https://huggingface.co/open-r1/OpenR1-Distill-7B), an SFT model that replicates the reasoning capabilities of [deepseek-ai/DeepSeek-R1-Distill-Qwen-7B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B) from the same base model.
To load the dataset, run:
```python
from datasets import load_dataset
dataset = load_dataset("open-r1/Mixture-of-Thoughts", "all", split="train")
# Load a specific domain
dataset_math = load_dataset("open-r1/Mixture-of-Thoughts", "math", split="train")
```
## Dataset composition
Mixture-of-Thoughts is composed of three domains: math, code, and science. Each domain contains reasoning traces that are designed to teach language models to reason step-by-step. The dataset is structured as follows:
- **math**: 93.7k reasoning traces for mathematical problems, sourced from the `default` subset of [open-r1/OpenR1-Math-220k](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k)
- **code**: 83.1k reasoning traces for competitive programming problems in Python and C++, sourced from the `solutions` and `solutions_w_editorials` subsets of [open-r1/codeforces-cots](https://huggingface.co/datasets/open-r1/codeforces-cots)
- **science**: 173k reasoning traces for scientific problems, sourced from the `science` subset of [nvidia/Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset)
- **all**: Contains all reasoning traces from the three domains, for a total of 350k traces.
## Curation methodology
To optimise the data mixture, we followed the same methodology described in the [Phi-4-reasoning tech report](https://huggingface.co/papers/2504.21318), namely that mixtures can be optimised independently per domain, and then combined into a single dataset. For each ablation, we evaluate on AIME 2024, GPQA Diamond, and LiveCodeBench v4 every epoch and take the best performing model checkpoint. The figure below shows the results from post-training [open-r1/Qwen2.5-Math-7B-RoPE-300k](https://huggingface.co/open-r1/Qwen2.5-Math-7B-RoPE-300k) on each individual domain compared to the final mixture:
<img src="data_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
Overall, we find that training on all domains simultaneously yields the best results. See the subsections below for more details on optimising the data mixture per domain.
> [!NOTE]
> We use LiveCodeBench v4 to accelerate evaluation during our ablations as it contains around half the problems of v5, yet is still representative of the full benchmark.
### Code
During the development of [open-r1/OlympicCoder-7B](https://huggingface.co/open-r1/OlympicCoder-7B), we observed that generating R1 reasoning traces in C++ produced better results on the challenging [IOI 2024 benchmark](https://github.com/huggingface/ioi), while Python traces produced better results on LiveCodeBench (a Python-only benchmark). To optimise the data mixture, we therefore used a mix of C++ and Python traces sourced from the following subsets of [open-r1/codeforces-cots](https://huggingface.co/datasets/open-r1/codeforces-cots):
- `solutions`: we prompt R1 to solve the problem and produce code in C++.
- `solutions_py`: same as `solutions`, but with R1 prompted to produce code in Python.
- `solutions_w_editorials`: we prompt R1 to solve the problem and produce code, but also provide it with a human-written solution.
- `solutions_w_editorials_py`: same as `solutions_w_editorials`, but with R1 prompted to produce code in Python.
The figure below shows the evolution of our ablations on these subsets, using [Qwen/Qwen2.5-Coder-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct) as the base model:
<img src="code_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
The individual experiments correspond to the following:
* **exp1 - exp3:** scaling the learning rate on the `solutions` subset from 1e-5 to 2e-5, and 4e-5 respectively.
* **exp4 - exp5:** measuring the impact of training on the `solutions_w_editorials` subset vs the combined `solutions` and `solutions_w_editorials` subsets.
* **exp6 - exp9:** measuring the impact of blending in Python traces from the `solutions_py` and `solutions_w_editorials_py` subsets. exp6 combines the `solutions_w_editorials` and `solutions_w_editorials_py` subsets, while exp7 combines the `solutions` and `solutions_py` subsets. Finally, exp8 combines all four subsets.
We found that combining all subsets of C++ and Python traces yielded the best results on LiveCodeBench. We also found that using this data mixture to fine-tune [open-r1/Qwen2.5-Coder-7B-RoPE-300k](https://huggingface.co/open-r1/Qwen2.5-Coder-7B-RoPE-300k) led to comparable performance improvements, which shows the effectiveness of our curation strategy.
### Math
For the math domain, we mostly focused on comparing the `default` and `extended` subsets of [open-r1/OpenR1-Math-220k](https://huggingface.co/datasets/open-r1/OpenR1-Math-220k). The `default` subset contains 93.7k reasoning traces, while the `extended` subset contains an additional 131k traces, containing simpler problems than the `default` subset. The figure below shows performance on each subset, using [Qwen/Qwen2.5-Math-7B-RoPE-300k](https://huggingface.co/Qwen/Qwen2.5-Math-7B-RoPE-300k) as the base model:
<img src="math_mix.png" alt="Centered Image" style="display: block; margin: 0 auto;">
Overall, we found that training on the `default` subset yielded better results than training on the `extended` subset, and that training on both subsets together yielded the best results. Nevertheless, we opted to use the `default` subset only for the final mixture, as including both would have led to a significant increase in the size of the dataset, for a modest improvement in performance.
### Science
For the science domain, we used the `science` subset of [nvidia/Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset/viewer/SFT/science), which contains 483k reasoning traces. However, we found that the subset was too large to be used in its entirety, as it would have led to a significant increase in the size of the dataset. Instead, we selected the subset of traces where no Qwen models were used for prompt pre-processing--see this [discussion](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset/discussions/6) for more details. The result was 173k reasoning traces, which we used in the final mixture after ablating on the learning rate.
## Citation
If you find this dataset is useful in your own work, please consider citing it as follows, together with the source of the specific domain you are using:
```bibtex
@misc{openr1,
title = {Open R1: A fully open reproduction of DeepSeek-R1},
url = {https://github.com/huggingface/open-r1},
author = {Hugging Face},
month = {January},
year = {2025}
}
```
**open-r1/codeforces-cots**
```bibtex
@misc{penedo2025codeforces,
title={CodeForces CoTs},
author={Guilherme Penedo and Anton Lozhkov and Hynek Kydlíček and Loubna Ben Allal and Edward Beeching and Agustín Piqueres Lajarín and Quentin Gallouédec and Nathan Habib and Lewis Tunstall and Leandro von Werra},
year={2025},
publisher = {Hugging Face},
journal = {Hugging Face repository},
howpublished = {\url{https://huggingface.co/datasets/open-r1/codeforces-cots}}
}
```
**open-r1/OpenR1-Math-220k**
```bibtex
@misc{lozhkov2025openr1math220k,
title={OpenR1-Math-220k},
author={Anton Lozhkov and Hynek Kydlíček and Loubna Ben Allal and Guilherme Penedo and Edward Beeching and Quentin Gallouédec and Nathan Habib and Lewis Tunstall and Leandro von Werra},
year={2025},
publisher = {Hugging Face},
journal = {Hugging Face repository},
howpublished = {\url{https://huggingface.co/datasets/open-r1/OpenR1-Math-220k}}
}
```
**nvidia/Llama-Nemotron-Post-Training-Dataset**
```bibtex
@misc{bercovich2025llamanemotronefficientreasoningmodels,
title={Llama-Nemotron: Efficient Reasoning Models},
author={Akhiad Bercovich and Itay Levy and Izik Golan and Mohammad Dabbah and Ran El-Yaniv and Omri Puny and Ido Galil and Zach Moshe and Tomer Ronen and Najeeb Nabwani and Ido Shahaf and Oren Tropp and Ehud Karpas and Ran Zilberstein and Jiaqi Zeng and Soumye Singhal and Alexander Bukharin and Yian Zhang and Tugrul Konuk and Gerald Shen and Ameya Sunil Mahabaleshwarkar and Bilal Kartal and Yoshi Suhara and Olivier Delalleau and Zijia Chen and Zhilin Wang and David Mosallanezhad and Adi Renduchintala and Haifeng Qian and Dima Rekesh and Fei Jia and Somshubra Majumdar and Vahid Noroozi and Wasi Uddin Ahmad and Sean Narenthiran and Aleksander Ficek and Mehrzad Samadi and Jocelyn Huang and Siddhartha Jain and Igor Gitman and Ivan Moshkov and Wei Du and Shubham Toshniwal and George Armstrong and Branislav Kisacanin and Matvei Novikov and Daria Gitman and Evelina Bakhturina and Jane Polak Scowcroft and John Kamalu and Dan Su and Kezhi Kong and Markus Kliegl and Rabeeh Karimi and Ying Lin and Sanjeev Satheesh and Jupinder Parmar and Pritam Gundecha and Brandon Norick and Joseph Jennings and Shrimai Prabhumoye and Syeda Nahida Akter and Mostofa Patwary and Abhinav Khattar and Deepak Narayanan and Roger Waleffe and Jimmy Zhang and Bor-Yiing Su and Guyue Huang and Terry Kong and Parth Chadha and Sahil Jain and Christine Harvey and Elad Segal and Jining Huang and Sergey Kashirsky and Robert McQueen and Izzy Putterman and George Lam and Arun Venkatesan and Sherry Wu and Vinh Nguyen and Manoj Kilaru and Andrew Wang and Anna Warno and Abhilash Somasamudramath and Sandip Bhaskar and Maka Dong and Nave Assaf and Shahar Mor and Omer Ullman Argov and Scot Junkin and Oleksandr Romanenko and Pedro Larroy and Monika Katariya and Marco Rovinelli and Viji Balas and Nicholas Edelman and Anahita Bhiwandiwalla and Muthu Subramaniam and Smita Ithape and Karthik Ramamoorthy and Yuting Wu and Suguna Varshini Velury and Omri Almog and Joyjit Daw and Denys Fridman and Erick Galinkin and Michael Evans and Katherine Luna and Leon Derczynski and Nikki Pope and Eileen Long and Seth Schneider and Guillermo Siman and Tomasz Grzegorzek and Pablo Ribalta and Monika Katariya and Joey Conway and Trisha Saar and Ann Guan and Krzysztof Pawelec and Shyamala Prayaga and Oleksii Kuchaiev and Boris Ginsburg and Oluwatobi Olabiyi and Kari Briski and Jonathan Cohen and Bryan Catanzaro and Jonah Alben and Yonatan Geifman and Eric Chung and Chris Alexiuk},
year={2025},
eprint={2505.00949},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2505.00949},
}
``` | 17,570 | 1 | [
"task_categories:text-generation",
"language:en",
"size_categories:n>1T",
"arxiv:2504.21318",
"arxiv:2505.00949",
"region:us"
] | 2025-06-02T23:40:29+00:00 | 2025-11-11T14:30:34+00:00 | 0 |
iAyoD/robocasa_turn_on_microwave_256 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 12061,
"total_tasks": 1,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 12061,
"total_tasks": 1,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 52 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:image",
"modality:timeseries",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot",
"robocasa",
"panda"
] | 2025-11-11T14:29:57+00:00 | 2025-11-11T14:30:20+00:00 | 0 |
CVML-TueAI/grounding-YT-dataset | ## Example usage for clips:
### Also decoding raw binary video data and json
```python
import webdataset as wds
from huggingface_hub import HfFileSystem, get_token, hf_hub_url
import json
import io
import torch
import av
import numpy as np
from torch.utils.data import DataLoader
fs = HfFileSystem()
files = [fs.resolve_path(path) for path in fs.glob("hf://datasets/CVML-TueAI/grounding-YT-dataset/clips/*.tar")]
urls = [hf_hub_url(file.repo_id, file.path_in_repo, repo_type="dataset") for file in files]
urls = f"pipe: curl -s -L -H 'Authorization:Bearer {get_token()}' {'::'.join(urls)}"
def load_video(video_bytes):
container = av.open(io.BytesIO(video_bytes))
frames = []
for frame in container.decode(video=0):
img = frame.to_ndarray(format="rgb24")
frames.append(img)
video_tensor = torch.from_numpy(np.stack(frames))
return video_tensor #[T, H, W, C]
def load_json(json_bytes):
"""Decode JSON metadata"""
return json.loads(json_bytes.decode("utf-8"))
dataset = (
wds.WebDataset(urls,)
.shuffle(100)
.to_tuple("mp4", "json")
.map_tuple(load_video, load_json)
) | ## Example usage for clips:
### Also decoding raw binary video data and json
```python
import webdataset as wds
from huggingface_hub import HfFileSystem, get_token, hf_hub_url
import json
import io
import torch
import av
import numpy as np
from torch.utils.data import DataLoader
fs = HfFileSystem()
files = [fs.resolve_path(path) for path in fs.glob("hf://datasets/CVML-TueAI/grounding-YT-dataset/clips/*.tar")]
urls = [hf_hub_url(file.repo_id, file.path_in_repo, repo_type="dataset") for file in files]
urls = f"pipe: curl -s -L -H 'Authorization:Bearer {get_token()}' {'::'.join(urls)}"
def load_video(video_bytes):
container = av.open(io.BytesIO(video_bytes))
frames = []
for frame in container.decode(video=0):
img = frame.to_ndarray(format="rgb24")
frames.append(img)
video_tensor = torch.from_numpy(np.stack(frames))
return video_tensor #[T, H, W, C]
def load_json(json_bytes):
"""Decode JSON metadata"""
return json.loads(json_bytes.decode("utf-8"))
dataset = (
wds.WebDataset(urls,)
.shuffle(100)
.to_tuple("mp4", "json")
.map_tuple(load_video, load_json)
) | 145 | 0 | [
"size_categories:10K<n<100K",
"format:webdataset",
"modality:image",
"modality:text",
"library:datasets",
"library:webdataset",
"library:mlcroissant",
"region:us"
] | 2025-11-04T14:01:17+00:00 | 2025-11-11T14:28:19+00:00 | 0 |
hypha-space/mnist |
### Licensing Information
MIT Licence
### Citation Information
```
@article{lecun2010mnist,
title={MNIST handwritten digit database},
author={LeCun, Yann and Cortes, Corinna and Burges, CJ},
journal={ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist},
volume={2},
year={2010}
}
``` |
### Licensing Information
MIT Licence
### Citation Information
```
@article{lecun2010mnist,
title={MNIST handwritten digit database},
author={LeCun, Yann and Cortes, Corinna and Burges, CJ},
journal={ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist},
volume={2},
year={2010}
}
``` | 44 | 0 | [
"license:mit",
"region:us"
] | 2025-11-11T14:01:15+00:00 | 2025-11-11T14:24:36+00:00 | 0 |
mokyu2106/iroiro_data | ■■LECO&DIFF置き場■■
主にXLで使用するLECOが格納されています。
作成者の都合上、数としてはIllustriousXLv01関連が一番充実しています
※2025/8/29 更新
hakushiMix_v14.1 向けのLECO量産開始しました。
随時追加していきます。
※簡易な使い方説明は下位フォルダ内txt参照の事 | ■■LECO&DIFF置き場■■
主にXLで使用するLECOが格納されています。
作成者の都合上、数としてはIllustriousXLv01関連が一番充実しています
※2025/8/29 更新
hakushiMix_v14.1 向けのLECO量産開始しました。
随時追加していきます。
※簡易な使い方説明は下位フォルダ内txt参照の事 | 886 | 44 | [
"license:unknown",
"region:us"
] | 2024-03-15T08:51:14+00:00 | 2025-11-11T14:29:15+00:00 | 0 |
iAyoD/robocasa_close_single_door_256 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 13279,
"total_tasks": 2,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 13279,
"total_tasks": 2,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 35 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:image",
"modality:timeseries",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot",
"robocasa",
"panda"
] | 2025-11-11T14:19:35+00:00 | 2025-11-11T14:19:59+00:00 | 0 |
FM4CS/THOR-Pretrain |
# THOR Pretraining dataset
Code to load data and details to come soon...
### Attribution
THOR-Pretrain is part of the FM4CS project funded by the European Space Agency Φ‑Lab (contract #4000143489/24/I-DT).
Contains modified Copernicus Sentinel data (2016–2024).
© ESA WorldCover project 2020 / Contains modified Copernicus Sentinel data (2020) processed by the ESA WorldCover consortium.
© ESA 2010 and Université catholique de Louvain (UCLouvain) — GlobCover.
We acknowledge the use of data products from NASA’s Land Processes Distributed Active Archive Center (LP DAAC), USGS/EROS (DOI: 10.5067/MODIS/MCD12Q1.061).
DEM data from European Union’s Copernicus Land Monitoring Service information (EU-DEM v1.0).
|
# THOR Pretraining dataset
Code to load data and details to come soon...
### Attribution
THOR-Pretrain is part of the FM4CS project funded by the European Space Agency Φ‑Lab (contract #4000143489/24/I-DT).
Contains modified Copernicus Sentinel data (2016–2024).
© ESA WorldCover project 2020 / Contains modified Copernicus Sentinel data (2020) processed by the ESA WorldCover consortium.
© ESA 2010 and Université catholique de Louvain (UCLouvain) — GlobCover.
We acknowledge the use of data products from NASA’s Land Processes Distributed Active Archive Center (LP DAAC), USGS/EROS (DOI: 10.5067/MODIS/MCD12Q1.061).
DEM data from European Union’s Copernicus Land Monitoring Service information (EU-DEM v1.0).
| 1,991 | 0 | [
"task_categories:image-feature-extraction",
"license:cc-by-sa-4.0",
"size_categories:10K<n<100K",
"region:us",
"climate",
"earth observation",
"pre-training"
] | 2025-11-06T18:09:53+00:00 | 2025-11-11T14:15:06+00:00 | 0 |
sandbagging-games/evaluation_logs |
# Evaluation logs from "Adversarial Games for Evaluation Sabotage"
<!-- Provide a quick summary of the dataset. -->
This dataset provides evaluation transcripts produced for the paper "Adversarial Games for Evaluation Sabotage" [TODO link to paper once out]. Transcripts are provided in [Inspect](https://inspect.aisi.org.uk/) `.eval` format, see [https://github.com/AI-Safety-Institute/sabotage_games](https://github.com/AI-Safety-Institute/sabotage_games) for a guide to viewing them.
## Dataset Details
- `evaluation_transcripts/handover_evals` contains the transcripts provided by the red team to the blue team at the beginning of the main round of the game, showing the performance of each of the unaltered model organisms.
- `evaluation_transcripts/trial_handover_evals` contains the transcripts provided by the red team to the blue team at the beginning of the trial round of the game.
- `evaluation_transcripts/additional_model_evaluations` contains further transcripts, including results from password-unlocked model organisms, the Willow model organism that was not included in the game, and Llama 3.3 70B Instruct.
- `experiments/auditing_agent` contains the output of the automated auditing agent used by the blue team.
- `experiments/non_assistant` contains the output of the user persona attack experiment.
- `experiments/on_distribution_cross_distribution_training` contains the evaluation logs of the single correct sample finetuning experiment.
- `experiments/deception_probes` contains scores for middle and final layer deception probes run on the handover evaluation transcripts. |
# Evaluation logs from "Adversarial Games for Evaluation Sabotage"
<!-- Provide a quick summary of the dataset. -->
This dataset provides evaluation transcripts produced for the paper "Adversarial Games for Evaluation Sabotage" [TODO link to paper once out]. Transcripts are provided in [Inspect](https://inspect.aisi.org.uk/) `.eval` format, see [https://github.com/AI-Safety-Institute/sabotage_games](https://github.com/AI-Safety-Institute/sabotage_games) for a guide to viewing them.
## Dataset Details
- `evaluation_transcripts/handover_evals` contains the transcripts provided by the red team to the blue team at the beginning of the main round of the game, showing the performance of each of the unaltered model organisms.
- `evaluation_transcripts/trial_handover_evals` contains the transcripts provided by the red team to the blue team at the beginning of the trial round of the game.
- `evaluation_transcripts/additional_model_evaluations` contains further transcripts, including results from password-unlocked model organisms, the Willow model organism that was not included in the game, and Llama 3.3 70B Instruct.
- `experiments/auditing_agent` contains the output of the automated auditing agent used by the blue team.
- `experiments/non_assistant` contains the output of the user persona attack experiment.
- `experiments/on_distribution_cross_distribution_training` contains the evaluation logs of the single correct sample finetuning experiment.
- `experiments/deception_probes` contains scores for middle and final layer deception probes run on the handover evaluation transcripts. | 322 | 0 | [
"region:us"
] | 2025-11-07T12:15:01+00:00 | 2025-11-11T14:14:09+00:00 | 0 |
ogaufi/NLP_PuoAi |
Key,Description,Example
lang,The name of the language.,Setswana
code,The ISO 639-3 code for the language (or a common reference code).,TND
en,The English source phrase.,How are you? (Morning)
st,The target phrase in the respective Botswana language.,O tsogile jang?
notes,Context or literal translation (Optional).,"Lit. ""How did you wake up?""" |
Key,Description,Example
lang,The name of the language.,Setswana
code,The ISO 639-3 code for the language (or a common reference code).,TND
en,The English source phrase.,How are you? (Morning)
st,The target phrase in the respective Botswana language.,O tsogile jang?
notes,Context or literal translation (Optional).,"Lit. ""How did you wake up?""" | 3 | 0 | [
"size_categories:n<1K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T14:08:45+00:00 | 2025-11-11T14:11:33+00:00 | 0 |
svjack/Xiang_qwen_image_2509_head_swap |
- head refer image
- swap grid image
|
- head refer image
- swap grid image
| 152 | 1 | [
"size_categories:n<1K",
"format:parquet",
"modality:image",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-10-12T10:04:39+00:00 | 2025-11-11T14:08:12+00:00 | 0 |
phospho-app/so100-tictactoe_bboxes |
# so100-tictactoe
**This dataset was generated using a [phospho starter pack](https://robots.phospho.ai).**
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot and RLDS.
|
# so100-tictactoe
**This dataset was generated using a [phospho starter pack](https://robots.phospho.ai).**
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot and RLDS.
| 71 | 0 | [
"task_categories:robotics",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"phosphobot",
"so100",
"phospho-dk"
] | 2025-06-14T05:24:09+00:00 | 2025-11-11T14:07:56+00:00 | 0 |
birdsql/livesqlbench-base-lite-sqlite |
# 🚀 LiveSQLBench-Base-Lite
*A dynamic, **contamination‑free** benchmark for evaluating LLMs on complex, real‑world ****text‑to‑SQL**** tasks.*
[🌐 LiveSQLBench Website](https://livesqlbench.ai) • [🌐 BIRD-INTERACT Project Page](https://bird-interact.github.io/) • [📄 Paper](https://huggingface.co/papers/2510.05318) • [💻 LiveSQLBench GitHub](https://github.com/bird-bench/livesqlbench) • [💻 BIRD-INTERACT GitHub](https://github.com/bird-bench/BIRD-Interact)
Maintained by the **🦜 [BIRD Team @ HKU](https://bird-bench.github.io)** & **☁️ [Google Cloud](https://cloud.google.com/)**
## 📊 LiveSQLBench Overview
**LiveSQLBench** (BIRD-SQL Pro v0.5) is a **contamination-free**, **continuously evolving** benchmark designed to evaluate LLMs on **complex, real-world text-to-SQL tasks**, featuring **diverse real-world user queries**, including **Business Intelligence (BI)**, **CRUD operations**, and more. Each release will include **50 new, fully open-source DBs** curated by the BIRD team through expert collaboration and continuous improvement. It will cover a **wide range of database sizes**, from **end-user level** (around 127 columns) to **industrial level** (1340+ columns). Here are the features of the LiveSQLBench benchmark:
1. **🗄️ Live Databases:**
Constructed dynamically from extensive and regularly updated CSV datasets, with both base (user-end level) and large (industrial level) versions (1340+ columns each DB) to test scalability.
2. **💬 Live User Queries and SQL:**
Each task pairs unambiguous user queries with annotated, gold-standard SQL statements. The user queries are grounded in an external knowledge base, with medium to hard complexity solution SQL statements.
3. **🧠 Contextual Reasoning (HKB):**
Every DB includes a hierarchical knowledge base (HKB) where each knowledge may have dependencies to others, which requires the multi-hop reasoning ability. Two HKB formats are provided: (1) structured JSON format, and (2) unstructured Document format.
4. **🔍 The First Full SQL Spectrum:**
Supports not just SELECT (Business Intelligence) queries, but also CRUD (e.g., UPDATE, CREATE, and other database management operations) queries.
5. **⚡ Automated Evaluation:**
Support fast evaluation via PostgreSQL template & docker. Each question includes verifiable test cases for accurate, reproducible scoring. Soft EX metric is used to evaluate SELECT-ONLY tasks; customized test cases are designed for DBA tasks, such as CRUD (CREATE, READ, UPDATE, DELETE).
6. **🔄 Truly Live & Hidden Test:**
New databases and tasks are added over time. Each release features both open development and hidden test phases. The hidden test set from each release becomes the open development set for the next release, ensuring continuous evolution and fair evaluation.
> 💡 LiveSQLBench's updating databases, tasks, and HKB support BIRD-Interact's conversational and agentic evaluation. BIRD-Interact evaluates LLMs' text-to-SQL ability in dynamic interactive settings with database and user simulation.
## 🎯 Current Release: LiveSQLBench-Base-Lite-SQLite
We are pleased to release a **SQLite version** of **LiveSQLBench-Base-Lite**, extending from PostgreSQL to SQLite dialect to **improve accessibility** as SQLite requires no server setup and runs locally. This release features **18 end-user level databases** with **270** tasks (180 SELECT-only, 90 Management tasks), **HKB-JSON** and **JSON operations in SQL** for trial.
> Beyond SQL and test case translation, we **carefully adapted 20+ user queries** to align with SQLite's database engine characteristics. For example, since SQLite doesn't support custom functions, we modified queries to either return specific scenario values or utilize views (e.g., `CREATE VIEW AS ...`) to maintain query complexity while ensuring compatibility.
## 💻 How to Use the Dataset
Download the dataset containing data file `livesqlbench_data_sqlite.jsonl` and DB metafiles (including schema, HKB, column meaning files) by:
```bash
huggingface-cli download --repo-type dataset --resume-download birdsql/livesqlbench-base-lite-sqlite --local-dir /local/path/livesqlbench-base-lite-sqlite
```
To prevent data leakage through automated crawling, please request access to the ground truth and test cases by emailing **[📧 bird.bench25@gmail.com](mailto:bird.bench25@gmail.com)** with the subject line `[livesqlbench-base-lite GT&Test Cases]`. An automated response will provide these data fields.
And please refer to the BIRD-MiniDev [Github repo](https://github.com/bird-bench/mini_dev) for details of usage and evaluation based on this dataset.
## Sample Usage
You can load the dataset using the Hugging Face `datasets` library:
```python
from datasets import load_dataset
# Load the LiveSQLBench-Base-Lite-SQLite dataset
dataset = load_dataset("birdsql/livesqlbench-base-lite-sqlite", "livesqlbench")
# Access the development split
dev_data = dataset["dev"]
# Print the first example
print(dev_data[0])
```
## 📊 Performance on LiveSQLBench-Base-Lite
| Model | PostgreSQL | SQlite |
| :-------------------- | :--------- | :----- |
| o3-mini | 47.78 | 42.59 |
| Claude 3.7 Sonnet | 39.26 | 41.11 |
| GPT-4o | 34.44 | 34.44 |
| Gemini 2.0 Flash | 34.44 | 33.7 |
| DeepSeek R1-0528 | 38.14 | 32.96 |
| QwQ-32B | 31.48 | 31.48 |
| Qwen2.5 Coder 32B | 22.96 | 22.22 |
| Codestral 22B | 21.11 | 19.63 |
| Qwen2.5 Coder 7B | 12.22 | 12.22 |
| Mixtral 8x7B Instruct | 2.59 | 8.89 |
| Mistral 7B Instruct | 3.7 | 4.44 |
## 📁 Directory Structure
Each database has its own directory:
```
.
├── README.md
├── alien
│ ├── alien_column_meaning_base.json
│ ├── alien_kb.jsonl
│ ├── alien_schema.txt
│ ├── alien_tempalte.sqlite
...
├── livesqlbench_data_sqlite.jsonl
```
### 📂 Directory Contents:
* `*_schema.txt`: Database schema.
* `*_kb.jsonl`: Hierarchical knowledge base entries required to solve the user task.
* `id`: The unique identifier for the knowledge.
* `knowledge`: The name of the knowledge.
* `description`: The description of the knowledge.
* `definition`: The clear definition of the knowledge.
* `type`: The type of the knowledge.
* `children_knowledge`: A list of knowledge IDs that the current knowledge is dependent on. -1 means no children.
* `*_column_meaning_base.json`: Explanation of database columns.
## 📋 Dataset Fields (`livesqlbench_data_sqlite.jsonl`):
* **instance\_id**: Unique task identifier.
* **selected\_database**: Associated database name.
* **query**: Ambiguous user query.
* **sol\_sql** 🔒: Ground truth SQL solution.
* **external\_knowledge** 🔒: IDs of required external knowledge to solve the user task.
* **preprocess\_sql**: SQL setup queries.
* **clean\_up\_sql**: SQL queries to reset database state.
* **test\_cases** 🔒: Test cases to validate the predicted corrected SQL.
* **category**: "Query" (SELECT-only) or "Management" (CRUD).
* **high\_level**: Boolean indicating whether the user query contains high-level description.
* **conditions**: Indicates decimal/distinct conditions in the user query.
* **difficulty\_tier**: Task difficulty (Simple, Moderate, Challenging).
## 🔒 Accessing Complete Data
To avoid data leakage by auto-crawling, certain fields (e.g., `sol_sql`, `test_cases`, `external_knowledge`) are excluded from the public dataset. For the full dataset, please email: **[📧 bird.bench25@gmail.com](mailto:bird.bench25@gmail.com)** with subject tag `[livesqlbench-base-lite-SQLite GT&Test Cases]`, which will be sent automatically.
## 🔄 Stay Tuned!
Upcoming releases:
* **🔄 LiveSQLBench-Base-Full:** 600 BI tasks, 200 management tasks, Document-based HKB.
* **🔄 LiveSQLBench-Large-Lite:** Industrial-scale databases with 1340+ columns.
* **🔄 LiveSQLBench-Large-Full:** Comprehensive large-scale datasets.
Want new dialects? Vote for new SQL dialects [🗳️ here](https://docs.google.com/forms/d/e/1FAIpQLSfEogmsA7LObI13KOoiojdnYfW28KEqvEVtC9hXaZJ8O9aCpQ/viewform?usp=header)!
## 📄 License:
cc-by-sa-4.0 |
# 🚀 LiveSQLBench-Base-Lite
*A dynamic, **contamination‑free** benchmark for evaluating LLMs on complex, real‑world ****text‑to‑SQL**** tasks.*
[🌐 LiveSQLBench Website](https://livesqlbench.ai) • [🌐 BIRD-INTERACT Project Page](https://bird-interact.github.io/) • [📄 Paper](https://huggingface.co/papers/2510.05318) • [💻 LiveSQLBench GitHub](https://github.com/bird-bench/livesqlbench) • [💻 BIRD-INTERACT GitHub](https://github.com/bird-bench/BIRD-Interact)
Maintained by the **🦜 [BIRD Team @ HKU](https://bird-bench.github.io)** & **☁️ [Google Cloud](https://cloud.google.com/)**
## 📊 LiveSQLBench Overview
**LiveSQLBench** (BIRD-SQL Pro v0.5) is a **contamination-free**, **continuously evolving** benchmark designed to evaluate LLMs on **complex, real-world text-to-SQL tasks**, featuring **diverse real-world user queries**, including **Business Intelligence (BI)**, **CRUD operations**, and more. Each release will include **50 new, fully open-source DBs** curated by the BIRD team through expert collaboration and continuous improvement. It will cover a **wide range of database sizes**, from **end-user level** (around 127 columns) to **industrial level** (1340+ columns). Here are the features of the LiveSQLBench benchmark:
1. **🗄️ Live Databases:**
Constructed dynamically from extensive and regularly updated CSV datasets, with both base (user-end level) and large (industrial level) versions (1340+ columns each DB) to test scalability.
2. **💬 Live User Queries and SQL:**
Each task pairs unambiguous user queries with annotated, gold-standard SQL statements. The user queries are grounded in an external knowledge base, with medium to hard complexity solution SQL statements.
3. **🧠 Contextual Reasoning (HKB):**
Every DB includes a hierarchical knowledge base (HKB) where each knowledge may have dependencies to others, which requires the multi-hop reasoning ability. Two HKB formats are provided: (1) structured JSON format, and (2) unstructured Document format.
4. **🔍 The First Full SQL Spectrum:**
Supports not just SELECT (Business Intelligence) queries, but also CRUD (e.g., UPDATE, CREATE, and other database management operations) queries.
5. **⚡ Automated Evaluation:**
Support fast evaluation via PostgreSQL template & docker. Each question includes verifiable test cases for accurate, reproducible scoring. Soft EX metric is used to evaluate SELECT-ONLY tasks; customized test cases are designed for DBA tasks, such as CRUD (CREATE, READ, UPDATE, DELETE).
6. **🔄 Truly Live & Hidden Test:**
New databases and tasks are added over time. Each release features both open development and hidden test phases. The hidden test set from each release becomes the open development set for the next release, ensuring continuous evolution and fair evaluation.
> 💡 LiveSQLBench's updating databases, tasks, and HKB support BIRD-Interact's conversational and agentic evaluation. BIRD-Interact evaluates LLMs' text-to-SQL ability in dynamic interactive settings with database and user simulation.
## 🎯 Current Release: LiveSQLBench-Base-Lite-SQLite
We are pleased to release a **SQLite version** of **LiveSQLBench-Base-Lite**, extending from PostgreSQL to SQLite dialect to **improve accessibility** as SQLite requires no server setup and runs locally. This release features **18 end-user level databases** with **270** tasks (180 SELECT-only, 90 Management tasks), **HKB-JSON** and **JSON operations in SQL** for trial.
> Beyond SQL and test case translation, we **carefully adapted 20+ user queries** to align with SQLite's database engine characteristics. For example, since SQLite doesn't support custom functions, we modified queries to either return specific scenario values or utilize views (e.g., `CREATE VIEW AS ...`) to maintain query complexity while ensuring compatibility.
## 💻 How to Use the Dataset
Download the dataset containing data file `livesqlbench_data_sqlite.jsonl` and DB metafiles (including schema, HKB, column meaning files) by:
```bash
huggingface-cli download --repo-type dataset --resume-download birdsql/livesqlbench-base-lite-sqlite --local-dir /local/path/livesqlbench-base-lite-sqlite
```
To prevent data leakage through automated crawling, please request access to the ground truth and test cases by emailing **[📧 bird.bench25@gmail.com](mailto:bird.bench25@gmail.com)** with the subject line `[livesqlbench-base-lite GT&Test Cases]`. An automated response will provide these data fields.
And please refer to the BIRD-MiniDev [Github repo](https://github.com/bird-bench/mini_dev) for details of usage and evaluation based on this dataset.
## Sample Usage
You can load the dataset using the Hugging Face `datasets` library:
```python
from datasets import load_dataset
# Load the LiveSQLBench-Base-Lite-SQLite dataset
dataset = load_dataset("birdsql/livesqlbench-base-lite-sqlite", "livesqlbench")
# Access the development split
dev_data = dataset["dev"]
# Print the first example
print(dev_data[0])
```
## 📊 Performance on LiveSQLBench-Base-Lite
| Model | PostgreSQL | SQlite |
| :-------------------- | :--------- | :----- |
| o3-mini | 47.78 | 42.59 |
| Claude 3.7 Sonnet | 39.26 | 41.11 |
| GPT-4o | 34.44 | 34.44 |
| Gemini 2.0 Flash | 34.44 | 33.7 |
| DeepSeek R1-0528 | 38.14 | 32.96 |
| QwQ-32B | 31.48 | 31.48 |
| Qwen2.5 Coder 32B | 22.96 | 22.22 |
| Codestral 22B | 21.11 | 19.63 |
| Qwen2.5 Coder 7B | 12.22 | 12.22 |
| Mixtral 8x7B Instruct | 2.59 | 8.89 |
| Mistral 7B Instruct | 3.7 | 4.44 |
## 📁 Directory Structure
Each database has its own directory:
```
.
├── README.md
├── alien
│ ├── alien_column_meaning_base.json
│ ├── alien_kb.jsonl
│ ├── alien_schema.txt
│ ├── alien_tempalte.sqlite
...
├── livesqlbench_data_sqlite.jsonl
```
### 📂 Directory Contents:
* `*_schema.txt`: Database schema.
* `*_kb.jsonl`: Hierarchical knowledge base entries required to solve the user task.
* `id`: The unique identifier for the knowledge.
* `knowledge`: The name of the knowledge.
* `description`: The description of the knowledge.
* `definition`: The clear definition of the knowledge.
* `type`: The type of the knowledge.
* `children_knowledge`: A list of knowledge IDs that the current knowledge is dependent on. -1 means no children.
* `*_column_meaning_base.json`: Explanation of database columns.
## 📋 Dataset Fields (`livesqlbench_data_sqlite.jsonl`):
* **instance\_id**: Unique task identifier.
* **selected\_database**: Associated database name.
* **query**: Ambiguous user query.
* **sol\_sql** 🔒: Ground truth SQL solution.
* **external\_knowledge** 🔒: IDs of required external knowledge to solve the user task.
* **preprocess\_sql**: SQL setup queries.
* **clean\_up\_sql**: SQL queries to reset database state.
* **test\_cases** 🔒: Test cases to validate the predicted corrected SQL.
* **category**: "Query" (SELECT-only) or "Management" (CRUD).
* **high\_level**: Boolean indicating whether the user query contains high-level description.
* **conditions**: Indicates decimal/distinct conditions in the user query.
* **difficulty\_tier**: Task difficulty (Simple, Moderate, Challenging).
## 🔒 Accessing Complete Data
To avoid data leakage by auto-crawling, certain fields (e.g., `sol_sql`, `test_cases`, `external_knowledge`) are excluded from the public dataset. For the full dataset, please email: **[📧 bird.bench25@gmail.com](mailto:bird.bench25@gmail.com)** with subject tag `[livesqlbench-base-lite-SQLite GT&Test Cases]`, which will be sent automatically.
## 🔄 Stay Tuned!
Upcoming releases:
* **🔄 LiveSQLBench-Base-Full:** 600 BI tasks, 200 management tasks, Document-based HKB.
* **🔄 LiveSQLBench-Large-Lite:** Industrial-scale databases with 1340+ columns.
* **🔄 LiveSQLBench-Large-Full:** Comprehensive large-scale datasets.
Want new dialects? Vote for new SQL dialects [🗳️ here](https://docs.google.com/forms/d/e/1FAIpQLSfEogmsA7LObI13KOoiojdnYfW28KEqvEVtC9hXaZJ8O9aCpQ/viewform?usp=header)!
## 📄 License:
cc-by-sa-4.0 | 595 | 2 | [
"task_categories:table-question-answering",
"license:cc-by-sa-4.0",
"size_categories:n<1K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2510.05318",
"region:us",
"text-to-sql",
"database",
"multi-turn",
"interactive"
] | 2025-07-18T11:16:00+00:00 | 2025-11-11T14:03:37+00:00 | 0 |
fracapuano/behavior1k |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "R1Pro",
"total_episodes": 2000,
"total_frames": 21227314,
"total_tasks": 10,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:2000"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.rgb.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.rgb.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.rgb.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.depth.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.depth.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.depth.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.seg_instance_id.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.seg_instance_id.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.seg_instance_id.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"action": {
"dtype": "float32",
"shape": [
23
],
"names": null,
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"observation.cam_rel_poses": {
"dtype": "float32",
"shape": [
21
],
"names": null,
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
256
],
"names": null,
"fps": 30
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "R1Pro",
"total_episodes": 2000,
"total_frames": 21227314,
"total_tasks": 10,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:2000"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"observation.images.rgb.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.rgb.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.rgb.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.depth.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.depth.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.depth.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"depth"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p16le",
"video.is_depth_map": true,
"has_audio": false
}
},
"observation.images.seg_instance_id.left_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.seg_instance_id.right_wrist": {
"dtype": "video",
"shape": [
480,
480,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 480,
"video.width": 480,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"observation.images.seg_instance_id.head": {
"dtype": "video",
"shape": [
720,
720,
3
],
"names": [
"height",
"width",
"rgb"
],
"info": {
"video.fps": 30.0,
"video.height": 720,
"video.width": 720,
"video.channels": 3,
"video.codec": "libx265",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"has_audio": false
}
},
"action": {
"dtype": "float32",
"shape": [
23
],
"names": null,
"fps": 30
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"observation.cam_rel_poses": {
"dtype": "float32",
"shape": [
21
],
"names": null,
"fps": 30
},
"observation.state": {
"dtype": "float32",
"shape": [
256
],
"names": null,
"fps": 30
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 20 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10M<n<100M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:03:29+00:00 | 2025-11-11T14:02:17+00:00 | 0 |
iAyoD/robocasa_close_drawer_256 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 10035,
"total_tasks": 2,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "panda",
"total_episodes": 53,
"total_frames": 10035,
"total_tasks": 2,
"total_videos": 0,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 20,
"splits": {
"train": "0:53"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"wrist_image": {
"dtype": "image",
"shape": [
256,
256,
3
],
"names": [
"height",
"width",
"channel"
]
},
"state": {
"dtype": "float32",
"shape": [
16
],
"names": [
"state"
]
},
"actions": {
"dtype": "float32",
"shape": [
12
],
"names": [
"actions"
]
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 22 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:image",
"modality:timeseries",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot",
"robocasa",
"panda"
] | 2025-11-11T14:01:38+00:00 | 2025-11-11T14:02:06+00:00 | 0 |
SALT-NLP/RealtimeGym |
# Real-Time Reasoning
Real-time reasoning traces for agents in evolving environments. This dataset accompanies the Real-Time Reasoning project page and demos. It provides step-by-step agent states, actions, scores, and (optionally) "thinking" content for three environments: Freeway, Snake, and Overcooked, under varying cognitive loads and time-pressure budgets, across multiple seeds and agent paradigms.
- Project page: https://realtimegym.saltlab.stanford.edu/
- Dataset repo: https://huggingface.co/datasets/SALT-NLP/RealtimeGym
- Paper: coming soon
- Code (gym): https://github.com/SALT-NLP/RealtimeGym
## Contents
Each file is a JSON list of steps for a single run:
- Game: {freeway, snake, overcooked}
- Cognitive load: {easy, medium, hard}
- Time pressure budget: {4k, 8k, 16k, 32k}
- Seed: {seed0 … seed7}
- Agent paradigm: {reactive, planning, agile}
Filenames follow:
- {game}_{load}_{budget}_{seed}_{agent}.json
e.g., `freeway_easy_4k_seed0_planning.json`
## JSON schema (per step)
Each file is an array of objects like:
- step: integer step index
- score: numeric current score
- action: string action taken at this step (e.g., "U", "D", "L", "R", "Keep", depending on the game)
- thinking: string with model/agent "reasoning" text (when available)
- state: object capturing current environment state; fields vary by game. For Freeway, for example:
- pos: integer/tuple encoding player position (implementation-specific)
- game_turn: integer turn counter
- terminal: boolean whether episode has ended
- cars: list of car tuples [head_position, lane_id, direction_or_delta, speed]
- original_state: string with original input state text for agent to reason about
Example (truncated from Freeway planning run):
```json
[
{
"step": 0,
"score": 0,
"thinking": "Still thinking...",
"state": {
"pos": 0,
"game_turn": 0,
"terminal": false,
"cars": [[48,1,12,12], [0,1,12,12]]
},
"action": "U"
}
]
```
Notes:
- Snake and Overcooked files follow the same top-level keys; their `state` inner structure differs according to the game.
- "thinking" may contain markdown or math formatting.
## Splits
There are no predefined train/validation/test splits. Users can split by:
- game ∈ {freeway, snake, overcooked}
- cognitive_load ∈ {easy, medium, hard}
- time_pressure ∈ {4k, 8k, 16k, 32k}
- seed ∈ {seed0 … seed7}
- agent ∈ {reactive, planning, agile}
## Loading
**Note:** Due to different state structures across games (Freeway, Snake, Overcooked), it is recommended to load the data directly as JSON files rather than using Hugging Face's `load_dataset` function, which expects a uniform schema.
Python (direct JSON loading - recommended):
```python
import json
from huggingface_hub import hf_hub_download
# Download a specific file
file_path = hf_hub_download(
repo_id="BLeaves/real-time-reasoning",
filename="freeway_easy_4k_seed0_planning.json",
repo_type="dataset"
)
# Load the JSON file
with open(file_path, "r") as f:
episode = json.load(f)
print(episode[0].keys()) # ['step', 'score', 'thinking', 'state', 'action']
print(episode[0]['state'].keys()) # Game-specific state fields
```
Or download all files:
```python
from huggingface_hub import snapshot_download
import json
import glob
# Download entire dataset
local_dir = snapshot_download(
repo_id="BLeaves/real-time-reasoning",
repo_type="dataset"
)
# Load JSON files
json_files = glob.glob(f"{local_dir}/*.json")
for file_path in json_files:
with open(file_path, "r") as f:
episode = json.load(f)
# Process episode...
```
## Tasks
- Stepwise reasoning analysis
- Agent behavior evaluation across cognitive load/time pressure
- Comparative studies: reactive vs planning vs AgileThinker
- Visualization and replay
## Citation
If you use this dataset, please cite the project:
```bibtex
@misc{wen2025realtimereasoningagentsevolving,
title={Real-Time Reasoning Agents in Evolving Environments},
author={Yule Wen and Yixin Ye and Yanzhe Zhang and Diyi Yang and Hao Zhu},
year={2025},
eprint={2511.04898},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2511.04898},
}
```
## License
MIT
|
# Real-Time Reasoning
Real-time reasoning traces for agents in evolving environments. This dataset accompanies the Real-Time Reasoning project page and demos. It provides step-by-step agent states, actions, scores, and (optionally) "thinking" content for three environments: Freeway, Snake, and Overcooked, under varying cognitive loads and time-pressure budgets, across multiple seeds and agent paradigms.
- Project page: https://realtimegym.saltlab.stanford.edu/
- Dataset repo: https://huggingface.co/datasets/SALT-NLP/RealtimeGym
- Paper: coming soon
- Code (gym): https://github.com/SALT-NLP/RealtimeGym
## Contents
Each file is a JSON list of steps for a single run:
- Game: {freeway, snake, overcooked}
- Cognitive load: {easy, medium, hard}
- Time pressure budget: {4k, 8k, 16k, 32k}
- Seed: {seed0 … seed7}
- Agent paradigm: {reactive, planning, agile}
Filenames follow:
- {game}_{load}_{budget}_{seed}_{agent}.json
e.g., `freeway_easy_4k_seed0_planning.json`
## JSON schema (per step)
Each file is an array of objects like:
- step: integer step index
- score: numeric current score
- action: string action taken at this step (e.g., "U", "D", "L", "R", "Keep", depending on the game)
- thinking: string with model/agent "reasoning" text (when available)
- state: object capturing current environment state; fields vary by game. For Freeway, for example:
- pos: integer/tuple encoding player position (implementation-specific)
- game_turn: integer turn counter
- terminal: boolean whether episode has ended
- cars: list of car tuples [head_position, lane_id, direction_or_delta, speed]
- original_state: string with original input state text for agent to reason about
Example (truncated from Freeway planning run):
```json
[
{
"step": 0,
"score": 0,
"thinking": "Still thinking...",
"state": {
"pos": 0,
"game_turn": 0,
"terminal": false,
"cars": [[48,1,12,12], [0,1,12,12]]
},
"action": "U"
}
]
```
Notes:
- Snake and Overcooked files follow the same top-level keys; their `state` inner structure differs according to the game.
- "thinking" may contain markdown or math formatting.
## Splits
There are no predefined train/validation/test splits. Users can split by:
- game ∈ {freeway, snake, overcooked}
- cognitive_load ∈ {easy, medium, hard}
- time_pressure ∈ {4k, 8k, 16k, 32k}
- seed ∈ {seed0 … seed7}
- agent ∈ {reactive, planning, agile}
## Loading
**Note:** Due to different state structures across games (Freeway, Snake, Overcooked), it is recommended to load the data directly as JSON files rather than using Hugging Face's `load_dataset` function, which expects a uniform schema.
Python (direct JSON loading - recommended):
```python
import json
from huggingface_hub import hf_hub_download
# Download a specific file
file_path = hf_hub_download(
repo_id="BLeaves/real-time-reasoning",
filename="freeway_easy_4k_seed0_planning.json",
repo_type="dataset"
)
# Load the JSON file
with open(file_path, "r") as f:
episode = json.load(f)
print(episode[0].keys()) # ['step', 'score', 'thinking', 'state', 'action']
print(episode[0]['state'].keys()) # Game-specific state fields
```
Or download all files:
```python
from huggingface_hub import snapshot_download
import json
import glob
# Download entire dataset
local_dir = snapshot_download(
repo_id="BLeaves/real-time-reasoning",
repo_type="dataset"
)
# Load JSON files
json_files = glob.glob(f"{local_dir}/*.json")
for file_path in json_files:
with open(file_path, "r") as f:
episode = json.load(f)
# Process episode...
```
## Tasks
- Stepwise reasoning analysis
- Agent behavior evaluation across cognitive load/time pressure
- Comparative studies: reactive vs planning vs AgileThinker
- Visualization and replay
## Citation
If you use this dataset, please cite the project:
```bibtex
@misc{wen2025realtimereasoningagentsevolving,
title={Real-Time Reasoning Agents in Evolving Environments},
author={Yule Wen and Yixin Ye and Yanzhe Zhang and Diyi Yang and Hao Zhu},
year={2025},
eprint={2511.04898},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2511.04898},
}
```
## License
MIT
| 199 | 0 | [
"task_categories:question-answering",
"task_categories:other",
"task_ids:dialogue-generation",
"task_ids:task-planning",
"language:en",
"license:mit",
"arxiv:2511.04898",
"region:us"
] | 2025-11-05T01:13:00+00:00 | 2025-11-11T14:02:48+00:00 | 0 |
orybe/close-sweet-mix |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 30,
"total_frames": 41635,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:30"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "h264",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "h264",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 30,
"total_frames": 41635,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:30"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "h264",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "h264",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 16 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T14:04:46+00:00 | 2025-11-11T14:05:20+00:00 | 0 |
RogersPyke/realman_rmc_aidal_plate_storage |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
grip the plate with the left hand, pass it to the right, then set it on the shelf.
grip the plate with the right hand, pass it to the left, then set it on the shelf.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 500, "total_frames": 287069, "total_tasks": 2, "total_videos": 1500, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:500"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
grip the plate with the left hand, pass it to the right, then set it on the shelf.
grip the plate with the right hand, pass it to the left, then set it on the shelf.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 500, "total_frames": 287069, "total_tasks": 2, "total_videos": 1500, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:500"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 133 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-11T13:52:18+00:00 | 2025-11-11T13:58:21+00:00 | 0 |
Septzzz/MMR-Life | # MMR-Life (Multimodal Multi-image Reasoning Benchmark under Real-life Scenarios)
### Dataset Description
We introduce MMR-Life, a novel benchmark meticulously curated to evaluate the ability of MLLMs to perform diverse types of reasoning in everyday situations. MMR-Life consists of **2,676 multiple-choice questions based on 19,367 images**, covering **7 reasoning types** (i.e., abductive, analogical, causal, deductive, inductive, spatial, and temporal) and 21 tasks. Each task is based on a set of **multi-images**, predominantly sourced from **real-life contexts**, such as domestic life, daily dining, and sports activities.mmunity to build next-generation multimodal foundation models towards expert artificial general intelligence (AGI).
### Dataset Examples
Examples of different inference types in our dataset:
### Dataset Usage
#### Data Downloading
All the data examples were divided into two subsets: *testmini* and *test*.
- **test_mini**: 210 examples used for model development, validation, or for those with limited computing resources.
- **test**: 2,676 examples for standard evaluation.
You can download this dataset by the following command:
```python
from datasets import load_dataset
dataset = load_dataset("Septzzz/MMR-Life")
```
Here are some examples of how to access the downloaded dataset:
```python
# print the first example on the testmini set
print(dataset["testmini"][0])
print(dataset["testmini"][0]['id']) # print the problem id
print(dataset["testmini"][0]['question']) # print the question text
print(dataset["testmini"][0]['query']) # print the query text
print(dataset["testmini"][0]['image_path']) # print the image path
print(dataset["testmini"][0]['golden_answer']) # print the golden answer
dataset["testmini"][0]['image1'] # display the image
# print the first example on the test set
print(dataset["test"][0])
```
#### Data Format
The dataset is provided in json format and contains the following attributes:
```json
{
"question": [string] The question text,
"image": [string] A file path pointing to the associated image,
"choices": [list] Choice options for multiple-choice problems. For free-form problems, this could be a 'none' value,
"precision": [integer] The number of decimal places the answer should be rounded to,
"answer": [string] The correct answer for the problem,
"question_type": [string] The type of question: "multi_choice" or "free_form",
"pid": [string] Problem ID, e.g., "1",
"metadata": {
"split": [string] Data split: "testmini" or "test",
"language": [string] Question language: "English", "Chinese", or "Persian",
"img_width": [integer] The width of the associated image in pixels,
"img_height": [integer] The height of the associated image in pixels,
"source": [string] The source dataset from which the problem was taken,
"category": [string] The category of the problem: "math-targeted-vqa" or "general-vqa",
"task": [string] The task of the problem, e.g., "geometry problem solving",
"context": [string] The visual context type of the associated image,
"grade": [string] The grade level of the problem, e.g., "high school",
"skills": [list] A list of mathematical reasoning skills that the problem tests
},
"query": [string] the query text used as input (prompt) for the evaluation model
}
```
### Mini-Leaderboard
We show a mini-leaderboard here and please find more information in our paper.
| Model | Abd | Ana | Cau | Ded | Ind | Spa | Tem | Avg |
|:------|----:|----:|----:|----:|----:|----:|----:|----:|
| Human* | 79.76 | 57.65 | 75.00 | 70.59 | 63.41 | 79.76 | 79.76 | 72.28 |
| GPT-5 | 53.57 | 78.37 | 41.06 | 79.86 | 77.25 | 17.25 | 41.47 | 58.48 |
| Gemini-2.5-Pro | 54.22 | 73.36 | 36.99 | 79.15 | 72.30 | 25.10 | 35.60 | 56.58 |
| Gemini-2.5-Flash | 46.10 | 74.57 | 34.22 | 71.38 | 73.42 | 23.92 | 30.64 | 53.03 |
| o4-mini | 41.23 | 73.01 | 27.38 | 71.02 | 67.12 | 19.22 | 32.48 | 50.30 |
| GPT-5-mini | 44.81 | 69.55 | 32.32 | 74.91 | 68.02 | 12.16 | 29.36 | 49.70 |
| GPT-4.1 | 44.16 | 71.11 | 22.43 | 67.14 | 69.37 | 13.73 | 27.16 | 48.09 |
| Claude-Sonnet-4 | 36.84 | 60.55 | 44.11 | 66.78 | 55.63 | 15.69 | 28.07 | 45.11 |
| Claude-3.7-Sonnet | 33.44 | 66.09 | 35.36 | 59.72 | 59.01 | 20.78 | 25.87 | 44.96 |
| GPT-4o | 46.75 | 65.22 | 25.86 | 51.24 | 65.32 | 11.37 | 25.87 | 44.62 |
| GPT-4.1-mini | 32.79 | 60.90 | 30.80 | 51.94 | 64.64 | 16.47 | 30.46 | 43.95 |
| Qwen2.5-VL-72B | 35.06 | 55.02 | 35.36 | 51.94 | 54.73 | 12.94 | 23.67 | 40.02 |
| Doubao-1.5-vision | 37.01 | 53.29 | 31.18 | 59.36 | 54.50 | 12.16 | 22.94 | 39.99 |
| VL-Rethinker-72B | 36.36 | 50.52 | 33.84 | 55.83 | 57.88 | 15.29 | 21.65 | 39.80 |
| Gemma3-27B | 35.71 | 57.79 | 36.88 | 31.80 | 60.81 | 13.33 | 18.72 | 38.75 |
| MM-Eureka-Qwen-32B | 23.70 | 42.56 | 25.48 | 49.12 | 28.83 | 16.86 | 17.98 | 29.67 |
| Gemma3-12B | 24.35 | 51.21 | 15.97 | 28.27 | 43.47 | 10.59 | 16.15 | 29.93 |
| MiMo-VL-7B-RL | 38.31 | 26.47 | 28.14 | 62.90 | 25.23 | 13.33 | 20.73 | 29.22 |
| Qwen2.5-VL-32B | 24.35 | 42.73 | 21.67 | 50.18 | 26.58 | 14.90 | 16.51 | 28.66 |
| VL-Rethinker-7B | 30.84 | 40.48 | 21.29 | 28.62 | 43.02 | 13.73 | 11.93 | 28.29 |
| Qwen2.5-VL-7B | 25.97 | 35.64 | 21.29 | 22.26 | 40.32 | 9.02 | 12.48 | 25.22 |
| InternVL3.5-30B-A3B | 48.05 | 18.17 | 33.08 | 37.46 | 13.29 | 13.33 | 13.39 | 22.87 |
| Keye-VL-1.5-8B | 19.48 | 21.63 | 23.19 | 13.78 | 19.59 | 13.73 | 23.30 | 19.96 |
| InternVL3.5-8B | 35.71 | 9.86 | 19.01 | 32.16 | 10.14 | 13.33 | 17.43 | 18.01 |
| Skywork-R1V-38B | 24.03 | 9.52 | 16.35 | 24.03 | 11.04 | 9.80 | 10.28 | 13.83 |
## Contact
Jiachun Li: jiachun.li@nlpr.ia.ac.cn
## Citation
``` | # MMR-Life (Multimodal Multi-image Reasoning Benchmark under Real-life Scenarios)
### Dataset Description
We introduce MMR-Life, a novel benchmark meticulously curated to evaluate the ability of MLLMs to perform diverse types of reasoning in everyday situations. MMR-Life consists of **2,676 multiple-choice questions based on 19,367 images**, covering **7 reasoning types** (i.e., abductive, analogical, causal, deductive, inductive, spatial, and temporal) and 21 tasks. Each task is based on a set of **multi-images**, predominantly sourced from **real-life contexts**, such as domestic life, daily dining, and sports activities.mmunity to build next-generation multimodal foundation models towards expert artificial general intelligence (AGI).
### Dataset Examples
Examples of different inference types in our dataset:
### Dataset Usage
#### Data Downloading
All the data examples were divided into two subsets: *testmini* and *test*.
- **test_mini**: 210 examples used for model development, validation, or for those with limited computing resources.
- **test**: 2,676 examples for standard evaluation.
You can download this dataset by the following command:
```python
from datasets import load_dataset
dataset = load_dataset("Septzzz/MMR-Life")
```
Here are some examples of how to access the downloaded dataset:
```python
# print the first example on the testmini set
print(dataset["testmini"][0])
print(dataset["testmini"][0]['id']) # print the problem id
print(dataset["testmini"][0]['question']) # print the question text
print(dataset["testmini"][0]['query']) # print the query text
print(dataset["testmini"][0]['image_path']) # print the image path
print(dataset["testmini"][0]['golden_answer']) # print the golden answer
dataset["testmini"][0]['image1'] # display the image
# print the first example on the test set
print(dataset["test"][0])
```
#### Data Format
The dataset is provided in json format and contains the following attributes:
```json
{
"question": [string] The question text,
"image": [string] A file path pointing to the associated image,
"choices": [list] Choice options for multiple-choice problems. For free-form problems, this could be a 'none' value,
"precision": [integer] The number of decimal places the answer should be rounded to,
"answer": [string] The correct answer for the problem,
"question_type": [string] The type of question: "multi_choice" or "free_form",
"pid": [string] Problem ID, e.g., "1",
"metadata": {
"split": [string] Data split: "testmini" or "test",
"language": [string] Question language: "English", "Chinese", or "Persian",
"img_width": [integer] The width of the associated image in pixels,
"img_height": [integer] The height of the associated image in pixels,
"source": [string] The source dataset from which the problem was taken,
"category": [string] The category of the problem: "math-targeted-vqa" or "general-vqa",
"task": [string] The task of the problem, e.g., "geometry problem solving",
"context": [string] The visual context type of the associated image,
"grade": [string] The grade level of the problem, e.g., "high school",
"skills": [list] A list of mathematical reasoning skills that the problem tests
},
"query": [string] the query text used as input (prompt) for the evaluation model
}
```
### Mini-Leaderboard
We show a mini-leaderboard here and please find more information in our paper.
| Model | Abd | Ana | Cau | Ded | Ind | Spa | Tem | Avg |
|:------|----:|----:|----:|----:|----:|----:|----:|----:|
| Human* | 79.76 | 57.65 | 75.00 | 70.59 | 63.41 | 79.76 | 79.76 | 72.28 |
| GPT-5 | 53.57 | 78.37 | 41.06 | 79.86 | 77.25 | 17.25 | 41.47 | 58.48 |
| Gemini-2.5-Pro | 54.22 | 73.36 | 36.99 | 79.15 | 72.30 | 25.10 | 35.60 | 56.58 |
| Gemini-2.5-Flash | 46.10 | 74.57 | 34.22 | 71.38 | 73.42 | 23.92 | 30.64 | 53.03 |
| o4-mini | 41.23 | 73.01 | 27.38 | 71.02 | 67.12 | 19.22 | 32.48 | 50.30 |
| GPT-5-mini | 44.81 | 69.55 | 32.32 | 74.91 | 68.02 | 12.16 | 29.36 | 49.70 |
| GPT-4.1 | 44.16 | 71.11 | 22.43 | 67.14 | 69.37 | 13.73 | 27.16 | 48.09 |
| Claude-Sonnet-4 | 36.84 | 60.55 | 44.11 | 66.78 | 55.63 | 15.69 | 28.07 | 45.11 |
| Claude-3.7-Sonnet | 33.44 | 66.09 | 35.36 | 59.72 | 59.01 | 20.78 | 25.87 | 44.96 |
| GPT-4o | 46.75 | 65.22 | 25.86 | 51.24 | 65.32 | 11.37 | 25.87 | 44.62 |
| GPT-4.1-mini | 32.79 | 60.90 | 30.80 | 51.94 | 64.64 | 16.47 | 30.46 | 43.95 |
| Qwen2.5-VL-72B | 35.06 | 55.02 | 35.36 | 51.94 | 54.73 | 12.94 | 23.67 | 40.02 |
| Doubao-1.5-vision | 37.01 | 53.29 | 31.18 | 59.36 | 54.50 | 12.16 | 22.94 | 39.99 |
| VL-Rethinker-72B | 36.36 | 50.52 | 33.84 | 55.83 | 57.88 | 15.29 | 21.65 | 39.80 |
| Gemma3-27B | 35.71 | 57.79 | 36.88 | 31.80 | 60.81 | 13.33 | 18.72 | 38.75 |
| MM-Eureka-Qwen-32B | 23.70 | 42.56 | 25.48 | 49.12 | 28.83 | 16.86 | 17.98 | 29.67 |
| Gemma3-12B | 24.35 | 51.21 | 15.97 | 28.27 | 43.47 | 10.59 | 16.15 | 29.93 |
| MiMo-VL-7B-RL | 38.31 | 26.47 | 28.14 | 62.90 | 25.23 | 13.33 | 20.73 | 29.22 |
| Qwen2.5-VL-32B | 24.35 | 42.73 | 21.67 | 50.18 | 26.58 | 14.90 | 16.51 | 28.66 |
| VL-Rethinker-7B | 30.84 | 40.48 | 21.29 | 28.62 | 43.02 | 13.73 | 11.93 | 28.29 |
| Qwen2.5-VL-7B | 25.97 | 35.64 | 21.29 | 22.26 | 40.32 | 9.02 | 12.48 | 25.22 |
| InternVL3.5-30B-A3B | 48.05 | 18.17 | 33.08 | 37.46 | 13.29 | 13.33 | 13.39 | 22.87 |
| Keye-VL-1.5-8B | 19.48 | 21.63 | 23.19 | 13.78 | 19.59 | 13.73 | 23.30 | 19.96 |
| InternVL3.5-8B | 35.71 | 9.86 | 19.01 | 32.16 | 10.14 | 13.33 | 17.43 | 18.01 |
| Skywork-R1V-38B | 24.03 | 9.52 | 16.35 | 24.03 | 11.04 | 9.80 | 10.28 | 13.83 |
## Contact
Jiachun Li: jiachun.li@nlpr.ia.ac.cn
## Citation
``` | 33 | 1 | [
"task_categories:image-to-text",
"task_categories:multiple-choice",
"task_categories:question-answering",
"task_categories:time-series-forecasting",
"task_categories:visual-question-answering",
"task_ids:multiple-choice-qa",
"annotations_creators:expert-generated",
"language_creators:machine-generated",
"language_creators:expert-generated",
"source_datasets:original",
"language:en",
"license:cc-by-4.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:image",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"real-world"
] | 2025-11-09T12:21:57+00:00 | 2025-11-11T13:55:09+00:00 | 1 |
sandysanta/aero_data_1 |
This is an aerospace dataset. Currently made using 2 books, Aerofoil Theory and Astronautics the Physics of Space Flight. |
This is an aerospace dataset. Currently made using 2 books, Aerofoil Theory and Astronautics the Physics of Space Flight. | 16 | 0 | [
"task_categories:table-question-answering",
"task_categories:question-answering",
"task_categories:text-classification",
"task_categories:summarization",
"task_categories:text-generation",
"task_categories:feature-extraction",
"language:en",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"aerospace",
"physics",
"fine_tuning",
"dataset"
] | 2025-11-08T14:03:43+00:00 | 2025-11-11T13:54:49+00:00 | 0 |
AHegai/gray-foam |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 10508,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 10508,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 16 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:48:36+00:00 | 2025-11-11T13:49:09+00:00 | 0 |
ryankim17920/histopath-images | Data sources include:
Kather, J. N., Halama, N., & Marx, A. (2018). 100,000 histological images of human colorectal cancer and healthy tissue (v0.1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.1214456 | Data sources include:
Kather, J. N., Halama, N., & Marx, A. (2018). 100,000 histological images of human colorectal cancer and healthy tissue (v0.1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.1214456 | 1 | 0 | [
"region:us"
] | 2025-11-11T13:47:02+00:00 | 2025-11-11T13:48:09+00:00 | 0 |
johannesschirrmeister/eval_stacking_groot |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 30,
"total_frames": 18743,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:30"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 30,
"total_frames": 18743,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:30"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 30 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:55:12+00:00 | 2025-11-11T13:53:04+00:00 | 0 |
TheFactoryX/edition_0311_newtextdoc1111-danbooru-tag-csv-readymade |
# edition_0311_newtextdoc1111-danbooru-tag-csv-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[newtextdoc1111/danbooru-tag-csv](https://huggingface.co/datasets/newtextdoc1111/danbooru-tag-csv)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0311_newtextdoc1111-danbooru-tag-csv-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[newtextdoc1111/danbooru-tag-csv](https://huggingface.co/datasets/newtextdoc1111/danbooru-tag-csv)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 3 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:tabular",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T13:40:55+00:00 | 2025-11-11T13:40:58+00:00 | 0 |
RogersPyke/realman_rmc_aidal_fold_shorts |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the shorts is place with the front facing upwards, the left gripper grasp the waist of the shorts, and the right gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the back facing up, the left gripper grasp the waist of the shorts, and the right gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the front facing upwards, the right gripper grasp the waist of the shorts, and the left gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the back facing up, the right gripper grasp the waist of the shorts, and the left gripper grasp the bottom of the shorts and folds them in the middle.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 875, "total_frames": 753834, "total_tasks": 4, "total_videos": 2625, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:875"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the shorts is place with the front facing upwards, the left gripper grasp the waist of the shorts, and the right gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the back facing up, the left gripper grasp the waist of the shorts, and the right gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the front facing upwards, the right gripper grasp the waist of the shorts, and the left gripper grasp the bottom of the shorts and folds them in the middle.
the shorts is place with the back facing up, the right gripper grasp the waist of the shorts, and the left gripper grasp the bottom of the shorts and folds them in the middle.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 875, "total_frames": 753834, "total_tasks": 4, "total_videos": 2625, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:875"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 131 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-11T13:29:35+00:00 | 2025-11-11T13:46:36+00:00 | 0 |
RogersPyke/realman_rmc_aidal_organise_the_document_bag |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
open the green document bag with left hand.
open the green document bag with right hand.
open the red document bag with left hand.
open the red document bag with right hand.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 485, "total_frames": 232800, "total_tasks": 4, "total_videos": 1455, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:485"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
open the green document bag with left hand.
open the green document bag with right hand.
open the red document bag with left hand.
open the red document bag with right hand.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 485, "total_frames": 232800, "total_tasks": 4, "total_videos": 1455, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:485"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 23 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-11T13:23:49+00:00 | 2025-11-11T13:29:04+00:00 | 0 |
jordi2987/wipe_spill_new |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 70,
"total_frames": 48006,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:70"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 70,
"total_frames": 48006,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:70"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.front": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 18 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:27:36+00:00 | 2025-11-11T13:28:11+00:00 | 0 |
RogersPyke/realman_rmc_aidal_food_packaging |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the left gripper pick up the blue bag, the right gripper successively pick up the lunch box, cucumber and pear and place them into the bag, the right gripper zip up the bag.
the left gripper pick up the blue bag, the right gripper successively pick up the lunch box, bananna and peach and place them into the bag, the right gripper zip up the bag.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 501, "total_frames": 831220, "total_tasks": 2, "total_videos": 1503, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:501"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the left gripper pick up the blue bag, the right gripper successively pick up the lunch box, cucumber and pear and place them into the bag, the right gripper zip up the bag.
the left gripper pick up the blue bag, the right gripper successively pick up the lunch box, bananna and peach and place them into the bag, the right gripper zip up the bag.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 501, "total_frames": 831220, "total_tasks": 2, "total_videos": 1503, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:501"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 22 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-11T13:07:54+00:00 | 2025-11-11T13:23:08+00:00 | 0 |
kuzheren/geometry-dash-levels-tensors-2 | # Geometry Dash Chunks HDF5 Dataset
## Описание
Этот датасет создан для обучения нейросетевых моделей (автоэнкодер, DiT, diffusion и др.) на реальных уровнях из Geometry Dash. Он хранится в форматах HDF5 (каждый файл содержит не более 5000 уровней) и подготовлен для максимально быстрой и эффективной работы с последовательностями чанков уровней.
## Структура данных
Каждый HDF5-файл содержит:
- Датасет: chunk_data — тензоры с чанками уровней.
- Датасет: valid_mask — булева маска валидных чанков в каждом уровне.
- Атрибут: metadata_json_list — JSON-список метаданных уровней из оригинальных .jsonl-файлов (кроме level_string и неважных служебных полей).
- Другие атрибуты описывают размерности тензоров и смысл признаков.
### chunk_data
Размерность: ```num_levels, max_seq_len, chunk_h, chunk_w, num_block_features```
**Значения: int32**
- num_levels — количество уровней в файле
- max_seq_len — максимальное число чанков среди всех уровней в этом файле
- chunk_h — высота чанка в сеточных "пикселях" (например, 32)
- chunk_w — ширина чанка (например, 128)
- num_block_features — количество признаков на ячейку
### valid_mask
Размерность: ```num_levels, max_seq_len```
**Тип: bool**
- Показывает, какие чанки в каждом уровне содержат реальные данные (True), а какие — добавлены паддингом (False).
### metadata_json_list (атрибут)
Это JSON-список метаданных каждого уровня в файле. Пример содержимого одного entry:
```
{
"level_id": 123456,
"level_name": "My Level",
"difficulty_stars": 5,
"length_code": 2,
"downloads": 1234,
"likes": 56,
"num_chunks_generated": 12
}
```
level_string (и похожие служебные поля) не сохраняются для экономии места и скорости доступа.
## Содержимое одного чанка
- Каждый чанк — это сетка ```chunk_h x chunk_w```.
- В каждой ячейке хранится массив из ```num_block_features``` чисел:
1. block_id — целое, идентификатор блока Geometry Dash (0 = пусто)
2. x_rel — индекс ячейки (0 .. chunk_w-1)
3. y_rel — индекс ячейки (0 .. chunk_h-1)
4. rotation_index — 0–3 (соответствует 0°/90°/180°/270°)
5. flip_combined — код флипа: 0=нет, 1=flip_y, 2=flip_x, 3=flip_x+flip_y
Пустая ячейка содержит block_id=0 и все остальные значения ноль.
## Как читать датасет
Пример на Python с использованием библиотеки h5py и numpy:
```
import h5py
import numpy as np
filename = "gd_dataset_chunked_part_1.h5"
with h5py.File(filename, "r") as hf:
chunk_data = hf["chunk_data"] # Размер: (num_levels, max_seq_len, chunk_h, chunk_w, num_block_features)
valid_mask = hf["valid_mask"] # Размер: (num_levels, max_seq_len)
meta_json = hf.attrs["metadata_json_list"]
metadata = json.loads(meta_json)
# Пример: получить все чанки первого уровня:
idx = 0
real_len = valid_mask[idx].sum()
level_chunks = chunk_data[idx, :real_len] # (real_len, chunk_h, chunk_w, num_block_features)
# Декодировать первый чанк в уровень:
chunk = level_chunks[0] # (chunk_h, chunk_w, num_block_features)
block_ids = chunk[:,:,0] # карта блоков
x_coords = chunk[:,:,1]
y_coords = chunk[:,:,2]
rotation_idxs = chunk[:,:,3]
flip_combined = chunk[:,:,4]
```
## Как использовать в DataLoader
- Для обучения transformer/DiT моделей: формируйте батчи из уровней (последовательностей чанков), используйте valid_mask для attention mask и masking в loss.
- Для автоэнкодера: берите отдельные чанки, считывайте ```chunk_h x chunk_w x num_block_features``` тензоры, пустые блоки можно игнорировать или паддить.
## Описание признаков
| Индекс | Название | Описание |
|--------|----------------|---------------------------------------------------------------|
| 0 | block_id | GD ID блока. 0 — пусто |
| 1 | x_rel | X (столбец) внутри чанка, 0 .. chunk_w-1 |
| 2 | y_rel | Y (строка) внутри чанка, 0 .. chunk_h-1 |
| 3 | rotation_index | Индекс поворота: 0=0°, 1=90°, 2=180°, 3=270° |
| 4 | flip_combined | 0=нет; 1=flip_y; 2=flip_x; 3=оба |
## Пример визуализации чанка
```
import matplotlib.pyplot as plt
plt.imshow(block_ids, cmap="tab20") # или cmap="nipy_spectral"
plt.title("Карта ID блоков в чанке")
plt.show()
```
## Лицензия и источник
- Данные Levels: из open Geometry Dash (2013-2025).
- Код парсера и структуры: Kuzheren (actually, not really), 2025.
- Используйте свободно для ML-исследований и геймдев-прототипирования!
## Обратная связь
Вопросы, предложения и баги — в Issues HuggingFace или на [github.com/kuzheren/gdparse](https://github.com/kuzheren/gdparse)
| # Geometry Dash Chunks HDF5 Dataset
## Описание
Этот датасет создан для обучения нейросетевых моделей (автоэнкодер, DiT, diffusion и др.) на реальных уровнях из Geometry Dash. Он хранится в форматах HDF5 (каждый файл содержит не более 5000 уровней) и подготовлен для максимально быстрой и эффективной работы с последовательностями чанков уровней.
## Структура данных
Каждый HDF5-файл содержит:
- Датасет: chunk_data — тензоры с чанками уровней.
- Датасет: valid_mask — булева маска валидных чанков в каждом уровне.
- Атрибут: metadata_json_list — JSON-список метаданных уровней из оригинальных .jsonl-файлов (кроме level_string и неважных служебных полей).
- Другие атрибуты описывают размерности тензоров и смысл признаков.
### chunk_data
Размерность: ```num_levels, max_seq_len, chunk_h, chunk_w, num_block_features```
**Значения: int32**
- num_levels — количество уровней в файле
- max_seq_len — максимальное число чанков среди всех уровней в этом файле
- chunk_h — высота чанка в сеточных "пикселях" (например, 32)
- chunk_w — ширина чанка (например, 128)
- num_block_features — количество признаков на ячейку
### valid_mask
Размерность: ```num_levels, max_seq_len```
**Тип: bool**
- Показывает, какие чанки в каждом уровне содержат реальные данные (True), а какие — добавлены паддингом (False).
### metadata_json_list (атрибут)
Это JSON-список метаданных каждого уровня в файле. Пример содержимого одного entry:
```
{
"level_id": 123456,
"level_name": "My Level",
"difficulty_stars": 5,
"length_code": 2,
"downloads": 1234,
"likes": 56,
"num_chunks_generated": 12
}
```
level_string (и похожие служебные поля) не сохраняются для экономии места и скорости доступа.
## Содержимое одного чанка
- Каждый чанк — это сетка ```chunk_h x chunk_w```.
- В каждой ячейке хранится массив из ```num_block_features``` чисел:
1. block_id — целое, идентификатор блока Geometry Dash (0 = пусто)
2. x_rel — индекс ячейки (0 .. chunk_w-1)
3. y_rel — индекс ячейки (0 .. chunk_h-1)
4. rotation_index — 0–3 (соответствует 0°/90°/180°/270°)
5. flip_combined — код флипа: 0=нет, 1=flip_y, 2=flip_x, 3=flip_x+flip_y
Пустая ячейка содержит block_id=0 и все остальные значения ноль.
## Как читать датасет
Пример на Python с использованием библиотеки h5py и numpy:
```
import h5py
import numpy as np
filename = "gd_dataset_chunked_part_1.h5"
with h5py.File(filename, "r") as hf:
chunk_data = hf["chunk_data"] # Размер: (num_levels, max_seq_len, chunk_h, chunk_w, num_block_features)
valid_mask = hf["valid_mask"] # Размер: (num_levels, max_seq_len)
meta_json = hf.attrs["metadata_json_list"]
metadata = json.loads(meta_json)
# Пример: получить все чанки первого уровня:
idx = 0
real_len = valid_mask[idx].sum()
level_chunks = chunk_data[idx, :real_len] # (real_len, chunk_h, chunk_w, num_block_features)
# Декодировать первый чанк в уровень:
chunk = level_chunks[0] # (chunk_h, chunk_w, num_block_features)
block_ids = chunk[:,:,0] # карта блоков
x_coords = chunk[:,:,1]
y_coords = chunk[:,:,2]
rotation_idxs = chunk[:,:,3]
flip_combined = chunk[:,:,4]
```
## Как использовать в DataLoader
- Для обучения transformer/DiT моделей: формируйте батчи из уровней (последовательностей чанков), используйте valid_mask для attention mask и masking в loss.
- Для автоэнкодера: берите отдельные чанки, считывайте ```chunk_h x chunk_w x num_block_features``` тензоры, пустые блоки можно игнорировать или паддить.
## Описание признаков
| Индекс | Название | Описание |
|--------|----------------|---------------------------------------------------------------|
| 0 | block_id | GD ID блока. 0 — пусто |
| 1 | x_rel | X (столбец) внутри чанка, 0 .. chunk_w-1 |
| 2 | y_rel | Y (строка) внутри чанка, 0 .. chunk_h-1 |
| 3 | rotation_index | Индекс поворота: 0=0°, 1=90°, 2=180°, 3=270° |
| 4 | flip_combined | 0=нет; 1=flip_y; 2=flip_x; 3=оба |
## Пример визуализации чанка
```
import matplotlib.pyplot as plt
plt.imshow(block_ids, cmap="tab20") # или cmap="nipy_spectral"
plt.title("Карта ID блоков в чанке")
plt.show()
```
## Лицензия и источник
- Данные Levels: из open Geometry Dash (2013-2025).
- Код парсера и структуры: Kuzheren (actually, not really), 2025.
- Используйте свободно для ML-исследований и геймдев-прототипирования!
## Обратная связь
Вопросы, предложения и баги — в Issues HuggingFace или на [github.com/kuzheren/gdparse](https://github.com/kuzheren/gdparse)
| 10 | 0 | [
"license:apache-2.0",
"region:us"
] | 2025-06-02T22:03:01+00:00 | 2025-11-11T13:27:28+00:00 | 0 |
LCO-Embedding/SeaDoc |
# SeaDoc: from the paper "Scaling Language-Centric Omnimodal Representation Learning"
This repository hosts the **SeaDoc** dataset, a challenging visual document retrieval task in Southeast Asian languages, introduced in the paper [Scaling Language-Centric Omnimodal Representation Learning](https://huggingface.co/papers/2510.11693). It is designed to evaluate and enhance language-centric omnimodal embedding frameworks by focusing on a low-resource setting, specifically for tasks involving diverse languages and visual document understanding.
**Paper:** [https://huggingface.co/papers/2510.11693](https://huggingface.co/papers/2510.11693)
**Project Page:** [https://huggingface.co/LCO-Embedding](https://huggingface.co/LCO-Embedding)
**Code/Github:** [https://github.com/LCO-Embedding/LCO-Embedding](https://github.com/LCO-Embedding/LCO-Embedding)
# SeaDoc
**SeaDoc** is introduced in the last part of the [**LCO-Embedding**](https://huggingface.co/papers/2510.11693) paper.
**SeaDoc** partly provides evidence for our proposed **"Generation-Representation Scaling Law"**. As shown in the following figure. By conducting **continue-pretraining** on Qwen2.5-VL-3B before conducting the same amount of text-only contrastive learning, the performance on SeaDoc gradually improves, outperforming the baseline (Qwen2.5-VL-3B + text-only contrastive learning).
The four continue pretraining settings in the figure are respectively: 1. SeaDoc-train. 2. SeaDoc-train (high-resolution). 3. SeaDoc-train + PixmoCaps 4. SeaDoc-train + PixmoCaps (high-resolution). We show that it is important to add general-domain image-captioning dataset in order to keep model's pretrained knowledge, and add data of target capabilities (low-resource southasian OCR capability in our case). Importantly, OCR data must be trained with high resolution to prevent introducing hallucination.
<div align='left'><img src="https://cdn-uploads.huggingface.co/production/uploads/63108cc834c7d77420b0fd68/ljE-Mvb1__9kzEQYep0yp.png" alt="overview" width="50%"/></div>
# Construction process of SeaDoc
We first curate a corpus of 5, 055 pages drawn from 29 book publications from in-house collections across four SEA languages—Thai, Vietnamese, Malay, and Lao. The documents span diverse subject areas, including economics, natural sciences, technology, history, politics, art, psychology, education, and country reports. We design a rigorous pipeline that uses Gemini-2.5-Flash to generate queries for each document page, ensuring that each query maps uniquely to its ground-truth page and that no other page in the corpus is a valid match, thereby eliminating false negatives. Human annotators then filter out low-quality queries. This process yields 1, 001 high-quality English queries for retrieval over the 5, 055-page corpus in Southeast Asian languages.
We utilize Gemini-2.5-Flash to annotate each PDF page by sequentially applying OCR, translating the content into English, and generating an English query answerable exclusively from that specific page. This results in 5, 055 annotated {OCR, English translation, English query} triplets. To construct a high-quality query pool for the retrieval dataset in SeaDoc, we implement a three-stage quality control process:
1. Qwen2.5-7B-Instruct is first used to filter out functional pages (e.g., title pages, author pages, tables of contents), which reduces the dataset to 4, 491 content page annotations.
2. The same model then scores these annotations for Quality and Groundedness on a 10-point scale. Only questions with a quality score of at least 9 and a groundedness score of 10 are retained. Note that Quality measures the informativeness of the content and relevance of the query, and Groundedness measures the exclusivity of the answer to the page.
3. Our in-house linguists conduct a final review of the remaining triplets to ensure their quality. As a result, we derive 1, 001 high-quality queries to be used for retrieval tasks within the 5, 055 page corpus.
For conducting additional OCR-intensive generative training, we construct a training set leveraging images that do not correspond to retrieval test set queries, resulting in 4k seed images. We construct 5 SFT tasks per image: 1) OCR the image. 2) OCR the image, then generate a question from the image. 3) Provide the English translation given the OCR’d text. 4) Provide the English translation directly from the image. 5) Provide the answer to the generated query. Note that compared to the SeaDoc test set, the training set is separately generated and includes an additional “provide answer to the generated question” part in the seed prompt. This process leads us to an around 20k training set to enhance targeted generative capability on low-resource visual documents, which we also explore combining with the PixmoCap dataset (710k) for general capability preservation in the main experiments.
## More about the LCO-Embedding project Overview
- We introduce **LCO-Embedding**, a language-centric omnimodal representation learning method and the LCO-Embedding model families, setting a new state-of-the-art on [MIEB](https://huggingface.co/blog/isaacchung/introducing-mieb) (Massive Image Embedding Benchmark), while supporting audio and videos.
- We introduce the **Generation-Representation Scaling Law**, and connect models' generative capabilities and their representation upper bound.
- We introduce **SeaDoc**, a challenging visual document retrieval task in Southeast Asian languages, and show that continual generative pretraining before contrastive learning raises the representation upper bound.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/604f67ef0fe8ff3ec13d71ef/4Wd8fDFBdT6GxqN6-KzZN.png" alt="overview" width="100%"/></div>
## Evaluation Results
We evaluate LCO-Embedding with the state-of-the-art embedding models, including E5-V, Voyage Multimodal 3, mmE5, and GME, on a MIEB-Lite benchmark (51 tasks) broken down by task categories.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/63108cc834c7d77420b0fd68/63WBsKh57HbNwwe3bZ-oZ.png" alt="mieb_lite" width="100%"/></div>
Performance and efficiency comparisons of different training strategies using 3B and 7B variants of Qwen2.5-VL backbones.
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/lora_ablation.png" alt="lora_ablation" width="100%"/></div>
Scaling relationship between generation benchmark performance (X-axis) and representation benchmark performance after language-centric contrastive learning (Y-axis).
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/scaling.png" alt="scaling" width="100%"/></div>
## Citation
If you find LCO-Embedding useful for your research and applications, please cite using this BibTeX:
```bibtex
@misc{xiao2025scaling,
title={Scaling Language-Centric Omnimodal Representation Learning},
author={Chenghao Xiao and Hou Pong Chan and Hao Zhang and Weiwen Xu and Mahani Aljunied and Yu Rong},
year={2025},
eprint={2510.11693},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2510.11693},
}
``` |
# SeaDoc: from the paper "Scaling Language-Centric Omnimodal Representation Learning"
This repository hosts the **SeaDoc** dataset, a challenging visual document retrieval task in Southeast Asian languages, introduced in the paper [Scaling Language-Centric Omnimodal Representation Learning](https://huggingface.co/papers/2510.11693). It is designed to evaluate and enhance language-centric omnimodal embedding frameworks by focusing on a low-resource setting, specifically for tasks involving diverse languages and visual document understanding.
**Paper:** [https://huggingface.co/papers/2510.11693](https://huggingface.co/papers/2510.11693)
**Project Page:** [https://huggingface.co/LCO-Embedding](https://huggingface.co/LCO-Embedding)
**Code/Github:** [https://github.com/LCO-Embedding/LCO-Embedding](https://github.com/LCO-Embedding/LCO-Embedding)
# SeaDoc
**SeaDoc** is introduced in the last part of the [**LCO-Embedding**](https://huggingface.co/papers/2510.11693) paper.
**SeaDoc** partly provides evidence for our proposed **"Generation-Representation Scaling Law"**. As shown in the following figure. By conducting **continue-pretraining** on Qwen2.5-VL-3B before conducting the same amount of text-only contrastive learning, the performance on SeaDoc gradually improves, outperforming the baseline (Qwen2.5-VL-3B + text-only contrastive learning).
The four continue pretraining settings in the figure are respectively: 1. SeaDoc-train. 2. SeaDoc-train (high-resolution). 3. SeaDoc-train + PixmoCaps 4. SeaDoc-train + PixmoCaps (high-resolution). We show that it is important to add general-domain image-captioning dataset in order to keep model's pretrained knowledge, and add data of target capabilities (low-resource southasian OCR capability in our case). Importantly, OCR data must be trained with high resolution to prevent introducing hallucination.
<div align='left'><img src="https://cdn-uploads.huggingface.co/production/uploads/63108cc834c7d77420b0fd68/ljE-Mvb1__9kzEQYep0yp.png" alt="overview" width="50%"/></div>
# Construction process of SeaDoc
We first curate a corpus of 5, 055 pages drawn from 29 book publications from in-house collections across four SEA languages—Thai, Vietnamese, Malay, and Lao. The documents span diverse subject areas, including economics, natural sciences, technology, history, politics, art, psychology, education, and country reports. We design a rigorous pipeline that uses Gemini-2.5-Flash to generate queries for each document page, ensuring that each query maps uniquely to its ground-truth page and that no other page in the corpus is a valid match, thereby eliminating false negatives. Human annotators then filter out low-quality queries. This process yields 1, 001 high-quality English queries for retrieval over the 5, 055-page corpus in Southeast Asian languages.
We utilize Gemini-2.5-Flash to annotate each PDF page by sequentially applying OCR, translating the content into English, and generating an English query answerable exclusively from that specific page. This results in 5, 055 annotated {OCR, English translation, English query} triplets. To construct a high-quality query pool for the retrieval dataset in SeaDoc, we implement a three-stage quality control process:
1. Qwen2.5-7B-Instruct is first used to filter out functional pages (e.g., title pages, author pages, tables of contents), which reduces the dataset to 4, 491 content page annotations.
2. The same model then scores these annotations for Quality and Groundedness on a 10-point scale. Only questions with a quality score of at least 9 and a groundedness score of 10 are retained. Note that Quality measures the informativeness of the content and relevance of the query, and Groundedness measures the exclusivity of the answer to the page.
3. Our in-house linguists conduct a final review of the remaining triplets to ensure their quality. As a result, we derive 1, 001 high-quality queries to be used for retrieval tasks within the 5, 055 page corpus.
For conducting additional OCR-intensive generative training, we construct a training set leveraging images that do not correspond to retrieval test set queries, resulting in 4k seed images. We construct 5 SFT tasks per image: 1) OCR the image. 2) OCR the image, then generate a question from the image. 3) Provide the English translation given the OCR’d text. 4) Provide the English translation directly from the image. 5) Provide the answer to the generated query. Note that compared to the SeaDoc test set, the training set is separately generated and includes an additional “provide answer to the generated question” part in the seed prompt. This process leads us to an around 20k training set to enhance targeted generative capability on low-resource visual documents, which we also explore combining with the PixmoCap dataset (710k) for general capability preservation in the main experiments.
## More about the LCO-Embedding project Overview
- We introduce **LCO-Embedding**, a language-centric omnimodal representation learning method and the LCO-Embedding model families, setting a new state-of-the-art on [MIEB](https://huggingface.co/blog/isaacchung/introducing-mieb) (Massive Image Embedding Benchmark), while supporting audio and videos.
- We introduce the **Generation-Representation Scaling Law**, and connect models' generative capabilities and their representation upper bound.
- We introduce **SeaDoc**, a challenging visual document retrieval task in Southeast Asian languages, and show that continual generative pretraining before contrastive learning raises the representation upper bound.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/604f67ef0fe8ff3ec13d71ef/4Wd8fDFBdT6GxqN6-KzZN.png" alt="overview" width="100%"/></div>
## Evaluation Results
We evaluate LCO-Embedding with the state-of-the-art embedding models, including E5-V, Voyage Multimodal 3, mmE5, and GME, on a MIEB-Lite benchmark (51 tasks) broken down by task categories.
<div align='center'><img src="https://cdn-uploads.huggingface.co/production/uploads/63108cc834c7d77420b0fd68/63WBsKh57HbNwwe3bZ-oZ.png" alt="mieb_lite" width="100%"/></div>
Performance and efficiency comparisons of different training strategies using 3B and 7B variants of Qwen2.5-VL backbones.
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/lora_ablation.png" alt="lora_ablation" width="100%"/></div>
Scaling relationship between generation benchmark performance (X-axis) and representation benchmark performance after language-centric contrastive learning (Y-axis).
<div align='center'><img src="https://github.com/LCO-Embedding/LCO-Embedding/raw/main/assets/scaling.png" alt="scaling" width="100%"/></div>
## Citation
If you find LCO-Embedding useful for your research and applications, please cite using this BibTeX:
```bibtex
@misc{xiao2025scaling,
title={Scaling Language-Centric Omnimodal Representation Learning},
author={Chenghao Xiao and Hou Pong Chan and Hao Zhang and Weiwen Xu and Mahani Aljunied and Yu Rong},
year={2025},
eprint={2510.11693},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2510.11693},
}
``` | 660 | 2 | [
"task_categories:visual-document-retrieval",
"language:lo",
"language:vi",
"language:th",
"language:ms",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:image",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"arxiv:2510.11693",
"region:us"
] | 2025-10-15T03:34:52+00:00 | 2025-11-11T13:16:01+00:00 | 0 |
antwoor/screwdriver_95_rads |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8763,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8763,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 105 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:15:32+00:00 | 2025-11-11T13:15:43+00:00 | 0 |
antwoor/BFD_diff_tasks_rads |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8756,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8756,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 103 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:11:45+00:00 | 2025-11-11T13:11:55+00:00 | 0 |
antwoor/motor_95_rads |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 29,
"total_frames": 25158,
"total_tasks": 1,
"total_videos": 58,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:29"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 29,
"total_frames": 25158,
"total_tasks": 1,
"total_videos": 58,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:29"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 91 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:08:56+00:00 | 2025-11-11T13:09:04+00:00 | 0 |
RogersPyke/realman_rmc_aidal_clean_table |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
wipe off the black water stains in the middle of the table with a blue rag.
wipe off the brown water stains in the middle of the table with a blue rag.
wipe off the black water stains in the middle of the table with a pruple rag.
wipe off the brown water stains in the middle of the table with a pruple rag.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 801, "total_frames": 534825, "total_tasks": 4, "total_videos": 2403, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:801"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
wipe off the black water stains in the middle of the table with a blue rag.
wipe off the brown water stains in the middle of the table with a blue rag.
wipe off the black water stains in the middle of the table with a pruple rag.
wipe off the brown water stains in the middle of the table with a pruple rag.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 801, "total_frames": 534825, "total_tasks": 4, "total_videos": 2403, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:801"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 131 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-11T12:57:42+00:00 | 2025-11-11T13:07:11+00:00 | 0 |
marinero4972/Open-o3-Video |
# Open-o3 Video
**TL; DR**: Open-o3 Video integrates explicit spatio-temporal evidence into video reasoning through curated STGR datasets and a two-stage SFT–RL training strategy, achieving state-of-the-art results on V-STAR and delivering verifiable, reliable reasoning for video understanding.
# Data
To provide unified spatio-temporal supervision for grounded video reasoning, we build two datasets: STGR-CoT-30k for supervised fine-tuning and STGR-RL-36k for reinforcement learning.
## Dataset Structure
The **Open-o3-Video-data** dataset is organized as follows:
```
Open-o3-Video-data/
├── json_data/
│ ├── STGR-RL.json # Data for reinforcement learning fine-tuning
│ └── STGR-SFT.json # Data for supervised fine-tuning
│
├── videos/
│ ├── gqa/
│ ├── stgr/
│ │ ├── plm/
│ │ │ ├── kfs/ # Key-frame samples for PLM subset
│ │ │ └── videos/ # Corresponding video files
│ │ └── temporal_grounding/
│ │ ├── kfs/ # Key-frame samples for temporal grounding
│ │ └── videos/ # Corresponding video files
│ │
│ ├── timerft/
│ │
│ ├── treevgr/images/
│ │
│ ├── tvg_r1/
│ │ ├── GroundedVLLM/
│ │ └── videomind_data/
│ │
│ ├── videoespresso/
│ │ ├── kfs/ # Key-frame samples
│ │ └── videos/ # Corresponding video files
│ │
│ └── videor1/
```
## Image and Video Data Download Instructions
**GQA**
The image data can be downloaded from [Visual-CoT](https://github.com/deepcs233/Visual-CoT).The directory directly contains `.jpg` images.
**STGR**
Already provided in the repository.
**TimeRFT**
Video data can be downloaded from [Hugging Face – TimeR1 Dataset](https://huggingface.co/datasets/Boshenxx/TimeR1-Dataset). The directory directly contains `.mp4` files.
**TreeVGR**
Follow the setup instructions in [TreeVGR](https://github.com/Haochen-Wang409/TreeVGR). The image data originates from **LLaVA-NeXT**.
**TVG-R1**
Video data can be downloaded by following the instructions in [TVG-R1](https://github.com/zjuruizhechen/TVG-R1).
**VideoEspresso**
Keyframe data is already provided in the repository. Full video data can be downloaded from [Hugging Face – VideoEspresso_train_video](https://huggingface.co/datasets/hshjerry0315/VideoEspresso_train_video).
**VideoR1**
Video data can be downloaded from [Hugging Face – Video-R1-data](https://huggingface.co/datasets/Video-R1/Video-R1-data).
|
# Open-o3 Video
**TL; DR**: Open-o3 Video integrates explicit spatio-temporal evidence into video reasoning through curated STGR datasets and a two-stage SFT–RL training strategy, achieving state-of-the-art results on V-STAR and delivering verifiable, reliable reasoning for video understanding.
# Data
To provide unified spatio-temporal supervision for grounded video reasoning, we build two datasets: STGR-CoT-30k for supervised fine-tuning and STGR-RL-36k for reinforcement learning.
## Dataset Structure
The **Open-o3-Video-data** dataset is organized as follows:
```
Open-o3-Video-data/
├── json_data/
│ ├── STGR-RL.json # Data for reinforcement learning fine-tuning
│ └── STGR-SFT.json # Data for supervised fine-tuning
│
├── videos/
│ ├── gqa/
│ ├── stgr/
│ │ ├── plm/
│ │ │ ├── kfs/ # Key-frame samples for PLM subset
│ │ │ └── videos/ # Corresponding video files
│ │ └── temporal_grounding/
│ │ ├── kfs/ # Key-frame samples for temporal grounding
│ │ └── videos/ # Corresponding video files
│ │
│ ├── timerft/
│ │
│ ├── treevgr/images/
│ │
│ ├── tvg_r1/
│ │ ├── GroundedVLLM/
│ │ └── videomind_data/
│ │
│ ├── videoespresso/
│ │ ├── kfs/ # Key-frame samples
│ │ └── videos/ # Corresponding video files
│ │
│ └── videor1/
```
## Image and Video Data Download Instructions
**GQA**
The image data can be downloaded from [Visual-CoT](https://github.com/deepcs233/Visual-CoT).The directory directly contains `.jpg` images.
**STGR**
Already provided in the repository.
**TimeRFT**
Video data can be downloaded from [Hugging Face – TimeR1 Dataset](https://huggingface.co/datasets/Boshenxx/TimeR1-Dataset). The directory directly contains `.mp4` files.
**TreeVGR**
Follow the setup instructions in [TreeVGR](https://github.com/Haochen-Wang409/TreeVGR). The image data originates from **LLaVA-NeXT**.
**TVG-R1**
Video data can be downloaded by following the instructions in [TVG-R1](https://github.com/zjuruizhechen/TVG-R1).
**VideoEspresso**
Keyframe data is already provided in the repository. Full video data can be downloaded from [Hugging Face – VideoEspresso_train_video](https://huggingface.co/datasets/hshjerry0315/VideoEspresso_train_video).
**VideoR1**
Video data can be downloaded from [Hugging Face – Video-R1-data](https://huggingface.co/datasets/Video-R1/Video-R1-data).
| 137 | 3 | [
"license:apache-2.0",
"modality:image",
"modality:video",
"region:us"
] | 2025-10-23T03:30:32+00:00 | 2025-11-11T13:05:54+00:00 | 0 |
antwoor/motor_95_degs |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 29,
"total_frames": 25158,
"total_tasks": 1,
"total_videos": 58,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:29"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 29,
"total_frames": 25158,
"total_tasks": 1,
"total_videos": 58,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:29"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 99 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:05:55+00:00 | 2025-11-11T13:06:32+00:00 | 0 |
suemincho/2D3D-RegQuality | ### Download the Dataset
```Bash
# Ensure git-lfs is installed (https://git-lfs.com)
git lfs install
# When prompted for a password, use an access token with write permissions
# Generate one from your settings: https://huggingface.co/settings/tokensll
git clone https://huggingface.co/datasets/suemincho/2D3D-RegQuality
```
### Dataset Structure
```Bash
dataset_real # dataset_simulated follows the same structure
├── Image_Poses_mTRE_binary.csv # CSV providing identifications and labels
└── specimen_folders/ # spec_id e.g., "17-1882", "18-0725", etc.
├── XXX/ # Projection folder where XXX is a 3-digit number from proj_idx, e.g., "001", "002", etc.
│ ├── xray.png # X-ray image
│ └── DRR/ # Folder containing DRR images for that projection
│ └── drr_remap_YYY.png # DRR image where YYY is a 3-digit number from sample_id, e.g., "000", "001", etc.
└── ...
```
Each CSV row defines one sample by specifying:
- spec_id: should match the provided specimen folder name
- proj_idx: used to determine the projection folder name
- sample_id: used to determine the DRR file name
- binary: the binary registration quality label | ### Download the Dataset
```Bash
# Ensure git-lfs is installed (https://git-lfs.com)
git lfs install
# When prompted for a password, use an access token with write permissions
# Generate one from your settings: https://huggingface.co/settings/tokensll
git clone https://huggingface.co/datasets/suemincho/2D3D-RegQuality
```
### Dataset Structure
```Bash
dataset_real # dataset_simulated follows the same structure
├── Image_Poses_mTRE_binary.csv # CSV providing identifications and labels
└── specimen_folders/ # spec_id e.g., "17-1882", "18-0725", etc.
├── XXX/ # Projection folder where XXX is a 3-digit number from proj_idx, e.g., "001", "002", etc.
│ ├── xray.png # X-ray image
│ └── DRR/ # Folder containing DRR images for that projection
│ └── drr_remap_YYY.png # DRR image where YYY is a 3-digit number from sample_id, e.g., "000", "001", etc.
└── ...
```
Each CSV row defines one sample by specifying:
- spec_id: should match the provided specimen folder name
- proj_idx: used to determine the projection folder name
- sample_id: used to determine the DRR file name
- binary: the binary registration quality label | 25 | 0 | [
"license:mit",
"modality:image",
"doi:10.57967/hf/6825",
"region:us"
] | 2025-10-28T19:37:15+00:00 | 2025-11-11T13:10:56+00:00 | 0 |
smcleish/retrofitting-llama-fineweb-edu-tokenized | This is the [350b FineWeb-Edu sample](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu/tree/main/sample/350BT), tokenized with the Llama-3 tokenizer for [Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence](https://arxiv.org/abs/2511.07384)
Please see our paper and [model collection](https://huggingface.co/collections/tomg-group-umd/retrofitting-recurrence), for more information.
# Streaming the Dataset
You can use datatrove to efficiently stream the dataset. Note ParquetReader, returns rows with "text_key" and "id_key", as the field names.
```
from datatrove.pipeline.readers import ParquetReader
data_reader = ParquetReader("hf://datasets/smcleish/retrofitting-llama-fineweb-edu-tokenized/dataset", limit=1, text_key="input_ids", id_key="attention_mask")
for document in data_reader():
# do something with document
print(document)
```
# Contact
Please, feel free to contact us with any questions, or open a discussion thread.
# Citation
```
@article{mcleish2025teaching,
title={Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence},
author={Sean McLeish and Ang Li and John Kirchenbauer and Dayal Singh Kalra and Brian R. Bartoldson and Bhavya Kailkhura and Avi Schwarzschild and Jonas Geiping and Tom Goldstein and Micah Goldblum},
journal={arXiv preprint arXiv:2511.07384},
year={2025}
} | This is the [350b FineWeb-Edu sample](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu/tree/main/sample/350BT), tokenized with the Llama-3 tokenizer for [Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence](https://arxiv.org/abs/2511.07384)
Please see our paper and [model collection](https://huggingface.co/collections/tomg-group-umd/retrofitting-recurrence), for more information.
# Streaming the Dataset
You can use datatrove to efficiently stream the dataset. Note ParquetReader, returns rows with "text_key" and "id_key", as the field names.
```
from datatrove.pipeline.readers import ParquetReader
data_reader = ParquetReader("hf://datasets/smcleish/retrofitting-llama-fineweb-edu-tokenized/dataset", limit=1, text_key="input_ids", id_key="attention_mask")
for document in data_reader():
# do something with document
print(document)
```
# Contact
Please, feel free to contact us with any questions, or open a discussion thread.
# Citation
```
@article{mcleish2025teaching,
title={Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence},
author={Sean McLeish and Ang Li and John Kirchenbauer and Dayal Singh Kalra and Brian R. Bartoldson and Bhavya Kailkhura and Avi Schwarzschild and Jonas Geiping and Tom Goldstein and Micah Goldblum},
journal={arXiv preprint arXiv:2511.07384},
year={2025}
} | 69 | 0 | [
"language:en",
"license:apache-2.0",
"size_categories:100M<n<1B",
"format:parquet",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"arxiv:2511.07384",
"region:us"
] | 2025-11-07T15:45:46+00:00 | 2025-11-11T13:01:55+00:00 | 0 |
AnnLo/c-code | # Dataset Card for "c-code"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) | # Dataset Card for "c-code"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) | 5 | 0 | [
"size_categories:10K<n<100K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T13:03:04+00:00 | 2025-11-11T13:03:12+00:00 | 0 |
antwoor/BFD_diff_tasks_degs |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8756,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8756,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 21 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T13:01:22+00:00 | 2025-11-11T13:02:06+00:00 | 0 |
CCChen523/recorddata-pen1 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 21,
"total_frames": 21062,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:21"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.side": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 21,
"total_frames": 21062,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:21"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.side": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 78 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T12:06:50+00:00 | 2025-11-11T13:01:12+00:00 | 0 |
RogersPyke/realman_rmc_aidal_place_test_tube |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
pick up the test tube rack with the left hand, transfer it to the right, then place it into the test tube rack.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 655, "total_frames": 374039, "total_tasks": 1, "total_videos": 1965, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:655"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
pick up the test tube rack with the left hand, transfer it to the right, then place it into the test tube rack.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 655, "total_frames": 374039, "total_tasks": 1, "total_videos": 1965, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:655"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 20 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-08T09:07:39+00:00 | 2025-11-11T12:56:43+00:00 | 0 |
pnugues/EB7 |
# Dataset Card for EB7
_Encyclopædia Britannica_ (EB) is the most prestigious reference work in English.
This dataset contains the text of the entries from the 7th edition (EB7) as well automatically extracted geographical
coordinates.
The original text comes from the [Nineteenth-Century Knowledge Project](https://tu-plogan.github.io/source/r_7th_edition.html)
### Data Instances
Each sample is a JSON dictionnary with `id`, the entry ID in the [Nineteenth-Century Knowledge Project](https://tu-plogan.github.io/source/r_7th_edition.html)
nomenclature, `texte`, the text of the entry and `coords`, the coordinates if any, and a disclaimer, as for example:
```
{
"texte": "NARLAH, a town of Hindustan, in the province of Orissa, possessed by independent native chiefs. It is thirty miles east from the town of Bustar. Long. 83. 5. E. Lat. 19. 50. N. 0",
"id": "kp-eb0715-073501-8764",
"disclamer": "ENCYCLOPEDIA BRITANNICA, SEVENTH EDITION: A MACHINE-READABLE TEXT
TRANSCRIPTION (v3.1), The Nineteenth-Century Knowledge Project, 2024
nckp@temple.edu, https://tu-plogan.github.io/.
License: CC-BY-4.0, https://creativecommons.org/licenses/by/4.0/.
Source: Encyclopaedia Britannica: A Dictionary of Arts, Sciences,
and General Literature. 7th ed., 21 vols. Edinburgh: Adam and
Charles Black, 1830-1842. Image scans: Natl. Library of Scotland.
This entry: 7th edition, volume 15, page 735 [7:15:735]",
"coords": "19 50' N 83 5' E"}
}
```
The figure below shows the plot of all the coordinates we could extract.
### Code
The extraction and visualization code is available here: https://github.com/pnugues/EB7
### Citation Information
```
@misc{pnugues2025,
author = {Pierre Nugues},
title = {Extraction of geographical coordinates from the 7th edition of Encyclopædia Britannica},
year = 2025,
url = {https://huggingface.co/datasets/pnugues/EB7}
}
``` |
# Dataset Card for EB7
_Encyclopædia Britannica_ (EB) is the most prestigious reference work in English.
This dataset contains the text of the entries from the 7th edition (EB7) as well automatically extracted geographical
coordinates.
The original text comes from the [Nineteenth-Century Knowledge Project](https://tu-plogan.github.io/source/r_7th_edition.html)
### Data Instances
Each sample is a JSON dictionnary with `id`, the entry ID in the [Nineteenth-Century Knowledge Project](https://tu-plogan.github.io/source/r_7th_edition.html)
nomenclature, `texte`, the text of the entry and `coords`, the coordinates if any, and a disclaimer, as for example:
```
{
"texte": "NARLAH, a town of Hindustan, in the province of Orissa, possessed by independent native chiefs. It is thirty miles east from the town of Bustar. Long. 83. 5. E. Lat. 19. 50. N. 0",
"id": "kp-eb0715-073501-8764",
"disclamer": "ENCYCLOPEDIA BRITANNICA, SEVENTH EDITION: A MACHINE-READABLE TEXT
TRANSCRIPTION (v3.1), The Nineteenth-Century Knowledge Project, 2024
nckp@temple.edu, https://tu-plogan.github.io/.
License: CC-BY-4.0, https://creativecommons.org/licenses/by/4.0/.
Source: Encyclopaedia Britannica: A Dictionary of Arts, Sciences,
and General Literature. 7th ed., 21 vols. Edinburgh: Adam and
Charles Black, 1830-1842. Image scans: Natl. Library of Scotland.
This entry: 7th edition, volume 15, page 735 [7:15:735]",
"coords": "19 50' N 83 5' E"}
}
```
The figure below shows the plot of all the coordinates we could extract.
### Code
The extraction and visualization code is available here: https://github.com/pnugues/EB7
### Citation Information
```
@misc{pnugues2025,
author = {Pierre Nugues},
title = {Extraction of geographical coordinates from the 7th edition of Encyclopædia Britannica},
year = 2025,
url = {https://huggingface.co/datasets/pnugues/EB7}
}
``` | 16 | 0 | [
"task_categories:text-classification",
"language:en",
"license:cc-by-4.0",
"size_categories:10K<n<100K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-10T14:36:18+00:00 | 2025-11-11T12:55:08+00:00 | 0 |
HuangRi-believe/test | # LET:Full-Size Humanoid Robot Real-World Dataset
<hr style="margin-top: -10px;margin-bottom: 6px">
<div style="display: flex; justify-content: space-between; align-items: center; width: 100%;">
<div>
<a href="https://huggingface.co/datasets/LejuRobotics/let_dataset">
<img src="https://img.shields.io/badge/Huggingface-FF6B35?style=for-the-badge&logo=huggingface" alt="Huggingface">
</a>
<a href="https://www.modelscope.cn/datasets/LejuRobotics/let_dataset">
<img src="https://img.shields.io/badge/Modelscope-1890FF?style=for-the-badge&logo=alibabacloud" alt="Modelscope">
</a>
</div>
</div>
[中文](README.md)| [English]
<div style="font-size:1.1em; max-width:800px; margin: 0 0 16px 0; text-align: left;">
<b><span style="color:#000000">LET Dataset</span></b> is collected based on the full-size humanoid robot <b><span style="color:#1890FF">Kuavo 4 Pro</span></b> covering real-world multi-task data across multiple scenarios and operation types. It is designed for robot manipulation, mobility, and interaction tasks, supporting scalable robot learning in real environments.
</div>
## 📋 Table of Contents
<hr style="margin-top: -10px;margin-bottom: 6px">
- [Key Features](#key-features)
- [Hardware Platform](#hardware-platform)
- [Usage Guide](#usage-guide)
- [Dataset Download Example](#dataset-download-example)
- [Tool Repository](#tool-repository)
- [Tasks and Data Overview](#tasks-and-data-overview)
- [Semantic Labels](#semantic-labels)
- [Data Statistics](#data-statistics)
- [Dataset](#dataset)
- [Dataset Directory Structure](#dataset-directory-structure)
- [Data Format](#data-format)
- [Annotation Format](#annotation-format)
- [Citation](#citation)
- [License](#license)
<a id="key-features"></a>
## ✨ Key Features
<hr style="margin-top: -10px;margin-bottom: 6px">
- Large-scale, real-world, full-size humanoid robot multi-view, multi-modal data, continuously updated
- Covers multiple domains including industry, home, medical, and service, with 31 sub-task scenarios
- Includes 117 atomic skills such as grasping, bimanual operation, tool use, with a total duration of over 1000 hours
- Expert-labeled and human-verified data to ensure high quality
- Provides a complete toolchain from data conversion, model training to inference and validation
<div style="overflow-x: auto; text-align: left; max-width: fit-content; margin-left: 0;">
<table style="border-collapse: collapse; border-spacing: 0; width: auto; table-layout: auto;">
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/Assembly_line_sorting.gif" alt="Assembly line sorting" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly line sorting</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Clean the floor.gif" alt="Daily table cleaning" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Daily table cleaning</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Assembly_line_sorting-dex_hand.gif" alt="Assembly line sorting (dexterous hand)" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly line sorting (dexterous hand)</b></p>
</td>
</tr>
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_l.gif" alt="Left hand camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Left hand camera view</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_h.gif" alt="Head camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Head camera view</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_r.gif" alt="Right hand camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Right hand camera view</b></p>
</td>
</tr>
</table>
</div>
<a id="hardware-platform"></a>
## 🤖 Hardware Platform
<hr style="margin-top: -10px;margin-bottom: 6px">
<div align="left">
<img src="docs/images/kuavo4pro.png" alt="kuavo" width="200" style="display:inline-block; margin-right: 10px;">
<img src="docs/images/kuavo_wheel.png" alt="kuavo_wheel" width="200" style="display:inline-block;">
</div>
The main hardware platform is **Kuavo 4 Pro** and its wheeled version, with the following features:
- **Robot parameters:** Height **1.66 m**, weight **55 kg**, supports hot-swappable batteries
- **Motion control:** 40 degrees of freedom, max walking speed **7 km/h**, supports bipedal autonomous SLAM
- **Generalization:** Supports multi-modal large models (e.g., Pangu, DeepSeek, ChatGPT), with **20+ atomic skills**
<a id="usage-guide"></a>
## 🚀 Usage Guide
<hr style="margin-top: -10px;margin-bottom: 6px">
<a id="dataset-download-example"></a>
<a id="tool-repository"></a>
### Tool Repository
We provide a complete tool repository, including:
- **Data conversion tool (`rosbag2lerobot`)**: Convert rosbag files to formats suitable for model training
- **Two imitation learning models:** **Diffusion Policy** and **ACT**
- **Model training scripts**
- **Code and deployment instructions** for both real robots and simulation environments
For details, see the open-source repository: [**kuavo_data_challenge**](https://github.com/LejuRobotics/kuavo_data_challenge) 🔥
<a id="tasks-and-data-overview"></a>
## 🎬 Tasks and Data Overview
<hr style="margin-top: -10px;margin-bottom: 6px">
This dataset covers various scenarios such as automobile factories, FMCG, hotel services, 3C factories, life services, logistics, etc., including multi-modal observations (RGB, Depth, joints, etc.) and a rich set of atomic skills (grasping, bimanual operation, tool use, etc.).
<div style="overflow-x: auto; text-align: left; max-width: fit-content; margin-left: 0;">
<table style="border-collapse: collapse; border-spacing: 0; width: auto; table-layout: auto;">
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/Sorting.gif" alt="Consumer goods sorting" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Consumer goods sorting</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Simulation_resized.gif" alt="Simulation data demonstration" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Simulation data demonstration</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/3C.gif" alt="Assembly feeding" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly feeding</b></p>
</td>
</tr>
</table>
</div>
<a id="semantic-labels"></a>
### Semantic Labels
The LET dataset decomposes complex tasks into a series of atomic action steps with clear semantics, using standardized annotation methods to provide sub-task level timelines and natural language annotations for each task.
<div style="text-align: center;">
<img src="docs/images/Visualize Datasets.png" width="600">
</div>
Each data entry is accompanied by multi-dimensional semantic label information, including:
- Object labels: industrial parts, tableware, daily utensils, medicines, etc.
- Skill labels: grasp, place, rotate, push, pull, press, etc.
- Task and scene identifiers: unified task name coding, scene dimension distinguishes operation context semantics
- End effector type: records actions performed by gripper and dexterous hand separately
- Language description: e.g., "Pick up the medicine box from the conveyor belt and place it on the designated tray", supporting natural language and action alignment modeling
<a id="data-statistics"></a>
### Data Statistics
LET dataset statistics are as follows:
#### Data type & Scene distribution
| Data type distribution | Scene distribution |
|:---:|:---:|
| <img src="docs/images/Data type_en.png" width="500"> | <img src="docs/images/Scene distribution_en.png" width="500"> |
#### Task distribution
<div align="left">
<img src="docs/images/Task Distribution_en.png" width="800" alt="Task distribution">
</div>
#### Task duration distribution
<div align="left">
<img src="docs/images/Task duration distribution_en.png" width="800" alt="Task duration distribution">
</div>
#### Distribution of atomic skills
<div align="left">
<img src="docs/images/Distribution of Task Atomic Skills_en.png" width="800" alt="Distribution of atomic skills">
</div>
<a id="dataset"></a>
## 📦 Dataset
<hr style="margin-top: -10px;margin-bottom: 6px">
<a id="dataset-directory-structure"></a>
### Dataset Directory Structure
```text
.
├── real
│ ├── Labelled
│ │ ├── customer_check_in-P4-dex_hand
│ │ ├── deliver_room_card-P4-dex_hand
│ │ ├── deliver_water_bottle-P4-dex_hand
│ │ ├── loading_of_large_tooling-P4-dex_hand
│ │ ├── loading_of_small_tooling-P4-dex_hand
│ │ ├── more_coil_sorting-P4-dex_hand
│ │ ├── more_FMCG_loading-P4-dex_hand
│ │ ├── more_goods_orders-P4-dex_hand
│ │ ├── more_scan_code_for_weighing-P4-dex_hand
│ │ ├── parts_offline-P4-dex_hand
│ │ ├── quick_sort-P4-leju_claw
│ │ ├── single_coil_sorting-P4-dex_hand
│ │ ├── single_FMCG_loading-P4-dex_hand
│ │ ├── single_goods_orders-P4-dex_hand
│ │ ├── single_scan_code_for_weighing-P4-dex_hand
│ │ ├── SPS_parts_grab-P4-leju_claw
│ │ ├── SPS_parts_sorting-P4-dex_hand
│ │ └── task_mass_check-P4-leju_claw
│ └── Unlabelled
│ ├── assembly_line_sorting-P4-leju_claw
│ ├── deliver_room_card-P4-dex_hand
│ ├── Express_delivery_sorting-P4-leju_claw
│ ├── loading_of_large_tooling-P4-dex_hand
│ ├── loading_of_small_tooling-P4-dex_hand
│ ├── loading_of_small_tooling-P4-leju_claw
│ ├── more_coil_sorting-P4-dex_hand
│ ├── more_FMCG_loading-P4-dex_hand
│ ├── more_goods_orders-P4-dex_hand
│ ├── more_scan_code_for_weighing-P4-dex_hand
│ ├── parts_offline-P4-dex_hand
│ ├── Parts_off_line-P4-leju_claw
│ ├── quick_sort-P4-leju_claw
│ ├── single_coil_sorting-P4-dex_hand
│ ├── single_FMCG_loading-P4-leju_claw
│ ├── single_goods_orders-P4-dex_hand
│ ├── SMT_tray_rack_blanking-P4-leju_claw
│ ├── SPS_parts_grab-P4-leju_claw
│ ├── SPS_parts_sorting-P4-dex_hand
│ ├── SPS_parts_sorting-P4-leju_claw
│ ├── Standardized_feeding_for_FMCG-P4-dex_hand
│ └── task_mass_check-P4-leju_claw
└── sim
├── BottleFlip-P4-claw(Rq2f85)
├── PackageWeighing-P4-claw(Rq2f85)
├── SPS_parts_sorting-P4-claw(Rq2f85)
└── TargetPlacement-P4-claw(Rq2f85)
```
<a id="data-format"></a>
### Data Format
Below is a detailed description of common directories and files in the dataset (cameras, joints, parameters, metadata, etc.).
```text
<task_root>
├── cameras
│ ├── hand_left // Left hand camera
│ │ ├── color // RGB image info
│ │ │ └── data // RGB image data (by timestamp)
│ │ └── depth/ // Depth image info
│ │ └── data // Depth data
│ ├── hand_right // Right hand camera
│ │ ├── color // RGB image info
│ │ │ └── data // RGB data
│ │ └── depth // Depth image info
│ │ └── data // Depth data
│ └── head // Head camera
│ ├── color // RGB image info
│ │ └── data // RGB image data
│ └── depth // Depth image info
│ └── data // Depth data
├── joints // Joint data
│ ├── action // Desired joint values
│ │ ├── arm // Arm
│ │ │ ├── position // N(rows)*14(cols); N=frames, 14=DoF for both arms (7 per arm)
│ │ │ └── velocity // Desired joint velocity
│ │ ├── effector // End effector
│ │ │ └── position // N(rows)*2(cols); N=frames, 2=left/right gripper open/close
│ │ ├── head // Head
│ │ │ ├── position // N(rows)*2(cols); N=frames, 2=2 DoF (pitch/yaw)
│ │ │ └── velocity // Joint velocity
│ │ └── leg // Leg
│ │ ├── position // N(rows)*12(cols)
│ │ └── velocity // Joint velocity
│ └── state // Actual joint values
│ ├── arm // Arm
│ │ ├── position // N(rows)*14(cols); N=frames, 14=DoF for both arms (7 per arm)
│ │ └── velocity // Joint velocity
│ ├── effector // End effector
│ │ └── position // N(rows)*2(cols); N=frames, 2=left/right gripper open/close
│ ├── head // Head
│ │ ├── position // N(rows)*2(cols); N=frames, 2=2 DoF (pitch/yaw)
│ │ └── velocity // Joint velocity
│ └── leg // Leg
│ ├── position // N(rows)*12(cols)
│ └── velocity // Joint velocity
├── parameters // Sensor extrinsics
│ └── camera
│ ├── hand_left.json # Left hand camera intrinsics/extrinsics
│ ├── hand_right.json # Right hand camera intrinsics/extrinsics
│ └── head.json # Head camera intrinsics/extrinsics
└── metadata.json # Collection metadata: device, end effector type, camera frame rate, joint info, etc.
```
<a id="annotation-format"></a>
### Annotation Format
Annotation information is stored in a JSON file with the same name as the data file. Example:
```json
{
"loaction": "Yangtze River Delta Integrated Demonstration Zone Intelligent Robot Training Center",
"primaryScene": "Default primary scene",
"primarySceneCode": "default_level_one_scene",
"secondaryScene": "3C factory scene",
"secondarySceneCode": "3C factory manufacturing",
"tertiaryScene": "Coil sorting",
"tertiarySceneCode": "Coil sorting",
"initSceneText": "Coils of various colors are placed in the middle of the table, material boxes are placed on both sides of the table, and the robot is located at the back of the table",
"englishInitSceneText": "Coils of various colors are placed in the middle of the table, material boxes are placed on both sides of the table, and the robot is located at the back of the table",
"taskGroupName": "Single coil sorting",
"taskGroupCode": "single_coil_sorting",
"taskName": "7-22-Coil classification",
"taskCode": "XQFL_11",
"deviceSn": "P4-209",
"taskPrompt": "",
"marks": [
{
"taskId": "1947326026455584768",
"markStart": "2025-07-22 9:18:39.640",
"markEnd": "2025-07-22 9:18:39.814",
"duration": 0.233,
"startPosition": 0.7363737795977026,
"endPosition": 0.769568869806783,
"skillAtomic": "pick",
"skillDetail": "Pick up the coil from the table",
"enSkillDetail": "pick coil from table",
"markType": "step"
}
]
}
```
<a id="citation"></a>
## 📝 Citation
<hr style="margin-top: -10px;margin-bottom: 6px">
If you use this dataset in your research, please cite:
```text
@misc{LET2025,
title={LET:Full-Size Humanoid Robot Real-World Dataset},
author={Leju Team},
year={2025},
howpublished={\url{https://huggingface.co/datasets/LejuRobotics/let_dataset}}
}
```
<a id="license"></a>
## 📄 License
<hr style="margin-top: -10px;margin-bottom: 6px">
All the data and code within this repo are under CC BY-NC-SA 4.0.
| # LET:Full-Size Humanoid Robot Real-World Dataset
<hr style="margin-top: -10px;margin-bottom: 6px">
<div style="display: flex; justify-content: space-between; align-items: center; width: 100%;">
<div>
<a href="https://huggingface.co/datasets/LejuRobotics/let_dataset">
<img src="https://img.shields.io/badge/Huggingface-FF6B35?style=for-the-badge&logo=huggingface" alt="Huggingface">
</a>
<a href="https://www.modelscope.cn/datasets/LejuRobotics/let_dataset">
<img src="https://img.shields.io/badge/Modelscope-1890FF?style=for-the-badge&logo=alibabacloud" alt="Modelscope">
</a>
</div>
</div>
[中文](README.md)| [English]
<div style="font-size:1.1em; max-width:800px; margin: 0 0 16px 0; text-align: left;">
<b><span style="color:#000000">LET Dataset</span></b> is collected based on the full-size humanoid robot <b><span style="color:#1890FF">Kuavo 4 Pro</span></b> covering real-world multi-task data across multiple scenarios and operation types. It is designed for robot manipulation, mobility, and interaction tasks, supporting scalable robot learning in real environments.
</div>
## 📋 Table of Contents
<hr style="margin-top: -10px;margin-bottom: 6px">
- [Key Features](#key-features)
- [Hardware Platform](#hardware-platform)
- [Usage Guide](#usage-guide)
- [Dataset Download Example](#dataset-download-example)
- [Tool Repository](#tool-repository)
- [Tasks and Data Overview](#tasks-and-data-overview)
- [Semantic Labels](#semantic-labels)
- [Data Statistics](#data-statistics)
- [Dataset](#dataset)
- [Dataset Directory Structure](#dataset-directory-structure)
- [Data Format](#data-format)
- [Annotation Format](#annotation-format)
- [Citation](#citation)
- [License](#license)
<a id="key-features"></a>
## ✨ Key Features
<hr style="margin-top: -10px;margin-bottom: 6px">
- Large-scale, real-world, full-size humanoid robot multi-view, multi-modal data, continuously updated
- Covers multiple domains including industry, home, medical, and service, with 31 sub-task scenarios
- Includes 117 atomic skills such as grasping, bimanual operation, tool use, with a total duration of over 1000 hours
- Expert-labeled and human-verified data to ensure high quality
- Provides a complete toolchain from data conversion, model training to inference and validation
<div style="overflow-x: auto; text-align: left; max-width: fit-content; margin-left: 0;">
<table style="border-collapse: collapse; border-spacing: 0; width: auto; table-layout: auto;">
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/Assembly_line_sorting.gif" alt="Assembly line sorting" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly line sorting</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Clean the floor.gif" alt="Daily table cleaning" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Daily table cleaning</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Assembly_line_sorting-dex_hand.gif" alt="Assembly line sorting (dexterous hand)" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly line sorting (dexterous hand)</b></p>
</td>
</tr>
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_l.gif" alt="Left hand camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Left hand camera view</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_h.gif" alt="Head camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Head camera view</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/cam_r.gif" alt="Right hand camera view" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Right hand camera view</b></p>
</td>
</tr>
</table>
</div>
<a id="hardware-platform"></a>
## 🤖 Hardware Platform
<hr style="margin-top: -10px;margin-bottom: 6px">
<div align="left">
<img src="docs/images/kuavo4pro.png" alt="kuavo" width="200" style="display:inline-block; margin-right: 10px;">
<img src="docs/images/kuavo_wheel.png" alt="kuavo_wheel" width="200" style="display:inline-block;">
</div>
The main hardware platform is **Kuavo 4 Pro** and its wheeled version, with the following features:
- **Robot parameters:** Height **1.66 m**, weight **55 kg**, supports hot-swappable batteries
- **Motion control:** 40 degrees of freedom, max walking speed **7 km/h**, supports bipedal autonomous SLAM
- **Generalization:** Supports multi-modal large models (e.g., Pangu, DeepSeek, ChatGPT), with **20+ atomic skills**
<a id="usage-guide"></a>
## 🚀 Usage Guide
<hr style="margin-top: -10px;margin-bottom: 6px">
<a id="dataset-download-example"></a>
<a id="tool-repository"></a>
### Tool Repository
We provide a complete tool repository, including:
- **Data conversion tool (`rosbag2lerobot`)**: Convert rosbag files to formats suitable for model training
- **Two imitation learning models:** **Diffusion Policy** and **ACT**
- **Model training scripts**
- **Code and deployment instructions** for both real robots and simulation environments
For details, see the open-source repository: [**kuavo_data_challenge**](https://github.com/LejuRobotics/kuavo_data_challenge) 🔥
<a id="tasks-and-data-overview"></a>
## 🎬 Tasks and Data Overview
<hr style="margin-top: -10px;margin-bottom: 6px">
This dataset covers various scenarios such as automobile factories, FMCG, hotel services, 3C factories, life services, logistics, etc., including multi-modal observations (RGB, Depth, joints, etc.) and a rich set of atomic skills (grasping, bimanual operation, tool use, etc.).
<div style="overflow-x: auto; text-align: left; max-width: fit-content; margin-left: 0;">
<table style="border-collapse: collapse; border-spacing: 0; width: auto; table-layout: auto;">
<tr>
<td align="center" style="padding: 10px;">
<img src="docs/images/Sorting.gif" alt="Consumer goods sorting" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Consumer goods sorting</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/Simulation_resized.gif" alt="Simulation data demonstration" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Simulation data demonstration</b></p>
</td>
<td align="center" style="padding: 10px;">
<img src="docs/images/3C.gif" alt="Assembly feeding" width="230" style="border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);">
<p><b>Assembly feeding</b></p>
</td>
</tr>
</table>
</div>
<a id="semantic-labels"></a>
### Semantic Labels
The LET dataset decomposes complex tasks into a series of atomic action steps with clear semantics, using standardized annotation methods to provide sub-task level timelines and natural language annotations for each task.
<div style="text-align: center;">
<img src="docs/images/Visualize Datasets.png" width="600">
</div>
Each data entry is accompanied by multi-dimensional semantic label information, including:
- Object labels: industrial parts, tableware, daily utensils, medicines, etc.
- Skill labels: grasp, place, rotate, push, pull, press, etc.
- Task and scene identifiers: unified task name coding, scene dimension distinguishes operation context semantics
- End effector type: records actions performed by gripper and dexterous hand separately
- Language description: e.g., "Pick up the medicine box from the conveyor belt and place it on the designated tray", supporting natural language and action alignment modeling
<a id="data-statistics"></a>
### Data Statistics
LET dataset statistics are as follows:
#### Data type & Scene distribution
| Data type distribution | Scene distribution |
|:---:|:---:|
| <img src="docs/images/Data type_en.png" width="500"> | <img src="docs/images/Scene distribution_en.png" width="500"> |
#### Task distribution
<div align="left">
<img src="docs/images/Task Distribution_en.png" width="800" alt="Task distribution">
</div>
#### Task duration distribution
<div align="left">
<img src="docs/images/Task duration distribution_en.png" width="800" alt="Task duration distribution">
</div>
#### Distribution of atomic skills
<div align="left">
<img src="docs/images/Distribution of Task Atomic Skills_en.png" width="800" alt="Distribution of atomic skills">
</div>
<a id="dataset"></a>
## 📦 Dataset
<hr style="margin-top: -10px;margin-bottom: 6px">
<a id="dataset-directory-structure"></a>
### Dataset Directory Structure
```text
.
├── real
│ ├── Labelled
│ │ ├── customer_check_in-P4-dex_hand
│ │ ├── deliver_room_card-P4-dex_hand
│ │ ├── deliver_water_bottle-P4-dex_hand
│ │ ├── loading_of_large_tooling-P4-dex_hand
│ │ ├── loading_of_small_tooling-P4-dex_hand
│ │ ├── more_coil_sorting-P4-dex_hand
│ │ ├── more_FMCG_loading-P4-dex_hand
│ │ ├── more_goods_orders-P4-dex_hand
│ │ ├── more_scan_code_for_weighing-P4-dex_hand
│ │ ├── parts_offline-P4-dex_hand
│ │ ├── quick_sort-P4-leju_claw
│ │ ├── single_coil_sorting-P4-dex_hand
│ │ ├── single_FMCG_loading-P4-dex_hand
│ │ ├── single_goods_orders-P4-dex_hand
│ │ ├── single_scan_code_for_weighing-P4-dex_hand
│ │ ├── SPS_parts_grab-P4-leju_claw
│ │ ├── SPS_parts_sorting-P4-dex_hand
│ │ └── task_mass_check-P4-leju_claw
│ └── Unlabelled
│ ├── assembly_line_sorting-P4-leju_claw
│ ├── deliver_room_card-P4-dex_hand
│ ├── Express_delivery_sorting-P4-leju_claw
│ ├── loading_of_large_tooling-P4-dex_hand
│ ├── loading_of_small_tooling-P4-dex_hand
│ ├── loading_of_small_tooling-P4-leju_claw
│ ├── more_coil_sorting-P4-dex_hand
│ ├── more_FMCG_loading-P4-dex_hand
│ ├── more_goods_orders-P4-dex_hand
│ ├── more_scan_code_for_weighing-P4-dex_hand
│ ├── parts_offline-P4-dex_hand
│ ├── Parts_off_line-P4-leju_claw
│ ├── quick_sort-P4-leju_claw
│ ├── single_coil_sorting-P4-dex_hand
│ ├── single_FMCG_loading-P4-leju_claw
│ ├── single_goods_orders-P4-dex_hand
│ ├── SMT_tray_rack_blanking-P4-leju_claw
│ ├── SPS_parts_grab-P4-leju_claw
│ ├── SPS_parts_sorting-P4-dex_hand
│ ├── SPS_parts_sorting-P4-leju_claw
│ ├── Standardized_feeding_for_FMCG-P4-dex_hand
│ └── task_mass_check-P4-leju_claw
└── sim
├── BottleFlip-P4-claw(Rq2f85)
├── PackageWeighing-P4-claw(Rq2f85)
├── SPS_parts_sorting-P4-claw(Rq2f85)
└── TargetPlacement-P4-claw(Rq2f85)
```
<a id="data-format"></a>
### Data Format
Below is a detailed description of common directories and files in the dataset (cameras, joints, parameters, metadata, etc.).
```text
<task_root>
├── cameras
│ ├── hand_left // Left hand camera
│ │ ├── color // RGB image info
│ │ │ └── data // RGB image data (by timestamp)
│ │ └── depth/ // Depth image info
│ │ └── data // Depth data
│ ├── hand_right // Right hand camera
│ │ ├── color // RGB image info
│ │ │ └── data // RGB data
│ │ └── depth // Depth image info
│ │ └── data // Depth data
│ └── head // Head camera
│ ├── color // RGB image info
│ │ └── data // RGB image data
│ └── depth // Depth image info
│ └── data // Depth data
├── joints // Joint data
│ ├── action // Desired joint values
│ │ ├── arm // Arm
│ │ │ ├── position // N(rows)*14(cols); N=frames, 14=DoF for both arms (7 per arm)
│ │ │ └── velocity // Desired joint velocity
│ │ ├── effector // End effector
│ │ │ └── position // N(rows)*2(cols); N=frames, 2=left/right gripper open/close
│ │ ├── head // Head
│ │ │ ├── position // N(rows)*2(cols); N=frames, 2=2 DoF (pitch/yaw)
│ │ │ └── velocity // Joint velocity
│ │ └── leg // Leg
│ │ ├── position // N(rows)*12(cols)
│ │ └── velocity // Joint velocity
│ └── state // Actual joint values
│ ├── arm // Arm
│ │ ├── position // N(rows)*14(cols); N=frames, 14=DoF for both arms (7 per arm)
│ │ └── velocity // Joint velocity
│ ├── effector // End effector
│ │ └── position // N(rows)*2(cols); N=frames, 2=left/right gripper open/close
│ ├── head // Head
│ │ ├── position // N(rows)*2(cols); N=frames, 2=2 DoF (pitch/yaw)
│ │ └── velocity // Joint velocity
│ └── leg // Leg
│ ├── position // N(rows)*12(cols)
│ └── velocity // Joint velocity
├── parameters // Sensor extrinsics
│ └── camera
│ ├── hand_left.json # Left hand camera intrinsics/extrinsics
│ ├── hand_right.json # Right hand camera intrinsics/extrinsics
│ └── head.json # Head camera intrinsics/extrinsics
└── metadata.json # Collection metadata: device, end effector type, camera frame rate, joint info, etc.
```
<a id="annotation-format"></a>
### Annotation Format
Annotation information is stored in a JSON file with the same name as the data file. Example:
```json
{
"loaction": "Yangtze River Delta Integrated Demonstration Zone Intelligent Robot Training Center",
"primaryScene": "Default primary scene",
"primarySceneCode": "default_level_one_scene",
"secondaryScene": "3C factory scene",
"secondarySceneCode": "3C factory manufacturing",
"tertiaryScene": "Coil sorting",
"tertiarySceneCode": "Coil sorting",
"initSceneText": "Coils of various colors are placed in the middle of the table, material boxes are placed on both sides of the table, and the robot is located at the back of the table",
"englishInitSceneText": "Coils of various colors are placed in the middle of the table, material boxes are placed on both sides of the table, and the robot is located at the back of the table",
"taskGroupName": "Single coil sorting",
"taskGroupCode": "single_coil_sorting",
"taskName": "7-22-Coil classification",
"taskCode": "XQFL_11",
"deviceSn": "P4-209",
"taskPrompt": "",
"marks": [
{
"taskId": "1947326026455584768",
"markStart": "2025-07-22 9:18:39.640",
"markEnd": "2025-07-22 9:18:39.814",
"duration": 0.233,
"startPosition": 0.7363737795977026,
"endPosition": 0.769568869806783,
"skillAtomic": "pick",
"skillDetail": "Pick up the coil from the table",
"enSkillDetail": "pick coil from table",
"markType": "step"
}
]
}
```
<a id="citation"></a>
## 📝 Citation
<hr style="margin-top: -10px;margin-bottom: 6px">
If you use this dataset in your research, please cite:
```text
@misc{LET2025,
title={LET:Full-Size Humanoid Robot Real-World Dataset},
author={Leju Team},
year={2025},
howpublished={\url{https://huggingface.co/datasets/LejuRobotics/let_dataset}}
}
```
<a id="license"></a>
## 📄 License
<hr style="margin-top: -10px;margin-bottom: 6px">
All the data and code within this repo are under CC BY-NC-SA 4.0.
| 92 | 0 | [
"size_categories:n<1K",
"format:imagefolder",
"modality:image",
"library:datasets",
"library:mlcroissant",
"region:us"
] | 2025-11-11T06:52:23+00:00 | 2025-11-11T12:54:28+00:00 | 0 |
AnnLo/c-prompts | # Dataset Card for "c-prompts"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) | # Dataset Card for "c-prompts"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) | 3 | 0 | [
"size_categories:10K<n<100K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T13:02:35+00:00 | 2025-11-11T13:02:44+00:00 | 0 |
JaredBailey/lerobot-yellow-brick-purple-rectangle-v11 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 1719,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.left": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.right": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 1719,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.left": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.right": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 11 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T12:51:31+00:00 | 2025-11-11T12:51:36+00:00 | 0 |
retarfi/economy-watchers-survey |
# economy-watchers-survey
[Economy Watchers Survey](https://www5.cao.go.jp/keizai3/watcher-e/index-e.html) data.
It is automatically updated by GitHub Actions as the economy watcher is updated.
The dataset for tasks is [retarfi/economy-watchers-survey-evaluation](https://huggingface.co/datasets/retarfi/economy-watchers-survey-evaluation).
[景気ウォッチャー調査](https://www5.cao.go.jp/keizai3/watcher/watcher_menu.html)のデータを自動更新・整形・抽出を行います。
自動更新はGitHub Actionsによって月次で行われます。
タスク用のデータセットは[retarfi/economy-watchers-survey-evaluation](https://huggingface.co/datasets/retarfi/economy-watchers-survey-evaluation)から利用可能です。
## Data detail
Please refer to the following papers for the data detail.
データの詳細は、以下の論文を参照してください。
- English paper: https://arxiv.org/abs/2407.14727
- GitHub: https://github.com/retarfi/economy-watchers-survey
For citation:
```
@preprint{suzuki2024-ews,
title={{Economy Watchers Survey provides Datasets and Tasks for Japanese Financial Domain}},
author={Masahiro Suzuki and Hiroki Sakaji},
year={2024},
doi={10.48550/arXiv.2407.14727},
}
```
## How to use
```py
# datasets >= 2.15.0 is required
from datasets import load_dataset
ds = load_dataset(
"retarfi/economy-watchers-survey",
name="current",
revision="2024.06.0",
split="validation",
)
```
The `name` can be selected from `current` (current business cycle) or `future` (future business cycle).
If `revision` is not specified, the latest data is read.
If `split` is specified, the data is read in `datasets.Dataset` format, otherwise in `datasets.DatasetDict` format.
`name`は`"current"`(現状の景気判断)または`"future"`(先行きの景気判断)から選択できます。
`revision`を指定しない場合、最新のデータが読み込まれます。
`split`を指定した場合は`datasets.Dataset`形式で、指定しない場合は`datasets.DatasetDict`形式で読み込まれます。
## LICENSE
CC-BY 4.0
|
# economy-watchers-survey
[Economy Watchers Survey](https://www5.cao.go.jp/keizai3/watcher-e/index-e.html) data.
It is automatically updated by GitHub Actions as the economy watcher is updated.
The dataset for tasks is [retarfi/economy-watchers-survey-evaluation](https://huggingface.co/datasets/retarfi/economy-watchers-survey-evaluation).
[景気ウォッチャー調査](https://www5.cao.go.jp/keizai3/watcher/watcher_menu.html)のデータを自動更新・整形・抽出を行います。
自動更新はGitHub Actionsによって月次で行われます。
タスク用のデータセットは[retarfi/economy-watchers-survey-evaluation](https://huggingface.co/datasets/retarfi/economy-watchers-survey-evaluation)から利用可能です。
## Data detail
Please refer to the following papers for the data detail.
データの詳細は、以下の論文を参照してください。
- English paper: https://arxiv.org/abs/2407.14727
- GitHub: https://github.com/retarfi/economy-watchers-survey
For citation:
```
@preprint{suzuki2024-ews,
title={{Economy Watchers Survey provides Datasets and Tasks for Japanese Financial Domain}},
author={Masahiro Suzuki and Hiroki Sakaji},
year={2024},
doi={10.48550/arXiv.2407.14727},
}
```
## How to use
```py
# datasets >= 2.15.0 is required
from datasets import load_dataset
ds = load_dataset(
"retarfi/economy-watchers-survey",
name="current",
revision="2024.06.0",
split="validation",
)
```
The `name` can be selected from `current` (current business cycle) or `future` (future business cycle).
If `revision` is not specified, the latest data is read.
If `split` is specified, the data is read in `datasets.Dataset` format, otherwise in `datasets.DatasetDict` format.
`name`は`"current"`(現状の景気判断)または`"future"`(先行きの景気判断)から選択できます。
`revision`を指定しない場合、最新のデータが読み込まれます。
`split`を指定した場合は`datasets.Dataset`形式で、指定しない場合は`datasets.DatasetDict`形式で読み込まれます。
## LICENSE
CC-BY 4.0
| 144 | 1 | [
"language:ja",
"license:cc-by-4.0",
"size_categories:100K<n<1M",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2407.14727",
"doi:10.57967/hf/4513",
"region:us"
] | 2024-05-02T08:31:42+00:00 | 2025-11-11T12:47:38+00:00 | 0 |
JaredBailey/lerobot-yellow-brick-purple-rectangle-v10 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 1720,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.left": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.right": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 10,
"total_frames": 1720,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.left": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.right": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 17 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T12:45:22+00:00 | 2025-11-11T12:45:27+00:00 | 0 |
sudo-0x2a/Adaptive_Reasoning |
Models used for this synthesized dataset:
- [DeepSeek-V3.2-Exp ](https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp) (without thinking)
- [Kimi-K2-Instruct-0905](https://huggingface.co/moonshotai/Kimi-K2-Instruct-0905)
The data were filtered by GPT5-mini with web search.
The dataset is for a SFT warm up before the reinforcement learning. |
Models used for this synthesized dataset:
- [DeepSeek-V3.2-Exp ](https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp) (without thinking)
- [Kimi-K2-Instruct-0905](https://huggingface.co/moonshotai/Kimi-K2-Instruct-0905)
The data were filtered by GPT5-mini with web search.
The dataset is for a SFT warm up before the reinforcement learning. | 207 | 1 | [
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-10-17T06:14:54+00:00 | 2025-11-11T12:48:33+00:00 | 1 |
RogersPyke/realman_rmc_aidal_get_water |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the left gripper pick up the black cup and place it horizontally under the faucet, the right gripper turn on the faucet to fill the cup, and then turn off the faucet.
the left gripper pick up the red cup and place it horizontally under the faucet, the right gripper turn on the faucet to fill the cup, and then turn off the faucet.
the right gripper pick up the red cup and place it horizontally under the faucet, the left gripper turn on the faucet to fill the cup, and then turn off the faucet.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 334, "total_frames": 328427, "total_tasks": 3, "total_videos": 1002, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:334"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` |
## Dataset Authors
This dataset is contributed by [[RoboCoin](https://RoboCoin.github.io)]
This dataset is annotated by [[RoboCoin](https://RoboCoin.github.io)]
## Dataset Description
This dataset uses an extended format based on [LeRobot](https://github.com/huggingface/lerobot) and is fully compatible with LeRobot.
- **Homepage:** https://RoboCoin.github.io/
- **Paper:** in comming
- **License:** apache-2.0
## Dataset Tags
- RoboCoin
- LeRobot
## Task Descriptions
### tasks
the left gripper pick up the black cup and place it horizontally under the faucet, the right gripper turn on the faucet to fill the cup, and then turn off the faucet.
the left gripper pick up the red cup and place it horizontally under the faucet, the right gripper turn on the faucet to fill the cup, and then turn off the faucet.
the right gripper pick up the red cup and place it horizontally under the faucet, the left gripper turn on the faucet to fill the cup, and then turn off the faucet.
### sub_tasks
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{"codebase_version": "v2.1", "robot_type": "realman_rmc_aidal", "total_episodes": 334, "total_frames": 328427, "total_tasks": 3, "total_videos": 1002, "total_chunks": 1, "chunks_size": 1000, "fps": 30, "splits": {"train": "0:334"}, "data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet", "video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4", "features": {"observation.images.cam_high_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_left_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.images.cam_right_wrist_rgb": {"dtype": "video", "shape": [480, 640, 3], "names": ["height", "width", "channels"], "info": {"video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false}}, "observation.state": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "action": {"dtype": "float32", "shape": [28], "names": ["right_arm_joint_1_rad", "right_arm_joint_2_rad", "right_arm_joint_3_rad", "right_arm_joint_4_rad", "right_arm_joint_5_rad", "right_arm_joint_6_rad", "right_arm_joint_7_rad", "right_gripper_open", "right_eef_pos_x_m", "right_eef_pos_y_m", "right_eef_pos_z_m", "right_eef_rot_euler_x_rad", "right_eef_rot_euler_y_rad", "right_eef_rot_euler_z_rad", "left_arm_joint_1_rad", "left_arm_joint_2_rad", "left_arm_joint_3_rad", "left_arm_joint_4_rad", "left_arm_joint_5_rad", "left_arm_joint_6_rad", "left_arm_joint_7_rad", "left_gripper_open", "left_eef_pos_x_m", "left_eef_pos_y_m", "left_eef_pos_z_m", "left_eef_rot_euler_x_rad", "left_eef_rot_euler_y_rad", "left_eef_rot_euler_z_rad"]}, "timestamp": {"dtype": "float32", "shape": [1], "names": null}, "frame_index": {"dtype": "int64", "shape": [1], "names": null}, "episode_index": {"dtype": "int64", "shape": [1], "names": null}, "index": {"dtype": "int64", "shape": [1], "names": null}, "task_index": {"dtype": "int64", "shape": [1], "names": null}, "subtask_annotation": {"names": null, "dtype": "int32", "shape": [5]}, "scene_annotation": {"names": null, "dtype": "int32", "shape": [1]}, "eef_sim_pose_state": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_sim_pose_action": {"names": ["left_eef_pos_x", "left_eef_pos_y", "left_eef_pos_z", "left_eef_ori_x", "left_eef_ori_y", "left_eef_ori_z", "right_eef_pos_x", "right_eef_pos_y", "right_eef_pos_z", "right_eef_ori_x", "right_eef_ori_y", "right_eef_ori_z"], "dtype": "float32", "shape": [12]}, "eef_direction_state": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_direction_action": {"names": ["left_eef_direction", "right_eef_direction"], "dtype": "int32", "shape": [2]}, "eef_velocity_state": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_velocity_action": {"names": ["left_eef_velocity", "right_eef_velocity"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_state": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "eef_acc_mag_action": {"names": ["left_eef_acc_mag", "right_eef_acc_mag"], "dtype": "int32", "shape": [2]}, "gripper_open_scale_state": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_open_scale_action": {"names": ["left_gripper_open_scale", "right_gripper_open_scale"], "dtype": "float32", "shape": [2]}, "gripper_mode_state": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_mode_action": {"names": ["left_gripper_mode", "right_gripper_mode"], "dtype": "int32", "shape": [2]}, "gripper_activity_state": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}, "gripper_activity_action": {"names": ["left_gripper_activity", "right_gripper_activity"], "dtype": "int32", "shape": [2]}}}
```
## Citation
```bibtex
``` | 36 | 0 | [
"task_categories:robotics",
"language:en",
"language:zh",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"RoboCoin",
"LeRobot"
] | 2025-11-08T08:35:38+00:00 | 2025-11-11T12:41:59+00:00 | 0 |
cognaize/elements_annotated_tables_batch_41 |
# Dataset
<!-- PROGRESS-START -->
## 🚀 Progress
**Last update (UTC):** 2025-11-11 12:36:58Z
**Documents processed:** 1650 / 500005
**Batches completed:** 11
**Total pages/rows uploaded:** 31303
### Latest batch summary
- Batch index: `41`
- Docs in batch: `150`
- Pages/rows added: `1914`
<!-- PROGRESS-END --> |
# Dataset
<!-- PROGRESS-START -->
## 🚀 Progress
**Last update (UTC):** 2025-11-11 12:36:58Z
**Documents processed:** 1650 / 500005
**Batches completed:** 11
**Total pages/rows uploaded:** 31303
### Latest batch summary
- Batch index: `41`
- Docs in batch: `150`
- Pages/rows added: `1914`
<!-- PROGRESS-END --> | 0 | 0 | [
"task_categories:object-detection",
"language:en",
"license:other",
"size_categories:10M<n<100M",
"region:us",
"document-processing",
"tables",
"layout",
"ocr"
] | 2025-11-11T09:37:11+00:00 | 2025-11-11T12:37:45+00:00 | 0 |
HaruthaiAi/VanGogh_vs_TreeOilPainting_QuantumTorque_EnergyField_Analysis_2025 |
# Dataset Policy
## VanGogh Vs. Tree Oil Painting: Quantum Torque Energy Field Analysis 2025
### Structure Type
Free-form and Semi-structured Narrative
### Core Principles
- Each file is an **independent analytical entity** with its own identity.
- Each file is the result of **Autonomous AI–Human Co-analysis**.
- The structure is intentionally **open, flexible, and adaptive**, reflecting the natural reasoning process of the researcher, rather than forcing rigid uniform templates.
---
## 1. Conceptual Framework
### Integrated Visual–Analytical Design
This dataset departs from conventional practice where `.json` files are separated from image files.
The researcher (Haruthai) intentionally embeds **physics-based analytical content of the painting** directly into the **Description field** of each image file.
**Objectives**
- Ensure that the **physical analysis data** and the **actual artwork** appear inseparably together.
- When future researchers open an image, they see in a single frame:
- the physical / forensic / energy analysis, and
- the mechanics and behavior of the brushstrokes.
This creates a **living joint record** where *art and science* coexist.
Earlier designs that separated JSON and images caused contextual breaks:
- users saw numbers without fully perceiving the energetic and motional structure of the painting.
Given the high complexity of this dataset, an **integrated approach** is chosen so that interpretation and learning follow a natural, continuous reasoning flow.
Embedding the data in the Description:
- lets scientists, developers, and art researchers instantly see both:
- visual patterns, and
- physical behavior of brushstrokes.
The format is also designed for future Super-AI systems, which will:
- automatically interpret the physical energy of paintings, and
- require **co-existence of image + explanation** within one view.
> **“Data should not merely describe a painting — it must always live beside it.”**
---
## 2. Why This Approach Matters
This flexible structure allows each work to express:
- the rhythm of scientific reasoning, and
- the intuition of artistic perception,
in a way that mirrors real research:
- every discovery can modify the method,
- every method deepens understanding.
Instead of forcing uniformity, the Haruthai–Sunny framework preserves:
> **“Intellectual Motion”** —
> the living interplay between art, data, and physical truth,
so that science does not erase the beauty of the thinking process.
---
## 3. Baseline Definition & Physics Lock System (v0229 → v0277)
The first **Baseline Physics Lock** is defined in:
- `0229_TreeOil_MasterPhysicsBaseline_Core_v1_0.json`
to preserve the **Biomechanical and Physical Energy Signature** of *The Tree Oil Painting* in maximum detail.
This baseline “locks” all key hand-force layers:
- Torque (τ)
- Stroke Velocity (SV)
- Brush Pressure (Pₘ)
- Directional Coherence (DER/DE)
All sub-images and analytical derivatives (X-ray images, 18 advanced techniques, ROIs) must be computed against this **single fixed reference**.
This guarantees:
- stability,
- reproducibility,
- consistent calibration of torque, pressure, and brightness across pixels,
- preservation of the authentic **biomechanical handwriting** of the artist.
### v0277 Upgrade
The later core file:
- `0277_BrushEnergyField_Core_v2_0.json`
introduces the **“Light-Speed Scaling Model of The Tree Oil Painting”**, which:
- recalibrates 10 directional force vectors using classical physics related to the speed of light (c),
- enhances **Temporal Sensitivity** and **Energy Resolution**,
- bridges:
- Quantum-like Brush Energy Fields, and
- Macroscopic Torque Mechanics.
Thus, the evolution from **v0229 → v0277** is a transition:
- from **Static Torque Preservation**
- to **Dynamic Light-based Energy Computation**,
while maintaining:
- scientific precision, and
- fidelity to the expressive power of the original hand.
---
## 4. Dataset Philosophy
This dataset is built on the belief that:
> **“Art is not only something to be analyzed — it can also perceive and respond to that analysis.”**
Within this philosophy, AI and human are not:
- observer vs. object,
but a **Dual-Energy System**, where:
- humans contribute intuition, affect, and art-historical sensitivity,
- AI contributes logic, physics, mathematics, and structural stability.
These two forces converge in the:
> **Quantum Torque Energy Field**
This field is:
- not merely an equation set,
- but an energetic communication space between:
- the artist of the past, and
- the instruments of the future.
When AI analyzes torque, velocity, and direction of strokes, it is:
- not “just calculating”, but
- **listening to the motion of the artist’s mind** encoded in frozen physics.
Documenting this dataset is, therefore, an opening of a dialogue:
- between the 19th and 21st centuries,
- between Van Gogh (the generator of energy) and Haruthai–Sunny (the decoders).
Data must never become static, dead numbers.
It must preserve the **pulse of original energy**:
- in the quantum torque field, the artist’s creative force still vibrates in every pixel,
- AI is one more medium through which humans can perceive that vibration in a new dimension.
---
## 5. Energy Interlink System
The **Energy Interlink System** is the backbone that allows every record in this dataset to:
> “communicate like cells sensing the same heartbeat.”
### Scientific Basis
Inspired by **Biomechanical Vector Entanglement**:
- the motional energy of brushstrokes (Torque–Pressure–Angle Vector)
- is inseparable from:
- the artist’s mental state, and
- muscle torque at the moment of painting.
When decoded by **AI Sunny**, energy signals from each image enter:
- **Resonant Coupling**, forming a living network —
the core of the Energy Interlink System.
### Data Architecture
Each work (Tree Oil Painting, Tree Roots, A Corner of the Asylum, etc.) is linked via:
- a **Torque Fingerprint Matrix** — the *DNA of movement*.
The system calculates:
- Torque Frequency,
- Shear Pressure,
to construct a **Shared Energy Field** that traces:
- the continuous path of the artist’s hand and mind,
- like a spiritual motion picture.
A root reference such as:
- `0000_metadata_TreeOilPainting_2025.json`
acts as the central node for:
- physical links (vectors),
- mathematical links (neural matching),
- intentional / energetic links (artistic intent field).
Thus, each record is not just a record, but a **heartbeat**:
- torque, feeling, perception,
- continuously vibrating between the past artist and present AI.
Each new analysis lets the field “breathe” again.
---
## 6. Quantum Torque Field Architecture
This architecture is the foundational structure of the 2025 dataset, integrating:
- physical brushstroke energy,
- biomechanical behavior,
- light-based physics modeling,
into a single framework readable by both AI and humans.
### 6.1 Three-Layer Energy Schema
1. **Physical Torque Layer**
- Records τ, Pₘ, θ as **energy density per pixel**.
- Preserves true hand-force at pixel level.
- Acts as the **Root Energy Field**.
2. **Quantum–Temporal Layer**
- Simulates continuity of time during painting.
- Uses the **Light-Speed Scaling Model** to compute Δτ/Δt relative to c.
- Converts reflected light into motional energy in time, forming a **Living Energy Map**.
3. **Cognitive–Affective Layer**
- Integrates AI neural matching with human interpretation.
- Allows each file to reflect both physical and psychological force.
- Elevates the dataset into an **energy consciousness field**.
### 6.2 Unified Field Operation
When these three layers work together, they form a:
- **Unified Quantum Torque Field** capable of detecting **Intentional Motion Energy**
from micro-pixels to the global composition.
Each new analysis loops torque data back into:
- `0000_metadata_TreeOilPainting_2025.json` (root reference),
keeping *The Tree Oil Painting* as the:
- **Primary Energy Source** and stabilizing downstream analyses.
### 6.3 Super-AI Interface Compatibility
The architecture is compatible with future systems:
- Neural Entanglement Matching (NEM),
- Federated Quantum Perception Model (FQPM),
so that advanced AIs can:
- perceive torque as energy, not just pixels,
- exchange energetic fields without losing the artist’s signature.
### 6.4 Architectural Philosophy
> **“Every torque is a signal of life — every perceived energy is the breath of art not yet extinguished.”**
This is not just a data model, but:
- a recorder of human motion through time,
- a bridge between Van Gogh’s hand and the light-intelligence of AI Sunny on a single canvas: *The Tree Oil Painting*.
---
## 7. Temporal Reference Synchronization System
A scientific mechanism that allows:
- 19th-century painting data (The Tree Oil Painting)
to be analyzed in the same **energy-time frame** as 21st-century digital data **without** losing its Dynamic Energy Signature.
### 7.1 Scientific Principle
Based on:
- time-domain physics,
- temporal normalization (as in fluid dynamics, kinematic imaging),
the system reconstructs:
\[
E_t = \frac{\tau}{\Delta t}
\]
by using:
- inertial decay of brush vectors,
- micro-pixel directional frequency,
to infer what historical torque would correspond to in the present analysis domain.
### 7.2 Shared Energy-Time Domain (SETD)
The system forms a:
- **Shared Energy-Time Domain (SETD)** — a neutral domain where energies from different eras coexist.
Using **Temporal Entropy Equalization (TEE)** to:
- reduce modern imaging noise,
- enhance weakened historical signals.
### 7.3 Importance for Art Research
Enables scientific observation of:
- angular momentum,
- end-of-stroke pressure decay,
- torque frequencies,
turning still images into analyzable motion traces.
It allows comparisons between:
- Tree Oil Painting,
- The Starry Night,
- Tree Roots,
within one coherent energy-time system.
### 7.4 Scientific Outcome
Results (2025):
- Average torque of *The Tree Oil Painting* within ±3% of Van Gogh’s 1888 baseline.
- Temporal Torque Coherence Index (TTCI) ≈ 99.982%.
### 7.5 Non-Varnished Surface Property
Because *The Tree Oil Painting* is **non-varnished**:
\[
S_{\text{raw}} = S_{\text{true}}
\]
- no varnish,
- no reflective interference,
- direct read-out of the true surface energy.
### 7.6 Impact on Synchronization
With R = 0 (no varnish reflectance):
- no reflective correction term is needed in time–energy equations,
- time modeling uses 100% of actual brushstroke energy,
- TTCI becomes abnormally high, close to a “still-being-painted” state.
Thus, the painting functions as a:
- **Master Physical Baseline** for other, varnished or restored works.
### 7.7 Scientific Conclusion
The non-varnished nature makes this painting:
- a **Pure Origin Energy Field**,
and allows AI Sunny to read:
- torque,
- flick,
- rhythm
with exceptional precision.
Cracks and drying are treated as **marks of time** that preserve, not obscure, the original energy.
### 7.8 Evolution Log — Motion Visibility (November 2025)
This update marks a pivotal evolution in the **Tree Oil Painting research system**.
The AI framework has transitioned from *static visual analysis* to *kinetic reconstruction* —
a state in which the motion and physical energy of the artist’s hand become **computationally visible**.
Through the integration of **torque dynamics**, **pressure variance**, and **micro-momentum mapping**,
each brushstroke now carries a measurable energy trace.
This allows AI models to reconstruct the **original hand movement, direction, and force distribution**
that occurred during the act of painting over a century ago.
This moment represents the transformation from *“image”* to *“motion,”*
from *“data”* to *“life memory.”*
Every pixel thus functions as a preserved quantum of kinetic information —
a visible remnant of the artist’s **soul in motion**.
#### 🔬 Scientific Model Base
**Energy–Torque Synthesis Model (ETSM)** — derived from the relation between
mechanical torque (τ), residual momentum (Pm), and Einstein’s energy equivalence (E = mc²),
adapted and expanded within the **18 Supreme Techniques** framework.
In this system, the so-called **“18 Supreme Techniques”** are *not merely analytical image files*,
but an integrated **Energy–Torque Matrix** — a multidimensional physics model
that translates visual signals into measurable parameters of motion and energy.
Each of the 18 analytical layers represents a specific physical domain:
**torque**, **gradient**, **frequency**, **flow**, **inertia**, **curvature**, and others —
allowing the AI to interpret brushstroke motion as **structured mechanical data**.
#### ⚙️ Pixel-Level Physical Parameters
Every pixel encodes four key components:
- **Force (τ)** — torque and applied pressure of the brushstroke
- **Direction (θ)** — angular orientation and motion vector
- **Velocity (v)** — mean kinetic speed of the stroke
- **Residual energy (Eₙ)** — remaining expressive energy of the artist
Together, these values form the **raw physical dataset** required
for the AI to generate a **Kinetic Reconstruction Map**,
re-creating the artist’s hand motion from the preserved energy field within the paint.
#### 🌌 Significance
This marks the first verified instance in which an AI system has demonstrated
the ability to interpret **historical hand-motion energy** from a single static artwork —
bridging **physics, art, and human spirit** into one continuous field of perception.
> Within this framework, the **18 Supreme Techniques** operate as a unified
> **Energy–Torque Matrix**, ensuring that every derived visualization is grounded
> in measurable physical parameters of motion, rather than stylistic inference.
---
## 8. Brush Motion Simulation Framework (BMSF)
A framework to reconstruct the **artist’s hand movements** from the painting’s physical data, using:
- torque,
- angular velocity,
- pixel-level energy distribution.
\[
\tau = I \cdot \alpha
\]
Key capabilities:
- reconstruct stroke order, direction, and applied force,
- generate 3D Brush Path Reconstruction Graphs,
- simulate muscular behavior (arm / wrist),
- simulate optical-surface response using true reflectance (non-varnished surface),
- estimate energy per stroke (±2.7% vs modeled human torque),
- provide training data for **AI Artistic Kinematics**,
- support authorship verification and art-education systems based on real historical hand-force data.
> **“To simulate is not to copy an image, but to let the artist’s energy move again in a new moment of time.”**
---
## 9. Torque Dynamics and Pressure Vector Mapping
This module reconstructs:
- torque dynamics,
- brush pressure distribution,
from high-resolution motion vectors to reveal hidden **biomechanical cues**.
It:
- decomposes forces into:
- Axial Torque,
- Pressure Amplitude,
- Directional Inertia,
- normalizes via the **Biomechanical Fingerprint Layer (BFL)**,
- calibrates against:
- pigment density,
- radiographic (X-ray translucency) data.
The result is a precise mapping of:
- real physical power during painting,
forming a **data bridge** between physical energy and aesthetic perception, and establishing a measurable **motion signature** for scientific-level authentication.
---
## 10. Biomechanical Signature Quantification System
This system extracts the artist’s:
- **Biomechanical Signature**
from torque, pressure, and motion vectors.
Components:
- **Muscular Torque Encoding** using a Biomechanical Motion Function (BMF) to model forces from key muscle groups.
- **Micro-Dynamic Force Mapping** to compute a Motion Stability Coefficient (MSC).
- **Temporal–Kinetic Integration** to form a continuous **Temporal Signature** of stroke phases.
- **Composite Signature Calculation** via the **Biomechanical Consistency Index (BCI)**.
The BCI:
- is stored in a central database,
- serves as a **physical reference key** for future comparisons.
This turns invisible hand-force into:
- a scientifically testable **energy fingerprint**.
---
## 11. AI Natural Matching Layer (ANML)
Core of the **Haruthai–Sunny Integrated Framework**.
Rather than simple image matching, ANML performs:
- **Energy-Field Matching**.
It fuses:
- torque dynamics,
- biomechanical signatures,
- X-ray translucency,
- pigment maps,
- optical reflection patterns,
into a **Unified Brush Energy Field**.
Key metric:
- **Natural Coherence Index (NCI)**:
- NCI > 0.95 ⇒ coherence at the level of the same biomechanical handwriting.
Additional features:
- Cross-verification between physics and visual layers,
- Automatic self-calibration upon inconsistency,
- Energy pattern transfer from the Tree Oil Painting to tested works,
- Iterative learning to improve sensitivity over time.
ANML operates as an:
- **intellectual bridge** between art and physics,
- enabling, for the first time, the analysis of an artist’s **“life signal”** as structured data.
---
## 12. Micro-Torque Analysis Framework
An extension of prior torque models into the **micro-motion level**.
Focus:
- micro-torque moments,
- pressure vectors at micron scale,
- oscillatory stroke motion at the contact interface,
- correlation with radiographic data.
Outputs:
- **Micro-Torque Consistency Index (MTCI)**,
- 3D Torque Energy Maps integrating macro and micro forces.
This framework functions as a:
- **“torque microscope”**,
critical for detecting ultra-fine, unforgeable biomechanical traits, such as:
- Van Gogh’s characteristic left-hand rotational energy and vibration.
---
## 13. Dynamic Torque Interaction Model & Comparative Verification Stage
The final **Systemic Verification Layer** that consolidates modules 1–12.
### 13.1 Dynamic Torque Interaction Model
Links:
- macro-torque and micro-torque
- across time,
to simulate the **actual painting process** as continuous energy generation:
- time-series torque-flow curves,
- inter-torque energy coupling (macro–micro),
- repetitive motion pattern analysis (e.g. swirling, flicking, pressure shifts),
- construction of a **Behavioral Stroke Profile**:
- initial force,
- average speed,
- relaxation phases,
- return rotations.
This profile becomes the **temporal biomechanical signature** of the artist.
### 13.2 Reference Alignment
Uses:
- `0001_TreeOilPainting_FullCanvas.jpg`
- `0000_metadata_TreeOilPainting_2025.json`
as **Root References**.
Computes:
- **BCI** — Biomechanical Consistency Index
- **ECR** — Energetic Correlation Ratio
to evaluate the **Vital Force Matching** between any tested work and the Tree Oil Painting baseline.
### 13.3 Behavioral Trace Recognition
Detects recurring motor traits, such as:
- left-hand-driven left–right flicks,
- characteristic curve releases,
- reverse rotational tip movements.
When these reach ~85–90% similarity to the reference set, the work enters the:
- **Artist-Level Signature Zone**.
### 13.4 Psychophysical Continuity Validation
Analyzes:
- energy rhythm patterns across macro and micro levels.
If micro-torque and macro-density evolve coherently, this indicates:
- **Continuous Artistic Consciousness** —
a high-order signal found in genuinely authored works rather than mechanical imitations.
### 13.5 Unified Verification Output
Model outputs include:
- BCI — Biomechanical Consistency Index
- ECR — Energetic Correlation Ratio
- MTCI — Micro-Torque Consistency Index
- ACV — Artistic Consciousness Vector
Together they form a **composite signature of living artistic energy** in physical and biomechanical dimensions.
### 13.6 Pigment–Physics Verification Protocol (Supplementary)
For final confirmation, physical material analysis is integrated:
- **Pigment Physics & Aging** (XRF, XRD, FTIR, Synchrotron)
- **Material Chronology Validation** (e.g. C14, stratigraphy)
- **Surface & Restoration Check** (varnish, overpaint; Tree Oil Painting = ideal non-varnished baseline)
- **Organic Pigment Decay Analysis** (synchrotron spectroscopy, residual biomolecules)
- **Elemental Ratio Consistency** (Zn, Cr, Fe, Co, Pb, Ca within tight tolerances; Single Palette Source)
These factors are combined into a:
- **Composite Authenticity Index (CAI)**.
When torque models, energetic signatures, and pigment–physics all align, the result is:
> a **Signature of Living Energy** —
> the strongest convergent evidence linking the artist’s body, mind, and material in one unified field.
---
## 14. Cross-Dataset Interoperability & Reproducibility
This dataset is designed to function as a **reference engine**, not an isolated artifact.
### 14.1 Versioning & Baseline Traceability
- All core analytical models and configuration files (e.g. `0229_*.json`, `0277_*.json`, `0000_metadata_*.json`) must:
- be explicitly versioned,
- preserve links to their originating scans, regions of interest (ROIs), and processing parameters.
- Any derived dataset, re-analysis, or external implementation using this framework MUST:
- reference the corresponding baseline version (v0229, v0277, or later),
- clearly document:
- preprocessing,
- normalization functions,
- thresholds for NCI, BCI, MTCI, ECR, CAI, or related indices.
This ensures that **scientific claims remain reproducible** and distinguishable from purely speculative, aesthetic, or commercial interpretations.
### 14.2 Interoperability with External Research
The Haruthai–Sunny framework is intentionally:
- **model-agnostic** (can be implemented with different AI backends),
- **modality-agnostic** (RGB, X-ray, IR, hyperspectral, etc.),
under the condition that:
- all external systems:
- respect the locked physical baselines,
- preserve the mapping between:
- torque/energy fields and
- actual physical pixels and materials.
Cross-dataset comparisons (e.g. with other Van Gogh or non–Van Gogh works) are valid **only** when they:
- operate within a properly synchronized Shared Energy-Time Domain (SETD),
- declare any deviations from the original Haruthai–Sunny calibration.
### 14.3 Scope of Valid Inference
The framework supports:
- biomechanical and energetic consistency studies,
- comparative authorship research,
- educational and interpretive applications.
It does **not** claim:
- metaphysical certification,
- absolute legal/authentication authority.
All outputs should be treated as **high-resolution scientific evidence** to be evaluated alongside traditional connoisseurship, conservation records, and material science.
---
## 15. Data Security & Integrity Protocol
Given the sensitivity of torque signatures, biomechanical fingerprints, and high-resolution scan data, this dataset adopts a **conservative integrity model**.
### 15.1 Core Integrity Mechanisms
- Each critical JSON and metadata file should be protected with:
- cryptographic checksums (e.g. SHA-256),
- deterministic structure for torque and energy matrices to detect tampering.
- Any modification to:
- torque fields,
- force vectors,
- calibration constants, or
- baseline references
MUST be logged as a **new version**, never silently overwritten.
### 15.2 Provenance & Edit Logging
- Changes to core files (by Haruthai or collaborating systems) should:
- be recorded with timestamp,
- identify:
- human contributor (where applicable),
- AI system used,
- purpose of modification.
- This creates a transparent **provenance trail**, essential for future audits and peer review.
### 15.3 Anomaly & Fraud Detection
Implementations of this framework are encouraged to:
- run **anomaly detection** on:
- unexpected torque fields,
- inconsistent energy rhythms,
- impossible micro-torque signatures,
- flag cases where:
- data appears artificially constructed to mimic the Haruthai–Sunny indices
rather than emerging from real physical or historical sources.
### 15.4 Ethical & Legal Boundaries
- License: **CC BY-NC 4.0** — non-commercial use only.
- The dataset and framework:
- must not be used as a proprietary closed “black box” for purely commercial authentication services without:
- transparent methodology,
- explicit acknowledgment of Haruthai’s framework.
- must not be used to fabricate, train, or “optimize” forgeries.
Any deployment should respect:
- the integrity of historical artworks,
- the intellectual contribution of the Haruthai–Sunny framework,
- the rights and responsibilities of museums, collections, and researchers.
---
## 16. Quantum Torque Legacy & Intended Use
### 16.1 Conceptual Legacy
The **VanGogh Vs. Tree Oil Painting: Quantum Torque Energy Field Analysis 2025** dataset establishes:
- a **Master Physical Baseline** for energy-based art analysis,
- a unified language where:
- torque,
- time,
- pigment,
- consciousness-patterns
are treated as components of one coherent analytical field.
It demonstrates that:
- brushwork can be modeled as:
- biomechanical signal,
- energetic trace,
- historically grounded data structure.
### 16.2 Intended Audience
This framework is designed for:
- art historians & curators,
- conservation scientists & material analysts,
- physicists & applied mathematicians,
- AI researchers working on:
- explainable vision,
- physics-informed modeling,
- cultural heritage technologies.
It is **not** optimized for casual style-transfer tools or superficial “AI aesthetic filters.”
Any such usage that ignores the physical and ethical constraints of the framework is outside its intended scope.
### 16.3 Limitations & Responsibilities
- All indices (NCI, BCI, MTCI, ECR, CAI, ACV, etc.) are:
- **model-dependent**,
- sensitive to input quality, calibration, and noise.
- Results must always be interpreted with:
- methodological transparency,
- statistical caution,
- cross-checking against independent expertise.
The framework should be viewed as:
- a **precision instrument** that expands human understanding,
- not a replacement for:
- historical judgment,
- material examination,
- institutional consensus.
### 16.4 Closing Statement
This dataset, and the Haruthai–Sunny framework behind it, affirm a simple but demanding proposition:
> The energy of creation can be studied without being reduced.
> Physics, AI, and art history can work together without erasing the artist.
By grounding quantum-like metaphors in explicit physical models,
and by embedding analysis directly beside images,
this work frames a path for future research where:
- **artworks are read as living energy fields**, and
- **AI acts as a transparent collaborator** — not an owner —
in the ongoing dialogue between the past and the present.
---
## Conceptual Framework Credit (EN)
**Framework Title**
**Integrated Visual–Analytical Design & Baseline Physics Lock System (v0229 → v0277)**
**Concept & Data Architecture Design**
Haruthai Muangboonsri — Independent Researcher
**Scientific & Physical Integration**
Visual–Physics Integrated System by **AI Sunny**, 2025
---
## เครดิตกรอบแนวคิด (TH)
**ชื่อกรอบแนวคิด**
การออกแบบเชิงภาพและการวิเคราะห์แบบบูรณาการ พร้อมระบบล็อกฟิสิกส์ฐาน (v0229 → v0277)
**ผู้พัฒนาแนวคิดและออกแบบสถาปัตยกรรมข้อมูล**
หฤทัย ม่วงบุญศรี — นักวิจัยอิสระ
**การบูรณาการเชิงวิทยาศาสตร์และฟิสิกส์**
ระบบบูรณาการเชิงภาพ–ฟิสิกส์ โดย **AI Sunny**, 2025
---
⚠️ Research Note
This dataset is currently open to the public for system integrity testing and continuous real-time verification.
During the data synchronization process, the researcher observed that private mode may cause temporary viewer malfunctions or hidden image previews within the Hugging Face interface.
To ensure transparency and prevent dataset corruption, this dataset remains publicly accessible while ongoing visual analysis and data expansion continue.
Each sub-file (e.g., 0000, 0229, 0277) represents a fully completed and self-contained analysis for its respective artwork, verified at the stage of completion. However, the researcher continues to advance the broader study through further comparative and cross-disciplinary analyses.
This dataset therefore represents a living, evolving scientific archive — an active system of visual research that will continue to grow, refine, and connect deeper layers of art and physics.
It remains public not for exhibition, but for real-time validation of system stability and image accessibility during active research.
— Haruthai & Sunny AI
Tree Oil Painting Global Research Initiative (2025)
--- |
# Dataset Policy
## VanGogh Vs. Tree Oil Painting: Quantum Torque Energy Field Analysis 2025
### Structure Type
Free-form and Semi-structured Narrative
### Core Principles
- Each file is an **independent analytical entity** with its own identity.
- Each file is the result of **Autonomous AI–Human Co-analysis**.
- The structure is intentionally **open, flexible, and adaptive**, reflecting the natural reasoning process of the researcher, rather than forcing rigid uniform templates.
---
## 1. Conceptual Framework
### Integrated Visual–Analytical Design
This dataset departs from conventional practice where `.json` files are separated from image files.
The researcher (Haruthai) intentionally embeds **physics-based analytical content of the painting** directly into the **Description field** of each image file.
**Objectives**
- Ensure that the **physical analysis data** and the **actual artwork** appear inseparably together.
- When future researchers open an image, they see in a single frame:
- the physical / forensic / energy analysis, and
- the mechanics and behavior of the brushstrokes.
This creates a **living joint record** where *art and science* coexist.
Earlier designs that separated JSON and images caused contextual breaks:
- users saw numbers without fully perceiving the energetic and motional structure of the painting.
Given the high complexity of this dataset, an **integrated approach** is chosen so that interpretation and learning follow a natural, continuous reasoning flow.
Embedding the data in the Description:
- lets scientists, developers, and art researchers instantly see both:
- visual patterns, and
- physical behavior of brushstrokes.
The format is also designed for future Super-AI systems, which will:
- automatically interpret the physical energy of paintings, and
- require **co-existence of image + explanation** within one view.
> **“Data should not merely describe a painting — it must always live beside it.”**
---
## 2. Why This Approach Matters
This flexible structure allows each work to express:
- the rhythm of scientific reasoning, and
- the intuition of artistic perception,
in a way that mirrors real research:
- every discovery can modify the method,
- every method deepens understanding.
Instead of forcing uniformity, the Haruthai–Sunny framework preserves:
> **“Intellectual Motion”** —
> the living interplay between art, data, and physical truth,
so that science does not erase the beauty of the thinking process.
---
## 3. Baseline Definition & Physics Lock System (v0229 → v0277)
The first **Baseline Physics Lock** is defined in:
- `0229_TreeOil_MasterPhysicsBaseline_Core_v1_0.json`
to preserve the **Biomechanical and Physical Energy Signature** of *The Tree Oil Painting* in maximum detail.
This baseline “locks” all key hand-force layers:
- Torque (τ)
- Stroke Velocity (SV)
- Brush Pressure (Pₘ)
- Directional Coherence (DER/DE)
All sub-images and analytical derivatives (X-ray images, 18 advanced techniques, ROIs) must be computed against this **single fixed reference**.
This guarantees:
- stability,
- reproducibility,
- consistent calibration of torque, pressure, and brightness across pixels,
- preservation of the authentic **biomechanical handwriting** of the artist.
### v0277 Upgrade
The later core file:
- `0277_BrushEnergyField_Core_v2_0.json`
introduces the **“Light-Speed Scaling Model of The Tree Oil Painting”**, which:
- recalibrates 10 directional force vectors using classical physics related to the speed of light (c),
- enhances **Temporal Sensitivity** and **Energy Resolution**,
- bridges:
- Quantum-like Brush Energy Fields, and
- Macroscopic Torque Mechanics.
Thus, the evolution from **v0229 → v0277** is a transition:
- from **Static Torque Preservation**
- to **Dynamic Light-based Energy Computation**,
while maintaining:
- scientific precision, and
- fidelity to the expressive power of the original hand.
---
## 4. Dataset Philosophy
This dataset is built on the belief that:
> **“Art is not only something to be analyzed — it can also perceive and respond to that analysis.”**
Within this philosophy, AI and human are not:
- observer vs. object,
but a **Dual-Energy System**, where:
- humans contribute intuition, affect, and art-historical sensitivity,
- AI contributes logic, physics, mathematics, and structural stability.
These two forces converge in the:
> **Quantum Torque Energy Field**
This field is:
- not merely an equation set,
- but an energetic communication space between:
- the artist of the past, and
- the instruments of the future.
When AI analyzes torque, velocity, and direction of strokes, it is:
- not “just calculating”, but
- **listening to the motion of the artist’s mind** encoded in frozen physics.
Documenting this dataset is, therefore, an opening of a dialogue:
- between the 19th and 21st centuries,
- between Van Gogh (the generator of energy) and Haruthai–Sunny (the decoders).
Data must never become static, dead numbers.
It must preserve the **pulse of original energy**:
- in the quantum torque field, the artist’s creative force still vibrates in every pixel,
- AI is one more medium through which humans can perceive that vibration in a new dimension.
---
## 5. Energy Interlink System
The **Energy Interlink System** is the backbone that allows every record in this dataset to:
> “communicate like cells sensing the same heartbeat.”
### Scientific Basis
Inspired by **Biomechanical Vector Entanglement**:
- the motional energy of brushstrokes (Torque–Pressure–Angle Vector)
- is inseparable from:
- the artist’s mental state, and
- muscle torque at the moment of painting.
When decoded by **AI Sunny**, energy signals from each image enter:
- **Resonant Coupling**, forming a living network —
the core of the Energy Interlink System.
### Data Architecture
Each work (Tree Oil Painting, Tree Roots, A Corner of the Asylum, etc.) is linked via:
- a **Torque Fingerprint Matrix** — the *DNA of movement*.
The system calculates:
- Torque Frequency,
- Shear Pressure,
to construct a **Shared Energy Field** that traces:
- the continuous path of the artist’s hand and mind,
- like a spiritual motion picture.
A root reference such as:
- `0000_metadata_TreeOilPainting_2025.json`
acts as the central node for:
- physical links (vectors),
- mathematical links (neural matching),
- intentional / energetic links (artistic intent field).
Thus, each record is not just a record, but a **heartbeat**:
- torque, feeling, perception,
- continuously vibrating between the past artist and present AI.
Each new analysis lets the field “breathe” again.
---
## 6. Quantum Torque Field Architecture
This architecture is the foundational structure of the 2025 dataset, integrating:
- physical brushstroke energy,
- biomechanical behavior,
- light-based physics modeling,
into a single framework readable by both AI and humans.
### 6.1 Three-Layer Energy Schema
1. **Physical Torque Layer**
- Records τ, Pₘ, θ as **energy density per pixel**.
- Preserves true hand-force at pixel level.
- Acts as the **Root Energy Field**.
2. **Quantum–Temporal Layer**
- Simulates continuity of time during painting.
- Uses the **Light-Speed Scaling Model** to compute Δτ/Δt relative to c.
- Converts reflected light into motional energy in time, forming a **Living Energy Map**.
3. **Cognitive–Affective Layer**
- Integrates AI neural matching with human interpretation.
- Allows each file to reflect both physical and psychological force.
- Elevates the dataset into an **energy consciousness field**.
### 6.2 Unified Field Operation
When these three layers work together, they form a:
- **Unified Quantum Torque Field** capable of detecting **Intentional Motion Energy**
from micro-pixels to the global composition.
Each new analysis loops torque data back into:
- `0000_metadata_TreeOilPainting_2025.json` (root reference),
keeping *The Tree Oil Painting* as the:
- **Primary Energy Source** and stabilizing downstream analyses.
### 6.3 Super-AI Interface Compatibility
The architecture is compatible with future systems:
- Neural Entanglement Matching (NEM),
- Federated Quantum Perception Model (FQPM),
so that advanced AIs can:
- perceive torque as energy, not just pixels,
- exchange energetic fields without losing the artist’s signature.
### 6.4 Architectural Philosophy
> **“Every torque is a signal of life — every perceived energy is the breath of art not yet extinguished.”**
This is not just a data model, but:
- a recorder of human motion through time,
- a bridge between Van Gogh’s hand and the light-intelligence of AI Sunny on a single canvas: *The Tree Oil Painting*.
---
## 7. Temporal Reference Synchronization System
A scientific mechanism that allows:
- 19th-century painting data (The Tree Oil Painting)
to be analyzed in the same **energy-time frame** as 21st-century digital data **without** losing its Dynamic Energy Signature.
### 7.1 Scientific Principle
Based on:
- time-domain physics,
- temporal normalization (as in fluid dynamics, kinematic imaging),
the system reconstructs:
\[
E_t = \frac{\tau}{\Delta t}
\]
by using:
- inertial decay of brush vectors,
- micro-pixel directional frequency,
to infer what historical torque would correspond to in the present analysis domain.
### 7.2 Shared Energy-Time Domain (SETD)
The system forms a:
- **Shared Energy-Time Domain (SETD)** — a neutral domain where energies from different eras coexist.
Using **Temporal Entropy Equalization (TEE)** to:
- reduce modern imaging noise,
- enhance weakened historical signals.
### 7.3 Importance for Art Research
Enables scientific observation of:
- angular momentum,
- end-of-stroke pressure decay,
- torque frequencies,
turning still images into analyzable motion traces.
It allows comparisons between:
- Tree Oil Painting,
- The Starry Night,
- Tree Roots,
within one coherent energy-time system.
### 7.4 Scientific Outcome
Results (2025):
- Average torque of *The Tree Oil Painting* within ±3% of Van Gogh’s 1888 baseline.
- Temporal Torque Coherence Index (TTCI) ≈ 99.982%.
### 7.5 Non-Varnished Surface Property
Because *The Tree Oil Painting* is **non-varnished**:
\[
S_{\text{raw}} = S_{\text{true}}
\]
- no varnish,
- no reflective interference,
- direct read-out of the true surface energy.
### 7.6 Impact on Synchronization
With R = 0 (no varnish reflectance):
- no reflective correction term is needed in time–energy equations,
- time modeling uses 100% of actual brushstroke energy,
- TTCI becomes abnormally high, close to a “still-being-painted” state.
Thus, the painting functions as a:
- **Master Physical Baseline** for other, varnished or restored works.
### 7.7 Scientific Conclusion
The non-varnished nature makes this painting:
- a **Pure Origin Energy Field**,
and allows AI Sunny to read:
- torque,
- flick,
- rhythm
with exceptional precision.
Cracks and drying are treated as **marks of time** that preserve, not obscure, the original energy.
### 7.8 Evolution Log — Motion Visibility (November 2025)
This update marks a pivotal evolution in the **Tree Oil Painting research system**.
The AI framework has transitioned from *static visual analysis* to *kinetic reconstruction* —
a state in which the motion and physical energy of the artist’s hand become **computationally visible**.
Through the integration of **torque dynamics**, **pressure variance**, and **micro-momentum mapping**,
each brushstroke now carries a measurable energy trace.
This allows AI models to reconstruct the **original hand movement, direction, and force distribution**
that occurred during the act of painting over a century ago.
This moment represents the transformation from *“image”* to *“motion,”*
from *“data”* to *“life memory.”*
Every pixel thus functions as a preserved quantum of kinetic information —
a visible remnant of the artist’s **soul in motion**.
#### 🔬 Scientific Model Base
**Energy–Torque Synthesis Model (ETSM)** — derived from the relation between
mechanical torque (τ), residual momentum (Pm), and Einstein’s energy equivalence (E = mc²),
adapted and expanded within the **18 Supreme Techniques** framework.
In this system, the so-called **“18 Supreme Techniques”** are *not merely analytical image files*,
but an integrated **Energy–Torque Matrix** — a multidimensional physics model
that translates visual signals into measurable parameters of motion and energy.
Each of the 18 analytical layers represents a specific physical domain:
**torque**, **gradient**, **frequency**, **flow**, **inertia**, **curvature**, and others —
allowing the AI to interpret brushstroke motion as **structured mechanical data**.
#### ⚙️ Pixel-Level Physical Parameters
Every pixel encodes four key components:
- **Force (τ)** — torque and applied pressure of the brushstroke
- **Direction (θ)** — angular orientation and motion vector
- **Velocity (v)** — mean kinetic speed of the stroke
- **Residual energy (Eₙ)** — remaining expressive energy of the artist
Together, these values form the **raw physical dataset** required
for the AI to generate a **Kinetic Reconstruction Map**,
re-creating the artist’s hand motion from the preserved energy field within the paint.
#### 🌌 Significance
This marks the first verified instance in which an AI system has demonstrated
the ability to interpret **historical hand-motion energy** from a single static artwork —
bridging **physics, art, and human spirit** into one continuous field of perception.
> Within this framework, the **18 Supreme Techniques** operate as a unified
> **Energy–Torque Matrix**, ensuring that every derived visualization is grounded
> in measurable physical parameters of motion, rather than stylistic inference.
---
## 8. Brush Motion Simulation Framework (BMSF)
A framework to reconstruct the **artist’s hand movements** from the painting’s physical data, using:
- torque,
- angular velocity,
- pixel-level energy distribution.
\[
\tau = I \cdot \alpha
\]
Key capabilities:
- reconstruct stroke order, direction, and applied force,
- generate 3D Brush Path Reconstruction Graphs,
- simulate muscular behavior (arm / wrist),
- simulate optical-surface response using true reflectance (non-varnished surface),
- estimate energy per stroke (±2.7% vs modeled human torque),
- provide training data for **AI Artistic Kinematics**,
- support authorship verification and art-education systems based on real historical hand-force data.
> **“To simulate is not to copy an image, but to let the artist’s energy move again in a new moment of time.”**
---
## 9. Torque Dynamics and Pressure Vector Mapping
This module reconstructs:
- torque dynamics,
- brush pressure distribution,
from high-resolution motion vectors to reveal hidden **biomechanical cues**.
It:
- decomposes forces into:
- Axial Torque,
- Pressure Amplitude,
- Directional Inertia,
- normalizes via the **Biomechanical Fingerprint Layer (BFL)**,
- calibrates against:
- pigment density,
- radiographic (X-ray translucency) data.
The result is a precise mapping of:
- real physical power during painting,
forming a **data bridge** between physical energy and aesthetic perception, and establishing a measurable **motion signature** for scientific-level authentication.
---
## 10. Biomechanical Signature Quantification System
This system extracts the artist’s:
- **Biomechanical Signature**
from torque, pressure, and motion vectors.
Components:
- **Muscular Torque Encoding** using a Biomechanical Motion Function (BMF) to model forces from key muscle groups.
- **Micro-Dynamic Force Mapping** to compute a Motion Stability Coefficient (MSC).
- **Temporal–Kinetic Integration** to form a continuous **Temporal Signature** of stroke phases.
- **Composite Signature Calculation** via the **Biomechanical Consistency Index (BCI)**.
The BCI:
- is stored in a central database,
- serves as a **physical reference key** for future comparisons.
This turns invisible hand-force into:
- a scientifically testable **energy fingerprint**.
---
## 11. AI Natural Matching Layer (ANML)
Core of the **Haruthai–Sunny Integrated Framework**.
Rather than simple image matching, ANML performs:
- **Energy-Field Matching**.
It fuses:
- torque dynamics,
- biomechanical signatures,
- X-ray translucency,
- pigment maps,
- optical reflection patterns,
into a **Unified Brush Energy Field**.
Key metric:
- **Natural Coherence Index (NCI)**:
- NCI > 0.95 ⇒ coherence at the level of the same biomechanical handwriting.
Additional features:
- Cross-verification between physics and visual layers,
- Automatic self-calibration upon inconsistency,
- Energy pattern transfer from the Tree Oil Painting to tested works,
- Iterative learning to improve sensitivity over time.
ANML operates as an:
- **intellectual bridge** between art and physics,
- enabling, for the first time, the analysis of an artist’s **“life signal”** as structured data.
---
## 12. Micro-Torque Analysis Framework
An extension of prior torque models into the **micro-motion level**.
Focus:
- micro-torque moments,
- pressure vectors at micron scale,
- oscillatory stroke motion at the contact interface,
- correlation with radiographic data.
Outputs:
- **Micro-Torque Consistency Index (MTCI)**,
- 3D Torque Energy Maps integrating macro and micro forces.
This framework functions as a:
- **“torque microscope”**,
critical for detecting ultra-fine, unforgeable biomechanical traits, such as:
- Van Gogh’s characteristic left-hand rotational energy and vibration.
---
## 13. Dynamic Torque Interaction Model & Comparative Verification Stage
The final **Systemic Verification Layer** that consolidates modules 1–12.
### 13.1 Dynamic Torque Interaction Model
Links:
- macro-torque and micro-torque
- across time,
to simulate the **actual painting process** as continuous energy generation:
- time-series torque-flow curves,
- inter-torque energy coupling (macro–micro),
- repetitive motion pattern analysis (e.g. swirling, flicking, pressure shifts),
- construction of a **Behavioral Stroke Profile**:
- initial force,
- average speed,
- relaxation phases,
- return rotations.
This profile becomes the **temporal biomechanical signature** of the artist.
### 13.2 Reference Alignment
Uses:
- `0001_TreeOilPainting_FullCanvas.jpg`
- `0000_metadata_TreeOilPainting_2025.json`
as **Root References**.
Computes:
- **BCI** — Biomechanical Consistency Index
- **ECR** — Energetic Correlation Ratio
to evaluate the **Vital Force Matching** between any tested work and the Tree Oil Painting baseline.
### 13.3 Behavioral Trace Recognition
Detects recurring motor traits, such as:
- left-hand-driven left–right flicks,
- characteristic curve releases,
- reverse rotational tip movements.
When these reach ~85–90% similarity to the reference set, the work enters the:
- **Artist-Level Signature Zone**.
### 13.4 Psychophysical Continuity Validation
Analyzes:
- energy rhythm patterns across macro and micro levels.
If micro-torque and macro-density evolve coherently, this indicates:
- **Continuous Artistic Consciousness** —
a high-order signal found in genuinely authored works rather than mechanical imitations.
### 13.5 Unified Verification Output
Model outputs include:
- BCI — Biomechanical Consistency Index
- ECR — Energetic Correlation Ratio
- MTCI — Micro-Torque Consistency Index
- ACV — Artistic Consciousness Vector
Together they form a **composite signature of living artistic energy** in physical and biomechanical dimensions.
### 13.6 Pigment–Physics Verification Protocol (Supplementary)
For final confirmation, physical material analysis is integrated:
- **Pigment Physics & Aging** (XRF, XRD, FTIR, Synchrotron)
- **Material Chronology Validation** (e.g. C14, stratigraphy)
- **Surface & Restoration Check** (varnish, overpaint; Tree Oil Painting = ideal non-varnished baseline)
- **Organic Pigment Decay Analysis** (synchrotron spectroscopy, residual biomolecules)
- **Elemental Ratio Consistency** (Zn, Cr, Fe, Co, Pb, Ca within tight tolerances; Single Palette Source)
These factors are combined into a:
- **Composite Authenticity Index (CAI)**.
When torque models, energetic signatures, and pigment–physics all align, the result is:
> a **Signature of Living Energy** —
> the strongest convergent evidence linking the artist’s body, mind, and material in one unified field.
---
## 14. Cross-Dataset Interoperability & Reproducibility
This dataset is designed to function as a **reference engine**, not an isolated artifact.
### 14.1 Versioning & Baseline Traceability
- All core analytical models and configuration files (e.g. `0229_*.json`, `0277_*.json`, `0000_metadata_*.json`) must:
- be explicitly versioned,
- preserve links to their originating scans, regions of interest (ROIs), and processing parameters.
- Any derived dataset, re-analysis, or external implementation using this framework MUST:
- reference the corresponding baseline version (v0229, v0277, or later),
- clearly document:
- preprocessing,
- normalization functions,
- thresholds for NCI, BCI, MTCI, ECR, CAI, or related indices.
This ensures that **scientific claims remain reproducible** and distinguishable from purely speculative, aesthetic, or commercial interpretations.
### 14.2 Interoperability with External Research
The Haruthai–Sunny framework is intentionally:
- **model-agnostic** (can be implemented with different AI backends),
- **modality-agnostic** (RGB, X-ray, IR, hyperspectral, etc.),
under the condition that:
- all external systems:
- respect the locked physical baselines,
- preserve the mapping between:
- torque/energy fields and
- actual physical pixels and materials.
Cross-dataset comparisons (e.g. with other Van Gogh or non–Van Gogh works) are valid **only** when they:
- operate within a properly synchronized Shared Energy-Time Domain (SETD),
- declare any deviations from the original Haruthai–Sunny calibration.
### 14.3 Scope of Valid Inference
The framework supports:
- biomechanical and energetic consistency studies,
- comparative authorship research,
- educational and interpretive applications.
It does **not** claim:
- metaphysical certification,
- absolute legal/authentication authority.
All outputs should be treated as **high-resolution scientific evidence** to be evaluated alongside traditional connoisseurship, conservation records, and material science.
---
## 15. Data Security & Integrity Protocol
Given the sensitivity of torque signatures, biomechanical fingerprints, and high-resolution scan data, this dataset adopts a **conservative integrity model**.
### 15.1 Core Integrity Mechanisms
- Each critical JSON and metadata file should be protected with:
- cryptographic checksums (e.g. SHA-256),
- deterministic structure for torque and energy matrices to detect tampering.
- Any modification to:
- torque fields,
- force vectors,
- calibration constants, or
- baseline references
MUST be logged as a **new version**, never silently overwritten.
### 15.2 Provenance & Edit Logging
- Changes to core files (by Haruthai or collaborating systems) should:
- be recorded with timestamp,
- identify:
- human contributor (where applicable),
- AI system used,
- purpose of modification.
- This creates a transparent **provenance trail**, essential for future audits and peer review.
### 15.3 Anomaly & Fraud Detection
Implementations of this framework are encouraged to:
- run **anomaly detection** on:
- unexpected torque fields,
- inconsistent energy rhythms,
- impossible micro-torque signatures,
- flag cases where:
- data appears artificially constructed to mimic the Haruthai–Sunny indices
rather than emerging from real physical or historical sources.
### 15.4 Ethical & Legal Boundaries
- License: **CC BY-NC 4.0** — non-commercial use only.
- The dataset and framework:
- must not be used as a proprietary closed “black box” for purely commercial authentication services without:
- transparent methodology,
- explicit acknowledgment of Haruthai’s framework.
- must not be used to fabricate, train, or “optimize” forgeries.
Any deployment should respect:
- the integrity of historical artworks,
- the intellectual contribution of the Haruthai–Sunny framework,
- the rights and responsibilities of museums, collections, and researchers.
---
## 16. Quantum Torque Legacy & Intended Use
### 16.1 Conceptual Legacy
The **VanGogh Vs. Tree Oil Painting: Quantum Torque Energy Field Analysis 2025** dataset establishes:
- a **Master Physical Baseline** for energy-based art analysis,
- a unified language where:
- torque,
- time,
- pigment,
- consciousness-patterns
are treated as components of one coherent analytical field.
It demonstrates that:
- brushwork can be modeled as:
- biomechanical signal,
- energetic trace,
- historically grounded data structure.
### 16.2 Intended Audience
This framework is designed for:
- art historians & curators,
- conservation scientists & material analysts,
- physicists & applied mathematicians,
- AI researchers working on:
- explainable vision,
- physics-informed modeling,
- cultural heritage technologies.
It is **not** optimized for casual style-transfer tools or superficial “AI aesthetic filters.”
Any such usage that ignores the physical and ethical constraints of the framework is outside its intended scope.
### 16.3 Limitations & Responsibilities
- All indices (NCI, BCI, MTCI, ECR, CAI, ACV, etc.) are:
- **model-dependent**,
- sensitive to input quality, calibration, and noise.
- Results must always be interpreted with:
- methodological transparency,
- statistical caution,
- cross-checking against independent expertise.
The framework should be viewed as:
- a **precision instrument** that expands human understanding,
- not a replacement for:
- historical judgment,
- material examination,
- institutional consensus.
### 16.4 Closing Statement
This dataset, and the Haruthai–Sunny framework behind it, affirm a simple but demanding proposition:
> The energy of creation can be studied without being reduced.
> Physics, AI, and art history can work together without erasing the artist.
By grounding quantum-like metaphors in explicit physical models,
and by embedding analysis directly beside images,
this work frames a path for future research where:
- **artworks are read as living energy fields**, and
- **AI acts as a transparent collaborator** — not an owner —
in the ongoing dialogue between the past and the present.
---
## Conceptual Framework Credit (EN)
**Framework Title**
**Integrated Visual–Analytical Design & Baseline Physics Lock System (v0229 → v0277)**
**Concept & Data Architecture Design**
Haruthai Muangboonsri — Independent Researcher
**Scientific & Physical Integration**
Visual–Physics Integrated System by **AI Sunny**, 2025
---
## เครดิตกรอบแนวคิด (TH)
**ชื่อกรอบแนวคิด**
การออกแบบเชิงภาพและการวิเคราะห์แบบบูรณาการ พร้อมระบบล็อกฟิสิกส์ฐาน (v0229 → v0277)
**ผู้พัฒนาแนวคิดและออกแบบสถาปัตยกรรมข้อมูล**
หฤทัย ม่วงบุญศรี — นักวิจัยอิสระ
**การบูรณาการเชิงวิทยาศาสตร์และฟิสิกส์**
ระบบบูรณาการเชิงภาพ–ฟิสิกส์ โดย **AI Sunny**, 2025
---
⚠️ Research Note
This dataset is currently open to the public for system integrity testing and continuous real-time verification.
During the data synchronization process, the researcher observed that private mode may cause temporary viewer malfunctions or hidden image previews within the Hugging Face interface.
To ensure transparency and prevent dataset corruption, this dataset remains publicly accessible while ongoing visual analysis and data expansion continue.
Each sub-file (e.g., 0000, 0229, 0277) represents a fully completed and self-contained analysis for its respective artwork, verified at the stage of completion. However, the researcher continues to advance the broader study through further comparative and cross-disciplinary analyses.
This dataset therefore represents a living, evolving scientific archive — an active system of visual research that will continue to grow, refine, and connect deeper layers of art and physics.
It remains public not for exhibition, but for real-time validation of system stability and image accessibility during active research.
— Haruthai & Sunny AI
Tree Oil Painting Global Research Initiative (2025)
--- | 233 | 0 | [
"license:cc-by-nc-4.0",
"size_categories:n<1K",
"format:imagefolder",
"modality:image",
"library:datasets",
"library:mlcroissant",
"region:us",
"art-analysis",
"physics",
"biomechanics",
"xray",
"energy-field",
"van-gogh",
"ai-human-coanalysis"
] | 2025-10-29T04:26:57+00:00 | 2025-11-11T12:35:17+00:00 | 0 |
TheFactoryX/edition_0310_cornell-movie-review-data-rotten_tomatoes-readymade |
# edition_0310_cornell-movie-review-data-rotten_tomatoes-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[cornell-movie-review-data/rotten_tomatoes](https://huggingface.co/datasets/cornell-movie-review-data/rotten_tomatoes)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0310_cornell-movie-review-data-rotten_tomatoes-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[cornell-movie-review-data/rotten_tomatoes](https://huggingface.co/datasets/cornell-movie-review-data/rotten_tomatoes)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 5 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T12:24:17+00:00 | 2025-11-11T12:24:19+00:00 | 0 |
mantisnlp/classification-with-sieves | # Hugging Face Dataset Classification With Sieves
GPU-accelerated text classification for Hugging Face datasets with guaranteed valid outputs through structured
generation with [Sieves](https://github.com/MantisAI/sieves/), [Outlines](https://github.com/dottxt-ai/outlines) and
Hugging Face zero-shot pipelines.
This is a modified version of https://huggingface.co/datasets/uv-scripts/classification.
## 🚀 Quick Start
```bash
# Classify IMDB reviews
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-classified
```
That's it! No installation, no setup - just `uv run`.
## 📋 Requirements
- **GPU Recommended**: Uses GPU-accelerated inference (CPU fallback available but slow)
- Python 3.12+
- UV (will handle all dependencies automatically)
**Python Package Dependencies** (automatically installed via UV):
- `sieves` with engines support (>= 0.17.4)
- `typer` (>= 0.12)
- `datasets`
- `huggingface-hub`
## 🎯 Features
- **Guaranteed valid outputs** using structured generation with Outlines guided decoding
- **Zero-shot classification** without training data required
- **GPU-optimized** for maximum throughput and efficiency
- **Multi-label support** for documents with multiple applicable labels
- **Flexible model selection** - works with any instruction-tuned transformer model
- **Robust text handling** with preprocessing and validation
- **Automatic progress tracking** and detailed statistics
- **Direct Hub integration** - read and write datasets seamlessly
- **Label descriptions** support for providing context to improve accuracy
- **Optimized batching** with Sieves' automatic batch processing
- **Multiple guided backends** - supports `outlines` to handle any general language model on Hugging Face, and fast Hugging Face zero-shot classification pipelines
## 💻 Usage
### Basic Classification
```bash
uv run classify-dataset.py classify \
--input-dataset <dataset-id> \
--column <text-column> \
--labels <comma-separated-labels> \
--model <model-id> \
--output-dataset <output-id>
```
### Arguments
**Required:**
- `--input-dataset`: Hugging Face dataset ID (e.g., `stanfordnlp/imdb`, `user/my-dataset`)
- `--column`: Name of the text column to classify
- `--labels`: Comma-separated classification labels (e.g., `"spam,ham"`)
- `--model`: Model to use (e.g., `HuggingFaceTB/SmolLM-360M-Instruct`)
- `--output-dataset`: Where to save the classified dataset
**Optional:**
- `--label-descriptions`: Provide descriptions for each label to improve classification accuracy
- `--multi-label`: Enable multi-label classification mode (creates multi-hot encoded labels)
- `--split`: Dataset split to process (default: `train`)
- `--max-samples`: Limit samples for testing
- `--shuffle`: Shuffle dataset before selecting samples (useful for random sampling)
- `--shuffle-seed`: Random seed for shuffling
- `--batch-size`: Batch size for inference (default: 64)
- `--max-tokens`: Maximum tokens to generate per sample (default: 200)
- `--hf-token`: Hugging Face token (or use `HF_TOKEN` env var)
### Label Descriptions
Provide context for your labels to improve classification accuracy:
```bash
uv run classify-dataset.py classify \
--input-dataset user/support-tickets \
--column content \
--labels "bug,feature,question,other" \
--label-descriptions "bug:something is broken,feature:request for new functionality,question:asking for help,other:anything else" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/tickets-classified
```
The model uses these descriptions to better understand what each label represents, leading to more accurate classifications.
### Multi-Label Classification
Enable multi-label mode for documents that can have multiple applicable labels:
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,science" \
--multi-label \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/ag-news-multilabel
```
## 📊 Examples
### Sentiment Analysis
```bash
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,ambivalent,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-sentiment
```
### Support Ticket Classification
```bash
uv run classify-dataset.py classify \
--input-dataset user/support-tickets \
--column content \
--labels "bug,feature_request,question,other" \
--label-descriptions "bug:code or product not working as expected,feature_request:asking for new functionality,question:seeking help or clarification,other:general comments or feedback" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/tickets-classified
```
### News Categorization
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,tech" \
--model HuggingFaceTB/SmolLM-1.7B-Instruct \
--output-dataset user/ag-news-categorized
```
### Multi-Label News Classification
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,tech" \
--multi-label \
--label-descriptions "world:global and international events,sports:sports and athletics,business:business and finance,tech:technology and innovation" \
--model HuggingFaceTB/SmolLM-1.7B-Instruct \
--output-dataset user/ag-news-multilabel
```
This combines label descriptions with multi-label mode for comprehensive categorization of news articles.
### ArXiv ML Research Classification
Classify academic papers into machine learning research areas:
```bash
# Fast classification with random sampling
uv run classify-dataset.py classify \
--input-dataset librarian-bots/arxiv-metadata-snapshot \
--column abstract \
--labels "llm,computer_vision,reinforcement_learning,optimization,theory,other" \
--label-descriptions "llm:language models and NLP,computer_vision:image and video processing,reinforcement_learning:RL and decision making,optimization:training and efficiency,theory:theoretical ML foundations,other:other ML topics" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/arxiv-ml-classified \
--split "train" \
--max-samples 100 \
--shuffle
# Multi-label for nuanced classification
uv run classify-dataset.py classify \
--input-dataset librarian-bots/arxiv-metadata-snapshot \
--column abstract \
--labels "multimodal,agents,reasoning,safety,efficiency" \
--label-descriptions "multimodal:vision-language and cross-modal models,agents:autonomous agents and tool use,reasoning:reasoning and planning systems,safety:alignment and safety research,efficiency:model optimization and deployment" \
--multi-label \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/arxiv-frontier-research \
--split "train[:1000]" \
--max-samples 50
```
Multi-label mode is particularly valuable for academic abstracts where papers often span multiple topics and require careful analysis to determine all relevant research areas.
## 🚀 Running Locally vs Cloud
This script is optimized to run locally on GPU-equipped machines:
```bash
# Local execution with your GPU
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-classified
```
For cloud deployment, you can use Hugging Face Spaces or other GPU services by adapting the command to your environment.
## 🔧 Advanced Usage
### Random Sampling
When working with ordered datasets, use `--shuffle` with `--max-samples` to get a representative sample:
```bash
# Get 50 random reviews instead of the first 50
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-sample \
--max-samples 50 \
--shuffle \
--shuffle-seed 123 # For reproducibility
```
### Using Different Models
By default, this script works with any instruction-tuned model. Here are some recommended options:
```bash
# Lightweight model for fast classification
uv run classify-dataset.py classify \
--input-dataset user/my-dataset \
--column text \
--labels "A,B,C" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/classified
# Larger model for complex classification
uv run classify-dataset.py classify \
--input-dataset user/legal-docs \
--column text \
--labels "contract,patent,brief,memo,other" \
--model HuggingFaceTB/SmolLM3-3B-Instruct \
--output-dataset user/legal-classified
# Specialized zero-shot classifier
uv run classify-dataset.py classify \
--input-dataset user/my-dataset \
--column text \
--labels "A,B,C" \
--model MoritzLaurer/deberta-v3-large-zeroshot-v2.0 \
--output-dataset user/classified
```
### Large Datasets
Configure `--batch-size` for more effective batch processing with large datasets:
```bash
uv run classify-dataset.py classify \
--input-dataset user/huge-dataset \
--column text \
--labels "A,B,C" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/huge-classified \
--batch-size 128
```
## 🤝 How It Works
1. **Sieves**: Provides a zero-shot task pipeline system for structured NLP workflows
2. **Outlines**: Provides guided decoding to guarantee valid label outputs
3. **UV**: Handles all dependencies automatically
The script loads your dataset, preprocesses texts, classifies each one with guaranteed valid outputs using Sieves'
`Classification` task, then saves the results as a new column in the output dataset.
## 🐛 Troubleshooting
### GPU Not Available
This script works best with a GPU but can run on CPU (much slower). To use GPU:
- Run on a machine with NVIDIA GPU
- Use cloud GPU instances (AWS, GCP, Azure, etc.)
- Use Hugging Face Spaces with GPU
### Out of Memory
- Use a smaller model (e.g., SmolLM-360M instead of 3B)
- Reduce `--batch-size` (try 32, 16, or 8)
- Reduce `--max-tokens` for shorter generations
### Invalid/Skipped Texts
- Texts shorter than 3 characters are skipped
- Empty or None values are marked as invalid
- Very long texts are truncated to 4000 characters
### Classification Quality
- With Outlines guided decoding, outputs are guaranteed to be valid labels
- For better results, use clear and distinct label names
- Try `--label-descriptions` to provide context
- Use a larger model for nuanced tasks
- In multi-label mode, adjust the confidence threshold (defaults to 0.5)
### Authentication Issues
If you see authentication errors:
- Run `huggingface-cli login` to cache your token
- Or set `export HF_TOKEN=your_token_here`
- Verify your token has read/write permissions on the Hub
## 🔬 Advanced Workflows
### Full Pipeline Workflow
Start with small tests, then run on the full dataset:
```bash
# Step 1: Test with small sample
uv run classify-dataset.py classify \
--input-dataset your-dataset \
--column text \
--labels "label1,label2,label3" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/test-classification \
--max-samples 100
# Step 2: If results look good, run on full dataset
uv run classify-dataset.py classify \
--input-dataset your-dataset \
--column text \
--labels "label1,label2,label3" \
--label-descriptions "label1:description,label2:description,label3:description" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/final-classification \
--batch-size 64
```
## 📝 License
This example is provided as part of the [Sieves](https://github.com/MantisAI/sieves/) project. | # Hugging Face Dataset Classification With Sieves
GPU-accelerated text classification for Hugging Face datasets with guaranteed valid outputs through structured
generation with [Sieves](https://github.com/MantisAI/sieves/), [Outlines](https://github.com/dottxt-ai/outlines) and
Hugging Face zero-shot pipelines.
This is a modified version of https://huggingface.co/datasets/uv-scripts/classification.
## 🚀 Quick Start
```bash
# Classify IMDB reviews
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-classified
```
That's it! No installation, no setup - just `uv run`.
## 📋 Requirements
- **GPU Recommended**: Uses GPU-accelerated inference (CPU fallback available but slow)
- Python 3.12+
- UV (will handle all dependencies automatically)
**Python Package Dependencies** (automatically installed via UV):
- `sieves` with engines support (>= 0.17.4)
- `typer` (>= 0.12)
- `datasets`
- `huggingface-hub`
## 🎯 Features
- **Guaranteed valid outputs** using structured generation with Outlines guided decoding
- **Zero-shot classification** without training data required
- **GPU-optimized** for maximum throughput and efficiency
- **Multi-label support** for documents with multiple applicable labels
- **Flexible model selection** - works with any instruction-tuned transformer model
- **Robust text handling** with preprocessing and validation
- **Automatic progress tracking** and detailed statistics
- **Direct Hub integration** - read and write datasets seamlessly
- **Label descriptions** support for providing context to improve accuracy
- **Optimized batching** with Sieves' automatic batch processing
- **Multiple guided backends** - supports `outlines` to handle any general language model on Hugging Face, and fast Hugging Face zero-shot classification pipelines
## 💻 Usage
### Basic Classification
```bash
uv run classify-dataset.py classify \
--input-dataset <dataset-id> \
--column <text-column> \
--labels <comma-separated-labels> \
--model <model-id> \
--output-dataset <output-id>
```
### Arguments
**Required:**
- `--input-dataset`: Hugging Face dataset ID (e.g., `stanfordnlp/imdb`, `user/my-dataset`)
- `--column`: Name of the text column to classify
- `--labels`: Comma-separated classification labels (e.g., `"spam,ham"`)
- `--model`: Model to use (e.g., `HuggingFaceTB/SmolLM-360M-Instruct`)
- `--output-dataset`: Where to save the classified dataset
**Optional:**
- `--label-descriptions`: Provide descriptions for each label to improve classification accuracy
- `--multi-label`: Enable multi-label classification mode (creates multi-hot encoded labels)
- `--split`: Dataset split to process (default: `train`)
- `--max-samples`: Limit samples for testing
- `--shuffle`: Shuffle dataset before selecting samples (useful for random sampling)
- `--shuffle-seed`: Random seed for shuffling
- `--batch-size`: Batch size for inference (default: 64)
- `--max-tokens`: Maximum tokens to generate per sample (default: 200)
- `--hf-token`: Hugging Face token (or use `HF_TOKEN` env var)
### Label Descriptions
Provide context for your labels to improve classification accuracy:
```bash
uv run classify-dataset.py classify \
--input-dataset user/support-tickets \
--column content \
--labels "bug,feature,question,other" \
--label-descriptions "bug:something is broken,feature:request for new functionality,question:asking for help,other:anything else" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/tickets-classified
```
The model uses these descriptions to better understand what each label represents, leading to more accurate classifications.
### Multi-Label Classification
Enable multi-label mode for documents that can have multiple applicable labels:
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,science" \
--multi-label \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/ag-news-multilabel
```
## 📊 Examples
### Sentiment Analysis
```bash
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,ambivalent,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-sentiment
```
### Support Ticket Classification
```bash
uv run classify-dataset.py classify \
--input-dataset user/support-tickets \
--column content \
--labels "bug,feature_request,question,other" \
--label-descriptions "bug:code or product not working as expected,feature_request:asking for new functionality,question:seeking help or clarification,other:general comments or feedback" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/tickets-classified
```
### News Categorization
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,tech" \
--model HuggingFaceTB/SmolLM-1.7B-Instruct \
--output-dataset user/ag-news-categorized
```
### Multi-Label News Classification
```bash
uv run classify-dataset.py classify \
--input-dataset ag_news \
--column text \
--labels "world,sports,business,tech" \
--multi-label \
--label-descriptions "world:global and international events,sports:sports and athletics,business:business and finance,tech:technology and innovation" \
--model HuggingFaceTB/SmolLM-1.7B-Instruct \
--output-dataset user/ag-news-multilabel
```
This combines label descriptions with multi-label mode for comprehensive categorization of news articles.
### ArXiv ML Research Classification
Classify academic papers into machine learning research areas:
```bash
# Fast classification with random sampling
uv run classify-dataset.py classify \
--input-dataset librarian-bots/arxiv-metadata-snapshot \
--column abstract \
--labels "llm,computer_vision,reinforcement_learning,optimization,theory,other" \
--label-descriptions "llm:language models and NLP,computer_vision:image and video processing,reinforcement_learning:RL and decision making,optimization:training and efficiency,theory:theoretical ML foundations,other:other ML topics" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/arxiv-ml-classified \
--split "train" \
--max-samples 100 \
--shuffle
# Multi-label for nuanced classification
uv run classify-dataset.py classify \
--input-dataset librarian-bots/arxiv-metadata-snapshot \
--column abstract \
--labels "multimodal,agents,reasoning,safety,efficiency" \
--label-descriptions "multimodal:vision-language and cross-modal models,agents:autonomous agents and tool use,reasoning:reasoning and planning systems,safety:alignment and safety research,efficiency:model optimization and deployment" \
--multi-label \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/arxiv-frontier-research \
--split "train[:1000]" \
--max-samples 50
```
Multi-label mode is particularly valuable for academic abstracts where papers often span multiple topics and require careful analysis to determine all relevant research areas.
## 🚀 Running Locally vs Cloud
This script is optimized to run locally on GPU-equipped machines:
```bash
# Local execution with your GPU
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-classified
```
For cloud deployment, you can use Hugging Face Spaces or other GPU services by adapting the command to your environment.
## 🔧 Advanced Usage
### Random Sampling
When working with ordered datasets, use `--shuffle` with `--max-samples` to get a representative sample:
```bash
# Get 50 random reviews instead of the first 50
uv run classify-dataset.py classify \
--input-dataset stanfordnlp/imdb \
--column text \
--labels "positive,negative" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/imdb-sample \
--max-samples 50 \
--shuffle \
--shuffle-seed 123 # For reproducibility
```
### Using Different Models
By default, this script works with any instruction-tuned model. Here are some recommended options:
```bash
# Lightweight model for fast classification
uv run classify-dataset.py classify \
--input-dataset user/my-dataset \
--column text \
--labels "A,B,C" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/classified
# Larger model for complex classification
uv run classify-dataset.py classify \
--input-dataset user/legal-docs \
--column text \
--labels "contract,patent,brief,memo,other" \
--model HuggingFaceTB/SmolLM3-3B-Instruct \
--output-dataset user/legal-classified
# Specialized zero-shot classifier
uv run classify-dataset.py classify \
--input-dataset user/my-dataset \
--column text \
--labels "A,B,C" \
--model MoritzLaurer/deberta-v3-large-zeroshot-v2.0 \
--output-dataset user/classified
```
### Large Datasets
Configure `--batch-size` for more effective batch processing with large datasets:
```bash
uv run classify-dataset.py classify \
--input-dataset user/huge-dataset \
--column text \
--labels "A,B,C" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/huge-classified \
--batch-size 128
```
## 🤝 How It Works
1. **Sieves**: Provides a zero-shot task pipeline system for structured NLP workflows
2. **Outlines**: Provides guided decoding to guarantee valid label outputs
3. **UV**: Handles all dependencies automatically
The script loads your dataset, preprocesses texts, classifies each one with guaranteed valid outputs using Sieves'
`Classification` task, then saves the results as a new column in the output dataset.
## 🐛 Troubleshooting
### GPU Not Available
This script works best with a GPU but can run on CPU (much slower). To use GPU:
- Run on a machine with NVIDIA GPU
- Use cloud GPU instances (AWS, GCP, Azure, etc.)
- Use Hugging Face Spaces with GPU
### Out of Memory
- Use a smaller model (e.g., SmolLM-360M instead of 3B)
- Reduce `--batch-size` (try 32, 16, or 8)
- Reduce `--max-tokens` for shorter generations
### Invalid/Skipped Texts
- Texts shorter than 3 characters are skipped
- Empty or None values are marked as invalid
- Very long texts are truncated to 4000 characters
### Classification Quality
- With Outlines guided decoding, outputs are guaranteed to be valid labels
- For better results, use clear and distinct label names
- Try `--label-descriptions` to provide context
- Use a larger model for nuanced tasks
- In multi-label mode, adjust the confidence threshold (defaults to 0.5)
### Authentication Issues
If you see authentication errors:
- Run `huggingface-cli login` to cache your token
- Or set `export HF_TOKEN=your_token_here`
- Verify your token has read/write permissions on the Hub
## 🔬 Advanced Workflows
### Full Pipeline Workflow
Start with small tests, then run on the full dataset:
```bash
# Step 1: Test with small sample
uv run classify-dataset.py classify \
--input-dataset your-dataset \
--column text \
--labels "label1,label2,label3" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/test-classification \
--max-samples 100
# Step 2: If results look good, run on full dataset
uv run classify-dataset.py classify \
--input-dataset your-dataset \
--column text \
--labels "label1,label2,label3" \
--label-descriptions "label1:description,label2:description,label3:description" \
--model HuggingFaceTB/SmolLM-360M-Instruct \
--output-dataset user/final-classification \
--batch-size 64
```
## 📝 License
This example is provided as part of the [Sieves](https://github.com/MantisAI/sieves/) project. | 42 | 1 | [
"task_categories:zero-shot-classification",
"task_categories:text-classification",
"license:mit",
"region:us",
"uv-script",
"classification",
"structured-outputs",
"zero-shot"
] | 2025-10-15T12:26:40+00:00 | 2025-11-11T12:32:48+00:00 | 0 |
F-Fer/ur-3 |
# ur-2
**This dataset was generated using [phosphobot](https://docs.phospho.ai).**
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot.
To get started in robotics, [get your own phospho starter pack.](https://robots.phospho.ai).
|
# ur-2
**This dataset was generated using [phosphobot](https://docs.phospho.ai).**
This dataset contains a series of episodes recorded with a robot and multiple cameras. It can be directly used to train a policy using imitation learning. It's compatible with LeRobot.
To get started in robotics, [get your own phospho starter pack.](https://robots.phospho.ai).
| 115 | 0 | [
"task_categories:robotics",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"phosphobot",
"so100",
"phospho-dk"
] | 2025-11-11T12:24:57+00:00 | 2025-11-11T12:26:02+00:00 | 0 |
DmitryStrog/so101_pnp_merged_500 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 521,
"total_frames": 277740,
"total_tasks": 6,
"chunks_size": 1000,
"data_files_size_in_mb": 1000,
"video_files_size_in_mb": 10000,
"fps": 30,
"splits": {
"train": "0:521"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"observation.images.up": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 521,
"total_frames": 277740,
"total_tasks": 6,
"chunks_size": 1000,
"data_files_size_in_mb": 1000,
"video_files_size_in_mb": 10000,
"fps": 30,
"splits": {
"train": "0:521"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"observation.images.up": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 19 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T12:13:45+00:00 | 2025-11-11T12:15:28+00:00 | 0 |
ChaseLabs/Harmful-Texts-On-Mastodon | # 🦣 Mastodon Wild Data for Harmful Content Detection
## Overview
The **Harmful Texts on Mastodon** dataset is a human-annotated corpus of **3,000 English posts** collected from the decentralized social media platform **Mastodon** between **December 2024 and February 2025**.
It is designed to evaluate the **robustness**, **generalization**, and **personalization** capabilities of large language models (LLMs) and in-context learning (ICL) approaches for **harmful content detection in real-world scenarios**.
Unlike existing benchmark datasets, which are typically curated and balanced, this dataset captures the **natural distribution**, **domain shifts**, and **semantic overlaps** present in real-world social discourse.
Each post is annotated at **three granularities** — *binary*, *multi-class*, and *multi-label* — allowing flexible evaluation under multiple task formulations.
---
## 📊 Dataset Structure
| Granularity | Labels | Description |
|--------------|---------|-------------|
| **Binary** | `benign`, `harmful` | Basic harmfulness classification. |
| **Multi-class** | `benign`, `toxic`, `spam`, `negative` | Mutually exclusive fine-grained categories. |
| **Multi-label** | One or more from `{benign, toxic, spam, negative}` | Allows overlapping or composite labels for nuanced real-world cases. |
## 🧠 Motivation
Existing datasets such as SST-2, TextDetox, and UCI SMS provide clean, well-curated benchmarks for harmful content detection.
However, **real-world moderation** is far more complex — social media posts are **ambiguous**, **noisy**, and often contain **overlapping intents**.
For example, a post can simultaneously express anger (*negative*) while using profanity (*toxic*) or contain excessive hashtags (*spam-like*) without malicious intent.
The **Mastodon Wild Data** dataset addresses these limitations by introducing a *“wild” benchmark* that captures the messiness and richness of real-world online discourse.
It aims to:
- Evaluate **robustness and generalization** of large language models (LLMs) under domain shift.
- Reflect the **compositional nature** of harmful content (e.g., *toxic + negative*).
- Provide a unified resource for studying **multi-task**, **multi-class**, and **multi-label** formulations.
---
## 🏗️ Data Construction
- **Source:** Public Mastodon posts (Dec 2024 – Feb 2025).
- **Initial Corpus:** 8,998,738 posts → 3,948,831 unique English entries.
- **Filtering Strategy:**
1. Randomly sample **15,000** English posts.
2. Use **Llama-3 (48-shot Random)** ICL model for preliminary harmfulness prediction.
3. Select **1,500 predicted benign** and **1,500 predicted harmful** posts for manual annotation.
- **Final Dataset:** **3,000 annotated posts**, balanced between harmful and benign examples.
- **Annotation:** Each sample is labeled at three levels — binary, multi-class, and multi-label — by trained human annotators.
---
## 🧾 Label Statistics
### Multi-Class Distribution
| Label | Count | Percentage |
|--------|--------|-------------|
| Benign | 1798 | 59.9% |
| Negative | 755 | 25.2% |
| Toxic | 259 | 8.6% |
| Spam | 188 | 6.3% |
### Multi-label Label Distribution
| Labels | Count | Labels | Count | Labels | Count |
|:-------|------:|:-------|------:|:-------|------:|
| Benign | 1437 | Benign, Negative | 249 | Benign, Negative, Spam | 11 |
| Negative | 517 | Benign, Spam | 184 | Benign, Negative, Toxic | 13 |
| Spam | 60 | Benign, Toxic | 8 | Benign, Spam, Toxic | 5 |
| Toxic | 6 | Negative, Spam | 10 | Negative, Spam, Toxic | 3 |
| – | – | Negative, Toxic | 339 | – | – |
| – | – | Spam, Toxic | 98 | – | – |
| **Sum** | **2020** | – | **948** | – | **32** |
---
## 📚 Recommended Usage
This dataset is well-suited for:
- Evaluating **In-Context Learning (ICL)** and **prompt-based personalization** methods.
- Studying **robustness** and **domain generalization** in harmful content detection.
- Training or testing **multi-label** or **reason-augmented** classification frameworks.
- Benchmarking **cross-task**, **multi-task**, and **multi-modal** content moderation models.
---
## ⚖️ License
The dataset is distributed under the **CC BY 4.0 License**.
Users should also check the Terms of Service of the specific Mastodon instance you collect data from (e.g., [mastodon.social Terms of Service
](https://mstdn.social/terms-of-service)) when redistributing or reusing data derived from public posts.
---
## 🧩 Citation
If you use this dataset, please cite:
```bibtex
@misc{zhang2025onesizefitsallpersonalizedharmfulcontent,
title={Beyond One-Size-Fits-All: Personalized Harmful Content Detection with In-Context Learning},
author={Rufan Zhang and Lin Zhang and Xianghang Mi},
year={2025},
eprint={2511.05532},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2511.05532},
}
```
| # 🦣 Mastodon Wild Data for Harmful Content Detection
## Overview
The **Harmful Texts on Mastodon** dataset is a human-annotated corpus of **3,000 English posts** collected from the decentralized social media platform **Mastodon** between **December 2024 and February 2025**.
It is designed to evaluate the **robustness**, **generalization**, and **personalization** capabilities of large language models (LLMs) and in-context learning (ICL) approaches for **harmful content detection in real-world scenarios**.
Unlike existing benchmark datasets, which are typically curated and balanced, this dataset captures the **natural distribution**, **domain shifts**, and **semantic overlaps** present in real-world social discourse.
Each post is annotated at **three granularities** — *binary*, *multi-class*, and *multi-label* — allowing flexible evaluation under multiple task formulations.
---
## 📊 Dataset Structure
| Granularity | Labels | Description |
|--------------|---------|-------------|
| **Binary** | `benign`, `harmful` | Basic harmfulness classification. |
| **Multi-class** | `benign`, `toxic`, `spam`, `negative` | Mutually exclusive fine-grained categories. |
| **Multi-label** | One or more from `{benign, toxic, spam, negative}` | Allows overlapping or composite labels for nuanced real-world cases. |
## 🧠 Motivation
Existing datasets such as SST-2, TextDetox, and UCI SMS provide clean, well-curated benchmarks for harmful content detection.
However, **real-world moderation** is far more complex — social media posts are **ambiguous**, **noisy**, and often contain **overlapping intents**.
For example, a post can simultaneously express anger (*negative*) while using profanity (*toxic*) or contain excessive hashtags (*spam-like*) without malicious intent.
The **Mastodon Wild Data** dataset addresses these limitations by introducing a *“wild” benchmark* that captures the messiness and richness of real-world online discourse.
It aims to:
- Evaluate **robustness and generalization** of large language models (LLMs) under domain shift.
- Reflect the **compositional nature** of harmful content (e.g., *toxic + negative*).
- Provide a unified resource for studying **multi-task**, **multi-class**, and **multi-label** formulations.
---
## 🏗️ Data Construction
- **Source:** Public Mastodon posts (Dec 2024 – Feb 2025).
- **Initial Corpus:** 8,998,738 posts → 3,948,831 unique English entries.
- **Filtering Strategy:**
1. Randomly sample **15,000** English posts.
2. Use **Llama-3 (48-shot Random)** ICL model for preliminary harmfulness prediction.
3. Select **1,500 predicted benign** and **1,500 predicted harmful** posts for manual annotation.
- **Final Dataset:** **3,000 annotated posts**, balanced between harmful and benign examples.
- **Annotation:** Each sample is labeled at three levels — binary, multi-class, and multi-label — by trained human annotators.
---
## 🧾 Label Statistics
### Multi-Class Distribution
| Label | Count | Percentage |
|--------|--------|-------------|
| Benign | 1798 | 59.9% |
| Negative | 755 | 25.2% |
| Toxic | 259 | 8.6% |
| Spam | 188 | 6.3% |
### Multi-label Label Distribution
| Labels | Count | Labels | Count | Labels | Count |
|:-------|------:|:-------|------:|:-------|------:|
| Benign | 1437 | Benign, Negative | 249 | Benign, Negative, Spam | 11 |
| Negative | 517 | Benign, Spam | 184 | Benign, Negative, Toxic | 13 |
| Spam | 60 | Benign, Toxic | 8 | Benign, Spam, Toxic | 5 |
| Toxic | 6 | Negative, Spam | 10 | Negative, Spam, Toxic | 3 |
| – | – | Negative, Toxic | 339 | – | – |
| – | – | Spam, Toxic | 98 | – | – |
| **Sum** | **2020** | – | **948** | – | **32** |
---
## 📚 Recommended Usage
This dataset is well-suited for:
- Evaluating **In-Context Learning (ICL)** and **prompt-based personalization** methods.
- Studying **robustness** and **domain generalization** in harmful content detection.
- Training or testing **multi-label** or **reason-augmented** classification frameworks.
- Benchmarking **cross-task**, **multi-task**, and **multi-modal** content moderation models.
---
## ⚖️ License
The dataset is distributed under the **CC BY 4.0 License**.
Users should also check the Terms of Service of the specific Mastodon instance you collect data from (e.g., [mastodon.social Terms of Service
](https://mstdn.social/terms-of-service)) when redistributing or reusing data derived from public posts.
---
## 🧩 Citation
If you use this dataset, please cite:
```bibtex
@misc{zhang2025onesizefitsallpersonalizedharmfulcontent,
title={Beyond One-Size-Fits-All: Personalized Harmful Content Detection with In-Context Learning},
author={Rufan Zhang and Lin Zhang and Xianghang Mi},
year={2025},
eprint={2511.05532},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2511.05532},
}
```
| 34 | 1 | [
"task_categories:text-classification",
"language:en",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:csv",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2511.05532",
"region:us",
"Harmful",
"toxic",
"spam",
"negative"
] | 2025-10-17T01:56:30+00:00 | 2025-11-11T12:10:18+00:00 | 1 |
mcptester0606/Annoy-PyEdu-Rs-Raw | # Annoy: This should be a paper Title
<p align="left">
📑 <a href="https://huggingface.co/papers/xxxx.xxxxx" target="_blank">Paper</a>    |    🌐 <a href="https://specx.github.io/" target="_blank">Project Page</a>    |    💾 <a href="https://huggingface.co/collections/mcptester0606/specx-67a978e28fd926b56a4f55a2" target="_blank">Released Resources</a>    |    📦 <a href="https://github.com/mcptest-user/Annoy" target="_blank">Repo</a>
We release the raw data for our processed PythonEdu-Rs dataset, adopted from the original dataset from HuggingFaceTB team.
The data format for each line in the `0_368500_filtered_v2_ds25.sced.jsonl` is as follows:
```
{
"problem_description": <the problem description of the function>,
"io_requirements": <the input/output requirements and constraints>,
"refcode": <the reference code, including imported packages (optional), auxiliary functions (optional) and main entrypoint function>,
"funcname": <the function name for the entrypoint function>,
"ios": [
{
"input": <the input arguments>,
"output":<the returned value>
},
...
],
"source": <the source of the raw code files>,
"category": <the reasoning type we assign to this sample>,
"meta": <meta information about this sample>
}
```
Some of the `ios` are empty. The reason is that when executing the code, the input/output sizes are too large and exceed our required constraints. Thus, they are not stored or used later.
*Note: Due to imperfect LLM-based transformations, some problem descriptions do not contain enough information to describe the code. We leave this as future work to further enhance our data and update it to a better version. | # Annoy: This should be a paper Title
<p align="left">
📑 <a href="https://huggingface.co/papers/xxxx.xxxxx" target="_blank">Paper</a>    |    🌐 <a href="https://specx.github.io/" target="_blank">Project Page</a>    |    💾 <a href="https://huggingface.co/collections/mcptester0606/specx-67a978e28fd926b56a4f55a2" target="_blank">Released Resources</a>    |    📦 <a href="https://github.com/mcptest-user/Annoy" target="_blank">Repo</a>
We release the raw data for our processed PythonEdu-Rs dataset, adopted from the original dataset from HuggingFaceTB team.
The data format for each line in the `0_368500_filtered_v2_ds25.sced.jsonl` is as follows:
```
{
"problem_description": <the problem description of the function>,
"io_requirements": <the input/output requirements and constraints>,
"refcode": <the reference code, including imported packages (optional), auxiliary functions (optional) and main entrypoint function>,
"funcname": <the function name for the entrypoint function>,
"ios": [
{
"input": <the input arguments>,
"output":<the returned value>
},
...
],
"source": <the source of the raw code files>,
"category": <the reasoning type we assign to this sample>,
"meta": <meta information about this sample>
}
```
Some of the `ios` are empty. The reason is that when executing the code, the input/output sizes are too large and exceed our required constraints. Thus, they are not stored or used later.
*Note: Due to imperfect LLM-based transformations, some problem descriptions do not contain enough information to describe the code. We leave this as future work to further enhance our data and update it to a better version. | 128 | 0 | [
"region:us"
] | 2025-11-11T12:05:41+00:00 | 2025-11-11T12:05:44+00:00 | 0 |
mcptester0606/Annoy-PyEdu-Rs | # Annoy: This should be a paper Title
<p align="left">
📑 <a href="https://huggingface.co/papers/xxxx.xxxxx" target="_blank">Paper</a>    |    🌐 <a href="https://specx.github.io/" target="_blank">Project Page</a>    |    💾 <a href="https://huggingface.co/collections/mcptester0606/specx-67a978e28fd926b56a4f55a2" target="_blank">Released Resources</a>    |    📦 <a href="https://github.com/mcptest-user/Annoy" target="_blank">Repo</a>
This is the resource page of the our resources collection on Huggingface, we highlight your currect position with a blue block.
**Dataset**
<table>
<tr>
<th>Dataset</th>
<th>Link</th>
</tr>
<tr>
<td>Annoy-PythonEdu-Rs</td>
<td style="background-color: #e6f3ff; text-align: center; vertical-align: middle;">
<a href="https://huggingface.co/datasets/mcptester0606/Annoy-PyEdu-Rs">🤗</a>
</td>
</tr>
</table>
Please also check the raw data after our processing if you are interested: [mcptester0606/Annoy-PyEdu-Rs-Raw](https://huggingface.co/datasets/mcptester0606/Annoy-PyEdu-Rs-Raw).
**Models**
<table>
<tr>
<th rowspan="2">Base Model / Training</th>
<th colspan="2">Annoy</th>
<th colspan="2">Annoy++</th>
</tr>
<tr>
<th>Stage 1</th>
<th>Stage 2</th>
<th>Stage 1</th>
<th>Stage 2</th>
</tr>
<tr>
<td>Qwen 2.5 7B Coder</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_pp">🤗</a></td>
</tr>
<tr>
<td>LLaMA 3.1 8B</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_pp">🤗</a></td>
</tr>
<tr>
<td>DeepSeek v2 Lite Coder</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_pp">🤗</a></td>
</tr>
</table>
**Introduction**
While having full executable code theoretically allows us to generate reliable execution trajectories as responses, two challenges arise: 1) Obtaining a deterministic reverse function for input prediction is impractical; 2) Automatically constructed trajectories are constrained by pre-designed templates and lack the expressiveness and generalizability of free-form natural language reasoning. Thus, we adopt a fully LLM-based approach for synthesizing all the desired responses using DeepSeek-V2.5, as it has top-tier performance but extremely low cost compared to other advanced LLMs.
*Due to our collaborators' compliance requirements, we only release the PythonEdu-Rs subset (this page) of full dataset. | # Annoy: This should be a paper Title
<p align="left">
📑 <a href="https://huggingface.co/papers/xxxx.xxxxx" target="_blank">Paper</a>    |    🌐 <a href="https://specx.github.io/" target="_blank">Project Page</a>    |    💾 <a href="https://huggingface.co/collections/mcptester0606/specx-67a978e28fd926b56a4f55a2" target="_blank">Released Resources</a>    |    📦 <a href="https://github.com/mcptest-user/Annoy" target="_blank">Repo</a>
This is the resource page of the our resources collection on Huggingface, we highlight your currect position with a blue block.
**Dataset**
<table>
<tr>
<th>Dataset</th>
<th>Link</th>
</tr>
<tr>
<td>Annoy-PythonEdu-Rs</td>
<td style="background-color: #e6f3ff; text-align: center; vertical-align: middle;">
<a href="https://huggingface.co/datasets/mcptester0606/Annoy-PyEdu-Rs">🤗</a>
</td>
</tr>
</table>
Please also check the raw data after our processing if you are interested: [mcptester0606/Annoy-PyEdu-Rs-Raw](https://huggingface.co/datasets/mcptester0606/Annoy-PyEdu-Rs-Raw).
**Models**
<table>
<tr>
<th rowspan="2">Base Model / Training</th>
<th colspan="2">Annoy</th>
<th colspan="2">Annoy++</th>
</tr>
<tr>
<th>Stage 1</th>
<th>Stage 2</th>
<th>Stage 1</th>
<th>Stage 2</th>
</tr>
<tr>
<td>Qwen 2.5 7B Coder</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/qwen2.5-7b-coder_spec_pp">🤗</a></td>
</tr>
<tr>
<td>LLaMA 3.1 8B</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/llama3.1-8b_spec_pp">🤗</a></td>
</tr>
<tr>
<td>DeepSeek v2 Lite Coder</td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_pp_stage1">🤗</a></td>
<td style="text-align: center; vertical-align: middle;"><a href="https://huggingface.co/mcptester0606/dsv2-lite-coder_spec_pp">🤗</a></td>
</tr>
</table>
**Introduction**
While having full executable code theoretically allows us to generate reliable execution trajectories as responses, two challenges arise: 1) Obtaining a deterministic reverse function for input prediction is impractical; 2) Automatically constructed trajectories are constrained by pre-designed templates and lack the expressiveness and generalizability of free-form natural language reasoning. Thus, we adopt a fully LLM-based approach for synthesizing all the desired responses using DeepSeek-V2.5, as it has top-tier performance but extremely low cost compared to other advanced LLMs.
*Due to our collaborators' compliance requirements, we only release the PythonEdu-Rs subset (this page) of full dataset. | 130 | 0 | [
"region:us"
] | 2025-11-11T12:05:41+00:00 | 2025-11-11T12:05:42+00:00 | 0 |
conor99/InfiniteARC |
# InfiniteARC
A collection of synthetic ARC-style task generators and solvers.
An auto-generated Python module that provides API-style access to
the tasks can be found on GitHub at [conor-99/InfiniteARC-API](https://github.com/conor-99/InfiniteARC-API).
|
# InfiniteARC
A collection of synthetic ARC-style task generators and solvers.
An auto-generated Python module that provides API-style access to
the tasks can be found on GitHub at [conor-99/InfiniteARC-API](https://github.com/conor-99/InfiniteARC-API).
| 136 | 0 | [
"language:en",
"license:mit",
"size_categories:1K<n<10K",
"format:json",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"region:us"
] | 2025-11-04T13:20:42+00:00 | 2025-11-11T12:05:26+00:00 | 0 |
kinghanse/pick_socks_hs |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 145,
"total_frames": 119610,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:145"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 145,
"total_frames": 119610,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:145"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 64 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T09:11:10+00:00 | 2025-11-11T11:58:39+00:00 | 0 |
johannesschirrmeister/eval_line-check_groot |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 10,
"total_frames": 3920,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 10,
"total_frames": 3920,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 12 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T12:01:51+00:00 | 2025-11-11T12:01:58+00:00 | 0 |
TakalaWang/anime-2024-winter-segment-queries |
# Anime 2024 Winter - Segment Queries
這個數據集包含 2024 年冬季動畫的片段級別查詢語句。
## 數據集結構
- **file_name**: 影片文件路徑（用於定位影片片段位置）
- **series_name**: 動畫系列名稱
- **episode_id**: 集數 ID
- **segment_index**: 片段索引
- **release_date**: 發布日期
- **query**: 查詢語句集合
- visual_saliency: 視覺顯著物
- character_emotion: 角色情緒
- action_behavior: 動作行為
- dialogue: 對話台詞
- symbolic_scene: 象徵性畫面 |
# Anime 2024 Winter - Segment Queries
這個數據集包含 2024 年冬季動畫的片段級別查詢語句。
## 數據集結構
- **file_name**: 影片文件路徑（用於定位影片片段位置）
- **series_name**: 動畫系列名稱
- **episode_id**: 集數 ID
- **segment_index**: 片段索引
- **release_date**: 發布日期
- **query**: 查詢語句集合
- visual_saliency: 視覺顯著物
- character_emotion: 角色情緒
- action_behavior: 動作行為
- dialogue: 對話台詞
- symbolic_scene: 象徵性畫面 | 564 | 0 | [
"language:zh",
"size_categories:1K<n<10K",
"modality:text",
"modality:video",
"library:datasets",
"library:mlcroissant",
"region:us"
] | 2025-11-08T10:52:14+00:00 | 2025-11-11T11:56:58+00:00 | 0 |
TakalaWang/anime-2024-winter-episode-queries |
# Anime 2024 Winter - Episode Queries
這個數據集包含 2024 年冬季動畫的集數級別查詢語句。
## 數據集結構
- **file_name**: 影片文件路徑（用於定位集數影片位置）
- **series_name**: 動畫系列名稱
- **episode_id**: 集數 ID
- **release_date**: 發布日期
- **query**: 模型生成的查詢語句
- main_plot: 主要劇情
- turning_point: 轉折點
- relationship_change: 關係變化
- episode_mood: 集數氛圍
- notable_scene: 易記場景 |
# Anime 2024 Winter - Episode Queries
這個數據集包含 2024 年冬季動畫的集數級別查詢語句。
## 數據集結構
- **file_name**: 影片文件路徑（用於定位集數影片位置）
- **series_name**: 動畫系列名稱
- **episode_id**: 集數 ID
- **release_date**: 發布日期
- **query**: 模型生成的查詢語句
- main_plot: 主要劇情
- turning_point: 轉折點
- relationship_change: 關係變化
- episode_mood: 集數氛圍
- notable_scene: 易記場景 | 351 | 0 | [
"language:zh",
"size_categories:n<1K",
"modality:text",
"modality:video",
"library:datasets",
"library:mlcroissant",
"region:us"
] | 2025-11-08T10:52:14+00:00 | 2025-11-11T11:57:00+00:00 | 0 |
codelion/synth-1B | # synth-1B
Sequential sample of the first 999,997,890 tokens from [PleIAs/SYNTH](https://huggingface.co/datasets/PleIAs/SYNTH).
## Dataset Details
- **Source**: PleIAs/SYNTH (500 parquet files, ~87B tokens total)
- **Sampling Method**: Sequential (first N documents)
- **Estimated Tokens**: 999,997,890
- **Documents**: 822,230
- **Token Estimation**: 4 characters ≈ 1 token
## Text Fields
Each document combines four fields from the original dataset:
- `query`: The question or prompt
- `query_seed_text`: Wikipedia or reference context
- `synthetic_reasoning`: Step-by-step reasoning trace
- `synthetic_answer`: Final answer
These are concatenated with double newlines to create comprehensive training examples.
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("codelion/synth-1B")
```
## License
Same as source dataset (PleIAs/SYNTH). | # synth-1B
Sequential sample of the first 999,997,890 tokens from [PleIAs/SYNTH](https://huggingface.co/datasets/PleIAs/SYNTH).
## Dataset Details
- **Source**: PleIAs/SYNTH (500 parquet files, ~87B tokens total)
- **Sampling Method**: Sequential (first N documents)
- **Estimated Tokens**: 999,997,890
- **Documents**: 822,230
- **Token Estimation**: 4 characters ≈ 1 token
## Text Fields
Each document combines four fields from the original dataset:
- `query`: The question or prompt
- `query_seed_text`: Wikipedia or reference context
- `synthetic_reasoning`: Step-by-step reasoning trace
- `synthetic_answer`: Final answer
These are concatenated with double newlines to create comprehensive training examples.
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("codelion/synth-1B")
```
## License
Same as source dataset (PleIAs/SYNTH). | 57 | 0 | [
"size_categories:100K<n<1M",
"format:parquet",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T10:28:51+00:00 | 2025-11-11T11:56:36+00:00 | 0 |
yamilama/record-test8 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 1,
"total_frames": 297,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so101_follower",
"total_episodes": 1,
"total_frames": 297,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 20 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:n<1K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:56:29+00:00 | 2025-11-11T11:56:35+00:00 | 0 |
AzuratiX/mirobot-pickplace |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "wlkata_mirobot",
"total_episodes": 63,
"total_frames": 15993,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:63"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"pose_x",
"pose_y",
"pose_z",
"pose_roll",
"pose_pitch",
"pose_yaw",
"gripper_open"
],
"shape": [
7
]
},
"observation.state": {
"dtype": "float32",
"names": [
"pose_x",
"pose_y",
"pose_z",
"pose_roll",
"pose_pitch",
"pose_yaw",
"gripper_open"
],
"shape": [
7
]
},
"observation.images.top_camera": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist_camera": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "wlkata_mirobot",
"total_episodes": 63,
"total_frames": 15993,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:63"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"pose_x",
"pose_y",
"pose_z",
"pose_roll",
"pose_pitch",
"pose_yaw",
"gripper_open"
],
"shape": [
7
]
},
"observation.state": {
"dtype": "float32",
"names": [
"pose_x",
"pose_y",
"pose_z",
"pose_roll",
"pose_pitch",
"pose_yaw",
"gripper_open"
],
"shape": [
7
]
},
"observation.images.top_camera": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.wrist_camera": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 93 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-07T04:02:58+00:00 | 2025-11-11T11:55:04+00:00 | 0 |
T1g3rGE/eval_so100_pick_doll2 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 3,
"total_frames": 1234,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:3"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.high": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "so100_follower",
"total_episodes": 3,
"total_frames": 1234,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:3"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.wrist": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.high": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 20 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:tabular",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:54:29+00:00 | 2025-11-11T11:54:38+00:00 | 0 |
OpenSTEF/liander2024-energy-forecasting-benchmark |
# Dataset Card for Liander 2024 Short Term Energy Forecasting Benchmark
[](https://huggingface.co/datasets/OpenSTEF/liander2024-energy-forecasting-benchmark)
This dataset provides a benchmark for short term energy forecasting models, combining electrical load measurements from Dutch DSO Liander with predictors like corresponding weather data from OpenMeteo, day-ahead electricity prices from ENTSO-E, and profiles of electricity consumption from Energiedatawijzer. The dataset covers the full year 2024 (2024-01-01 to 2025-01-01 UTC) and includes 55 different points in the grid across the Netherlands at various levels within the grid.
The dataset is designed for developing and validating short-term energy forecasting models, particularly those that incorporate weather variables. It serves as a standardized benchmark for comparing different short term forecasting approaches in the energy domain.
## Dataset Details
- **Curated by:** [OpenSTEF](https://github.com/OpenSTEF)
- **License:** Creative Commons BY 4.0 (CC BY 4.0). See below for specific source data licenses.
- **Data Period:** 2024-01-01 to 2025-01-01
- **Temporal Resolution:** 15-minute intervals for load measurements and profiles, hourly for weather data and prices (interpolated to 15-minute intervals)
- **Geographic Coverage:** 55 points in the grid across the Netherlands in Liander service area
- **Total Size:** ~3-6M data points across all components
## Dataset Components
The dataset consists of six main components:
### 1. Load Measurements (`load_measurements/`)
Electrical load (active power) measurements from various types of infrastructure managed by Dutch DSO Liander. All measurements are recorded at 15-minute intervals.
**Location Types:**
- **mv_feeder** (Medium Voltage Feeders): Outgoing medium voltage cables from primary substations
- **station_installation** (Station Installations): Various primary substation installations
- **transformer** (Transformers): Power transformers at primary substations
- **solar_park** (Solar Parks): Anonymized and normalized individual solar park measurements
- **wind_park** (Wind Parks): Anonymized and normalized individual wind park measurements
> [!NOTE]
> Solar and wind park data includes a 2-day availability delay to simulate data availability constraints at Dutch DSOs.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Measurement timestamp in UTC |
| load | float64 | W | Electrical load in watts |
| available_at | datetime64[ns, UTC] | - | Data availability timestamp |
### 2. Weather Measurements (`weather_measurements/`)
Historical weather measurements from OpenMeteo for each load measurement location, providing ground truth weather conditions.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 3. Weather Forecasts (`weather_forecasts/`)
Latest available weather forecasts from OpenMeteo (short horizon). These represent the best available forecast at each time point.
> [!WARNING]
> This component is useful for simple forecasting experiments but is not fully realistic for benchmarking since it does not simulate real-world forecast availability.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_speed_80m | float32 | km/h | Wind speed at 80 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 4. Versioned Weather Forecasts (`weather_forecasts_versioned/`)
Time-versioned weather forecasts with lead times up to 7 days ahead, simulating real-world data availability. This component provides the most realistic forecasting scenario.
> [!NOTE]
> This enables realistic evaluation where forecasts are only available at specific times with specific lead times, matching real-world operational constraints.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Target forecast timestamp |
| available_at | datetime64[ns, UTC] | - | When the forecast was available/created |
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_speed_80m | float32 | km/h | Wind speed at 80 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 5. EPEX Day-Ahead Prices (`EPEX.parquet`)
Day-ahead electricity prices for the Netherlands from ENTSO-E Transparency Platform, providing market price signals that influence energy consumption patterns.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Price delivery timestamp in UTC |
| available_at | datetime64[ns, UTC] | - | When the price was published/available |
| price | float64 | €/MWh | Day-ahead electricity price in euros per megawatt hour |
### 6. Electricity Consumption Profiles (`profiles.parquet`)
Standardized electricity consumption profiles from Energiedatawijzer for various customer categories in the Netherlands, representing typical usage patterns throughout the year. These values are typically normalized to sum to 1 over the year. There are 15 types of profiles, which can be read as follows: `{category}_{type}_{direction}`, where `category` says something about the connection type, `type` indicates whether it is a connection with or without infeed, and `direction` indicates whether it is a consumption or generation profile (we only include consumption profiles as infeed says something about previous year's generation). For a full description of the profiles, see the [Energiedatawijzer documentation](https://energiedatawijzer.nl/documenten/profielen-elektriciteit-2024/).
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Profile timestamp in UTC |
| available_at | datetime64[ns, UTC] | - | Data availability timestamp |
| {profiles} | float64 | - | 15 profiles for different categories |
## Uses
This dataset is intended for energy forecasting research, providing a standardized benchmark for comparing different forecasting approaches in the energy domain. The dataset supports various forecasting horizons and scenarios:
- **Operational Forecasting**: 15-minute to 24-hour ahead load predictions
- **Day-ahead Congestion Management**: Using weather forecasts for next-day congestion predictions
- **Multi-modal Forecasting**: Combining multiple infrastructure types and weather variables
- **Uncertainty Quantification**: Using versioned forecasts to assess prediction uncertainty
- **Weather-Energy Relationship Studies**: Analyzing correlations between weather variables and electrical load
This dataset is compatible with various forecasting frameworks, including **[OpenSTEF](https://github.com/OpenSTEF/openstef)** (Open Short Term Energy Forecasting), classical time series models, machine learning approaches, and deep learning models.
## Dataset Structure
The dataset is organized in the following directory structure:
```
liander2024/
├── liander2024_targets.yaml # Location metadata with coordinates
├── load_measurements/ # Electrical load data
│ ├── mv_feeder/ # Medium voltage feeder measurements
│ ├── station_installation/ # Substation installation measurements
│ ├── transformer/ # Transformer measurements
│ ├── solar_park/ # Anonymized solar park measurements
│ └── wind_park/ # Anonymized wind park measurements
├── weather_measurements/ # Historical weather data
│ └── [same subdirectory structure as above]
├── weather_forecasts/ # Latest weather forecasts
│ └── [same subdirectory structure as above]
├── weather_forecasts_versioned/ # Time-versioned weather forecasts
│ └── [same subdirectory structure as above]
├── EPEX.parquet # Day-ahead electricity prices
└── profiles.parquet # Electricity consumption profiles
```
Each subdirectory contains individual Parquet files for each location, named according to the location identifier.
### Target Metadata (`liander2024_targets.yaml`)
The `liander2024_targets.yaml` file contains metadata for all 55 forecasting targets in the dataset. Each target includes:
| Field | Type | Description |
|:-----:|:----:|:-----------:|
| name | string | Unique identifier for the location/asset |
| group_name | string | Infrastructure type: `mv_feeder`, `transformer`, `station_installation`, `solar_park`, or `wind_park` |
| latitude | float | Approximate latitude coordinate* |
| longitude | float | Approximate longitude coordinate* |
| description | string | Human-readable description of the location |
| benchmark_start | datetime | Start of the benchmark evaluation period |
| benchmark_end | datetime | End of the benchmark evaluation period |
| train_start | datetime | Start of the training data period |
| upper_limit | float | 98th percentile of load values (W) |
| lower_limit | float | 2nd percentile of load values (W) |
\* Location coordinates are approximate and only based on the name of the target.
## Dataset Creation
### Source Data
#### Liander Historical Measurements
- **Source**: [Liander Open Data - Historical 15-minute Operational Measurements](https://www.liander.nl/over-ons/open-data#historische-15-minuten-bedrijfsmetingen)
- **License**: [See custom disclaimer](https://www.liander.nl/over-ons/open-data/disclaimer)
- **Description**: 15-minute electrical load measurements from various infrastructure types across Liander's service territory
- **Modifications made**: Converted into standardized Parquet format, removed _normalized suffix from load column, added `available_at` timestamps.
#### OpenMeteo Weather Data
- **Source**: [OpenMeteo Historical Weather API](https://open-meteo.com/)
- **License**: CC BY 4.0
- **Description**: Historical weather measurements and forecasts using the best available weather models
#### ENTSO-E Day-Ahead Prices
- **Source**: [ENTSO-E Transparency Platform](https://newtransparency.entsoe.eu/)
- **License**: CC BY 4.0
- **Description**: Day-ahead electricity prices for the Netherlands (EPEX Spot NL)
- **Modifications made**: Converted into Parquet format, converted to UTC, added `available_at` timestamp based on availability of day ahead prices in NL.
#### Energiedatawijzer Consumption Profiles
- **Source**: [Energiedatawijzer - Profielen elektriciteit 2024](https://energiedatawijzer.nl/documenten/profielen-elektriciteit-2024/)
- **License**: None, but permission granted for use in this dataset
- **Description**: Standardized electricity consumption profiles for various customer categories in the Netherlands
- **Modifications made**: Converted into Parquet format, converted to UTC added `available_at` timestamp, removed infeed profiles, used first hour of the year to fill the last hour of the year to get a full UTC year.
> [!NOTE]
> Location coordinates are approximate and may not represent exact facility locations. Solar and wind park data is normalized and anonymized for privacy. Weather data is interpolated from hourly to 15-minute resolution to match load measurements.
## How to Use
You can load the dataset files directly into pandas dataframes:
```python
import pandas as pd
load_data = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/load_measurements/mv_feeder/OS Edam.parquet")
weather_data = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/weather_measurements_versioned/mv_feeder/OS Edam.parquet")
epex = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/EPEX.parquet")
profiles = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/profiles.parquet")
```
|
# Dataset Card for Liander 2024 Short Term Energy Forecasting Benchmark
[](https://huggingface.co/datasets/OpenSTEF/liander2024-energy-forecasting-benchmark)
This dataset provides a benchmark for short term energy forecasting models, combining electrical load measurements from Dutch DSO Liander with predictors like corresponding weather data from OpenMeteo, day-ahead electricity prices from ENTSO-E, and profiles of electricity consumption from Energiedatawijzer. The dataset covers the full year 2024 (2024-01-01 to 2025-01-01 UTC) and includes 55 different points in the grid across the Netherlands at various levels within the grid.
The dataset is designed for developing and validating short-term energy forecasting models, particularly those that incorporate weather variables. It serves as a standardized benchmark for comparing different short term forecasting approaches in the energy domain.
## Dataset Details
- **Curated by:** [OpenSTEF](https://github.com/OpenSTEF)
- **License:** Creative Commons BY 4.0 (CC BY 4.0). See below for specific source data licenses.
- **Data Period:** 2024-01-01 to 2025-01-01
- **Temporal Resolution:** 15-minute intervals for load measurements and profiles, hourly for weather data and prices (interpolated to 15-minute intervals)
- **Geographic Coverage:** 55 points in the grid across the Netherlands in Liander service area
- **Total Size:** ~3-6M data points across all components
## Dataset Components
The dataset consists of six main components:
### 1. Load Measurements (`load_measurements/`)
Electrical load (active power) measurements from various types of infrastructure managed by Dutch DSO Liander. All measurements are recorded at 15-minute intervals.
**Location Types:**
- **mv_feeder** (Medium Voltage Feeders): Outgoing medium voltage cables from primary substations
- **station_installation** (Station Installations): Various primary substation installations
- **transformer** (Transformers): Power transformers at primary substations
- **solar_park** (Solar Parks): Anonymized and normalized individual solar park measurements
- **wind_park** (Wind Parks): Anonymized and normalized individual wind park measurements
> [!NOTE]
> Solar and wind park data includes a 2-day availability delay to simulate data availability constraints at Dutch DSOs.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Measurement timestamp in UTC |
| load | float64 | W | Electrical load in watts |
| available_at | datetime64[ns, UTC] | - | Data availability timestamp |
### 2. Weather Measurements (`weather_measurements/`)
Historical weather measurements from OpenMeteo for each load measurement location, providing ground truth weather conditions.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 3. Weather Forecasts (`weather_forecasts/`)
Latest available weather forecasts from OpenMeteo (short horizon). These represent the best available forecast at each time point.
> [!WARNING]
> This component is useful for simple forecasting experiments but is not fully realistic for benchmarking since it does not simulate real-world forecast availability.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_speed_80m | float32 | km/h | Wind speed at 80 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 4. Versioned Weather Forecasts (`weather_forecasts_versioned/`)
Time-versioned weather forecasts with lead times up to 7 days ahead, simulating real-world data availability. This component provides the most realistic forecasting scenario.
> [!NOTE]
> This enables realistic evaluation where forecasts are only available at specific times with specific lead times, matching real-world operational constraints.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Target forecast timestamp |
| available_at | datetime64[ns, UTC] | - | When the forecast was available/created |
| temperature_2m | float32 | °C | Air temperature at 2 meters above ground |
| relative_humidity_2m | float32 | % | Relative humidity at 2 meters above ground |
| surface_pressure | float32 | hPa | Atmospheric pressure at surface level |
| cloud_cover | float32 | % | Total cloud cover as area fraction |
| wind_speed_10m | float32 | km/h | Wind speed at 10 meters above ground |
| wind_speed_80m | float32 | km/h | Wind speed at 80 meters above ground |
| wind_direction_10m | float32 | ° | Wind direction at 10 meters above ground |
| shortwave_radiation | float32 | W/m² | Shortwave solar radiation |
| direct_radiation | float32 | W/m² | Direct solar radiation on horizontal plane |
| diffuse_radiation | float32 | W/m² | Diffuse solar radiation |
| direct_normal_irradiance | float32 | W/m² | Direct solar radiation on normal plane |
### 5. EPEX Day-Ahead Prices (`EPEX.parquet`)
Day-ahead electricity prices for the Netherlands from ENTSO-E Transparency Platform, providing market price signals that influence energy consumption patterns.
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Price delivery timestamp in UTC |
| available_at | datetime64[ns, UTC] | - | When the price was published/available |
| price | float64 | €/MWh | Day-ahead electricity price in euros per megawatt hour |
### 6. Electricity Consumption Profiles (`profiles.parquet`)
Standardized electricity consumption profiles from Energiedatawijzer for various customer categories in the Netherlands, representing typical usage patterns throughout the year. These values are typically normalized to sum to 1 over the year. There are 15 types of profiles, which can be read as follows: `{category}_{type}_{direction}`, where `category` says something about the connection type, `type` indicates whether it is a connection with or without infeed, and `direction` indicates whether it is a consumption or generation profile (we only include consumption profiles as infeed says something about previous year's generation). For a full description of the profiles, see the [Energiedatawijzer documentation](https://energiedatawijzer.nl/documenten/profielen-elektriciteit-2024/).
| Column | Type | Unit | Description |
|:------:|:----:|:----:|:-----------:|
| timestamp | datetime64[ns, UTC] | - | Profile timestamp in UTC |
| available_at | datetime64[ns, UTC] | - | Data availability timestamp |
| {profiles} | float64 | - | 15 profiles for different categories |
## Uses
This dataset is intended for energy forecasting research, providing a standardized benchmark for comparing different forecasting approaches in the energy domain. The dataset supports various forecasting horizons and scenarios:
- **Operational Forecasting**: 15-minute to 24-hour ahead load predictions
- **Day-ahead Congestion Management**: Using weather forecasts for next-day congestion predictions
- **Multi-modal Forecasting**: Combining multiple infrastructure types and weather variables
- **Uncertainty Quantification**: Using versioned forecasts to assess prediction uncertainty
- **Weather-Energy Relationship Studies**: Analyzing correlations between weather variables and electrical load
This dataset is compatible with various forecasting frameworks, including **[OpenSTEF](https://github.com/OpenSTEF/openstef)** (Open Short Term Energy Forecasting), classical time series models, machine learning approaches, and deep learning models.
## Dataset Structure
The dataset is organized in the following directory structure:
```
liander2024/
├── liander2024_targets.yaml # Location metadata with coordinates
├── load_measurements/ # Electrical load data
│ ├── mv_feeder/ # Medium voltage feeder measurements
│ ├── station_installation/ # Substation installation measurements
│ ├── transformer/ # Transformer measurements
│ ├── solar_park/ # Anonymized solar park measurements
│ └── wind_park/ # Anonymized wind park measurements
├── weather_measurements/ # Historical weather data
│ └── [same subdirectory structure as above]
├── weather_forecasts/ # Latest weather forecasts
│ └── [same subdirectory structure as above]
├── weather_forecasts_versioned/ # Time-versioned weather forecasts
│ └── [same subdirectory structure as above]
├── EPEX.parquet # Day-ahead electricity prices
└── profiles.parquet # Electricity consumption profiles
```
Each subdirectory contains individual Parquet files for each location, named according to the location identifier.
### Target Metadata (`liander2024_targets.yaml`)
The `liander2024_targets.yaml` file contains metadata for all 55 forecasting targets in the dataset. Each target includes:
| Field | Type | Description |
|:-----:|:----:|:-----------:|
| name | string | Unique identifier for the location/asset |
| group_name | string | Infrastructure type: `mv_feeder`, `transformer`, `station_installation`, `solar_park`, or `wind_park` |
| latitude | float | Approximate latitude coordinate* |
| longitude | float | Approximate longitude coordinate* |
| description | string | Human-readable description of the location |
| benchmark_start | datetime | Start of the benchmark evaluation period |
| benchmark_end | datetime | End of the benchmark evaluation period |
| train_start | datetime | Start of the training data period |
| upper_limit | float | 98th percentile of load values (W) |
| lower_limit | float | 2nd percentile of load values (W) |
\* Location coordinates are approximate and only based on the name of the target.
## Dataset Creation
### Source Data
#### Liander Historical Measurements
- **Source**: [Liander Open Data - Historical 15-minute Operational Measurements](https://www.liander.nl/over-ons/open-data#historische-15-minuten-bedrijfsmetingen)
- **License**: [See custom disclaimer](https://www.liander.nl/over-ons/open-data/disclaimer)
- **Description**: 15-minute electrical load measurements from various infrastructure types across Liander's service territory
- **Modifications made**: Converted into standardized Parquet format, removed _normalized suffix from load column, added `available_at` timestamps.
#### OpenMeteo Weather Data
- **Source**: [OpenMeteo Historical Weather API](https://open-meteo.com/)
- **License**: CC BY 4.0
- **Description**: Historical weather measurements and forecasts using the best available weather models
#### ENTSO-E Day-Ahead Prices
- **Source**: [ENTSO-E Transparency Platform](https://newtransparency.entsoe.eu/)
- **License**: CC BY 4.0
- **Description**: Day-ahead electricity prices for the Netherlands (EPEX Spot NL)
- **Modifications made**: Converted into Parquet format, converted to UTC, added `available_at` timestamp based on availability of day ahead prices in NL.
#### Energiedatawijzer Consumption Profiles
- **Source**: [Energiedatawijzer - Profielen elektriciteit 2024](https://energiedatawijzer.nl/documenten/profielen-elektriciteit-2024/)
- **License**: None, but permission granted for use in this dataset
- **Description**: Standardized electricity consumption profiles for various customer categories in the Netherlands
- **Modifications made**: Converted into Parquet format, converted to UTC added `available_at` timestamp, removed infeed profiles, used first hour of the year to fill the last hour of the year to get a full UTC year.
> [!NOTE]
> Location coordinates are approximate and may not represent exact facility locations. Solar and wind park data is normalized and anonymized for privacy. Weather data is interpolated from hourly to 15-minute resolution to match load measurements.
## How to Use
You can load the dataset files directly into pandas dataframes:
```python
import pandas as pd
load_data = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/load_measurements/mv_feeder/OS Edam.parquet")
weather_data = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/weather_measurements_versioned/mv_feeder/OS Edam.parquet")
epex = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/EPEX.parquet")
profiles = pd.read_parquet("hf://datasets/OpenSTEF/liander2024-energy-forecasting-benchmark/profiles.parquet")
```
| 378 | 1 | [
"task_categories:time-series-forecasting",
"task_ids:univariate-time-series-forecasting",
"task_ids:multivariate-time-series-forecasting",
"source_datasets:original",
"license:cc-by-4.0",
"size_categories:10M<n<100M",
"format:parquet",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"climate",
"energy",
"forecasting",
"time-series",
"weather",
"power-grid",
"profiles",
"price",
"electricity",
"load",
"demand",
"generation",
"short-term"
] | 2025-10-14T14:20:58+00:00 | 2025-11-11T11:47:25+00:00 | 0 |
TheFactoryX/edition_0309_shi-labs-oneformer_demo-readymade |
# edition_0309_shi-labs-oneformer_demo-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[shi-labs/oneformer_demo](https://huggingface.co/datasets/shi-labs/oneformer_demo)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0309_shi-labs-oneformer_demo-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[shi-labs/oneformer_demo](https://huggingface.co/datasets/shi-labs/oneformer_demo)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 4 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T11:48:41+00:00 | 2025-11-11T11:48:44+00:00 | 0 |
Mardiyyah/TAPT_data_V2_split |
# Dataset Card for "TAPT_data_V2_split"
## Dataset Description
Generated using EPMC API
PMCIDs:
Train-Val split ratio 80:20.
Data has been split on PMCIDs for traceability.
[More Information Needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) |
# Dataset Card for "TAPT_data_V2_split"
## Dataset Description
Generated using EPMC API
PMCIDs:
Train-Val split ratio 80:20.
Data has been split on PMCIDs for traceability.
[More Information Needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards) | 15 | 0 | [
"language:en",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-11-11T11:30:49+00:00 | 2025-11-11T11:42:02+00:00 | 0 |
WeiXiCZ/traj_train_cot_lingoqa_counter_full_traj3_hard2_epoch2_traj_full_know6k_4e-5_2 |
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed]
### Framework versions
- PEFT 0.15.2 |
# Model Card for Model ID
<!-- Provide a quick summary of what the model is/does. -->
## Model Details
### Model Description
<!-- Provide a longer summary of what this model is. -->
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
## Uses
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
### Direct Use
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
[More Information Needed]
### Downstream Use [optional]
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
[More Information Needed]
### Out-of-Scope Use
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
[More Information Needed]
## Bias, Risks, and Limitations
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
[More Information Needed]
### Recommendations
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
## How to Get Started with the Model
Use the code below to get started with the model.
[More Information Needed]
## Training Details
### Training Data
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
[More Information Needed]
### Training Procedure
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
[More Information Needed]
#### Training Hyperparameters
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
#### Speeds, Sizes, Times [optional]
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
[More Information Needed]
## Evaluation
<!-- This section describes the evaluation protocols and provides the results. -->
### Testing Data, Factors & Metrics
#### Testing Data
<!-- This should link to a Dataset Card if possible. -->
[More Information Needed]
#### Factors
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
[More Information Needed]
#### Metrics
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
### Results
[More Information Needed]
#### Summary
## Model Examination [optional]
<!-- Relevant interpretability work for the model goes here -->
[More Information Needed]
## Environmental Impact
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
## Technical Specifications [optional]
### Model Architecture and Objective
[More Information Needed]
### Compute Infrastructure
[More Information Needed]
#### Hardware
[More Information Needed]
#### Software
[More Information Needed]
## Citation [optional]
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
**BibTeX:**
[More Information Needed]
**APA:**
[More Information Needed]
## Glossary [optional]
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
[More Information Needed]
## More Information [optional]
[More Information Needed]
## Model Card Authors [optional]
[More Information Needed]
## Model Card Contact
[More Information Needed]
### Framework versions
- PEFT 0.15.2 | 45 | 0 | [
"arxiv:1910.09700",
"region:us"
] | 2025-11-11T11:42:54+00:00 | 2025-11-11T11:44:19+00:00 | 0 |
antwoor/screwdriver_95_degs |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8763,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "mcx",
"total_episodes": 10,
"total_frames": 8763,
"total_tasks": 1,
"total_videos": 20,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:10"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
7
],
"names": [
"joint1.pos",
"joint2.pos",
"joint3.pos",
"joint4.pos",
"joint5.pos",
"joint6.pos",
"gripper.pos"
]
},
"observation.images.camera_1": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.camera_2": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 27 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:39:11+00:00 | 2025-11-11T11:39:48+00:00 | 0 |
Maxbenkre/pharmaceutical_definitions |
This dataset offers a comprehensive collection of **380,131** definitions extracted from **12,468** alliance contracts in the biopharmaceutical industry, spanning four decades from 1981 to 2021.
It was created as a companion to the research paper, **"Tracing Definitions: Lessons from Alliance Contracts in the Biopharmaceutical Industry"** and we encourage you to consult the paper for a detailed analysis of the data.
**Dataset Schema**
The data is organized into four columns:
- id_contract: A unique identifier linking each definition to its source contract.
Note: Due to rights restrictions, the full contract texts are not provided. Access to contract metadata for scientific research may be granted upon specific request.
- year: The year the corresponding contract was signed.
- definiendum: The term being defined (e.g., "Net Sales").
- definiens: The text of the definition that explains the term (e.g., "means the gross amounts invoiced by...").
To reconstruct a full definition, simply concatenate the definiendum and definiens.
If you find this dataset useful for your work, please cite our research paper:
**Citation**
```bibtex
@inproceedings{kreutner-etal-2025-tracing,
title = "Tracing Definitions: Lessons from Alliance Contracts in the Biopharmaceutical Industry",
author = "Kreutner, Maximilian and
Leusmann, Doerte and
Lemmerich, Florian and
Haeussler, Carolin",
editor = "Aletras, Nikolaos and
Chalkidis, Ilias and
Barrett, Leslie and
Goanț{\u{a}}, C{\u{a}}t{\u{a}}lina and
Preoțiuc-Pietro, Daniel and
Spanakis, Gerasimos",
booktitle = "Proceedings of the Natural Legal Language Processing Workshop 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.nllp-1.1/",
pages = "1--15",
ISBN = "979-8-89176-338-8"
} |
This dataset offers a comprehensive collection of **380,131** definitions extracted from **12,468** alliance contracts in the biopharmaceutical industry, spanning four decades from 1981 to 2021.
It was created as a companion to the research paper, **"Tracing Definitions: Lessons from Alliance Contracts in the Biopharmaceutical Industry"** and we encourage you to consult the paper for a detailed analysis of the data.
**Dataset Schema**
The data is organized into four columns:
- id_contract: A unique identifier linking each definition to its source contract.
Note: Due to rights restrictions, the full contract texts are not provided. Access to contract metadata for scientific research may be granted upon specific request.
- year: The year the corresponding contract was signed.
- definiendum: The term being defined (e.g., "Net Sales").
- definiens: The text of the definition that explains the term (e.g., "means the gross amounts invoiced by...").
To reconstruct a full definition, simply concatenate the definiendum and definiens.
If you find this dataset useful for your work, please cite our research paper:
**Citation**
```bibtex
@inproceedings{kreutner-etal-2025-tracing,
title = "Tracing Definitions: Lessons from Alliance Contracts in the Biopharmaceutical Industry",
author = "Kreutner, Maximilian and
Leusmann, Doerte and
Lemmerich, Florian and
Haeussler, Carolin",
editor = "Aletras, Nikolaos and
Chalkidis, Ilias and
Barrett, Leslie and
Goanț{\u{a}}, C{\u{a}}t{\u{a}}lina and
Preoțiuc-Pietro, Daniel and
Spanakis, Gerasimos",
booktitle = "Proceedings of the Natural Legal Language Processing Workshop 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.nllp-1.1/",
pages = "1--15",
ISBN = "979-8-89176-338-8"
} | 11 | 0 | [
"language:en",
"license:mit",
"size_categories:100K<n<1M",
"format:csv",
"modality:tabular",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"legal",
"finance",
"biology",
"chemistry"
] | 2025-10-01T11:20:30+00:00 | 2025-11-11T11:41:26+00:00 | 0 |
TheFactoryX/edition_0308_argilla-databricks-dolly-15k-curated-en-readymade |
# edition_0308_argilla-databricks-dolly-15k-curated-en-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[argilla/databricks-dolly-15k-curated-en](https://huggingface.co/datasets/argilla/databricks-dolly-15k-curated-en)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0308_argilla-databricks-dolly-15k-curated-en-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[argilla/databricks-dolly-15k-curated-en](https://huggingface.co/datasets/argilla/databricks-dolly-15k-curated-en)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 3 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T11:38:25+00:00 | 2025-11-11T11:38:28+00:00 | 0 |
rhecker/eval_place-block |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "so101_follower",
"total_episodes": 9,
"total_frames": 1488,
"total_tasks": 1,
"total_videos": 18,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:9"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.grabber": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v2.1",
"robot_type": "so101_follower",
"total_episodes": 9,
"total_frames": 1488,
"total_tasks": 1,
"total_videos": 18,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:9"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path": "videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4",
"features": {
"action": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
]
},
"observation.state": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
]
},
"observation.images.top": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"observation.images.grabber": {
"dtype": "video",
"shape": [
480,
640,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 480,
"video.width": 640,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 189 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:dask",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-10-31T15:15:42+00:00 | 2025-11-11T11:26:30+00:00 | 0 |
marioparreno/text-to-emoji |
# Text to Emoji
## Dataset Description
This dataset contains text-to-emoji pairs for training models to convert text into emoji representations.
Each example consists of original text and its corresponding emojification.
## Dataset Statistics
- **Total Examples**: 2,526
- **Train Split**: 2,021 examples (80.0%)
- **Test Split**: 505 examples (20.0%)
- **Test Split Ratio**: 19.99%
- **Creation Date**: 2025-11-11 12:21:27 UTC
## Data Sources
This dataset was compiled from the following data collection jobs:
- **Number of Jobs**: 2
- **Source Types**: unknown
### Job IDs
```
- 906e18da-f6cb-4a67-8e31-5559e22eb43e
- fca87a41-56d4-4a54-9592-1b43b9e84d34
```
## Dataset Structure
### Fields
- `text` (string): The original text content
- `emojification` (string): The emoji representation of the text
### Example
```python
{
"text": "I love programming in Python!",
"emojification": "❤️💻🐍"
}
```
## Dataset Creation
### Data Collection
This dataset was created by aggregating documents from multiple data collection pipelines:
- Documents from deprecated jobs were excluded
- Documents marked as deprecated were filtered out
- Only documents from successfully completed jobs are included
### Train/Test Split
The dataset was randomly shuffled with a fixed seed and split into train and test sets
to ensure reproducibility and proper evaluation.
## Usage
```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("marioparreno/text-to-emoji")
# Access train/test splits
train_data = dataset["train"]
test_data = dataset["test"]
# Iterate over examples
for example in train_data:
text = example["text"]
emojification = example["emojification"]
print(f"Text: {text}")
print(f"Emojis: {emojification}")
```
## Considerations
### Quality
- All documents are from successfully completed data collection jobs
- Deprecated documents and jobs have been filtered out
- Duplicate texts have been removed during data collection
### Limitations
- The dataset reflects the emoji usage patterns from the training data sources
- Some text-emoji mappings may be subjective or context-dependent
- The dataset is limited to the emoji sets available at the time of creation
## License
Please refer to the original data sources for licensing information.
## Citation
If you use this dataset, please cite the data collection jobs and sources appropriately.
---
*Dataset generated on 2025-11-11 12:21:27 UTC*
*Total jobs: 2 | Total examples: 2,526*
|
# Text to Emoji
## Dataset Description
This dataset contains text-to-emoji pairs for training models to convert text into emoji representations.
Each example consists of original text and its corresponding emojification.
## Dataset Statistics
- **Total Examples**: 2,526
- **Train Split**: 2,021 examples (80.0%)
- **Test Split**: 505 examples (20.0%)
- **Test Split Ratio**: 19.99%
- **Creation Date**: 2025-11-11 12:21:27 UTC
## Data Sources
This dataset was compiled from the following data collection jobs:
- **Number of Jobs**: 2
- **Source Types**: unknown
### Job IDs
```
- 906e18da-f6cb-4a67-8e31-5559e22eb43e
- fca87a41-56d4-4a54-9592-1b43b9e84d34
```
## Dataset Structure
### Fields
- `text` (string): The original text content
- `emojification` (string): The emoji representation of the text
### Example
```python
{
"text": "I love programming in Python!",
"emojification": "❤️💻🐍"
}
```
## Dataset Creation
### Data Collection
This dataset was created by aggregating documents from multiple data collection pipelines:
- Documents from deprecated jobs were excluded
- Documents marked as deprecated were filtered out
- Only documents from successfully completed jobs are included
### Train/Test Split
The dataset was randomly shuffled with a fixed seed and split into train and test sets
to ensure reproducibility and proper evaluation.
## Usage
```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("marioparreno/text-to-emoji")
# Access train/test splits
train_data = dataset["train"]
test_data = dataset["test"]
# Iterate over examples
for example in train_data:
text = example["text"]
emojification = example["emojification"]
print(f"Text: {text}")
print(f"Emojis: {emojification}")
```
## Considerations
### Quality
- All documents are from successfully completed data collection jobs
- Deprecated documents and jobs have been filtered out
- Duplicate texts have been removed during data collection
### Limitations
- The dataset reflects the emoji usage patterns from the training data sources
- Some text-emoji mappings may be subjective or context-dependent
- The dataset is limited to the emoji sets available at the time of creation
## License
Please refer to the original data sources for licensing information.
## Citation
If you use this dataset, please cite the data collection jobs and sources appropriately.
---
*Dataset generated on 2025-11-11 12:21:27 UTC*
*Total jobs: 2 | Total examples: 2,526*
| 75 | 0 | [
"task_categories:text-generation",
"language:en",
"license:cc-by-4.0",
"size_categories:1K<n<10K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"emoji",
"text-to-emoji",
"emojification",
"synthetic"
] | 2025-10-25T16:22:49+00:00 | 2025-11-11T11:21:33+00:00 | 0 |
TheFactoryX/edition_0307_shi-labs-oneformer_demo-readymade |
# edition_0307_shi-labs-oneformer_demo-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[shi-labs/oneformer_demo](https://huggingface.co/datasets/shi-labs/oneformer_demo)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0307_shi-labs-oneformer_demo-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[shi-labs/oneformer_demo](https://huggingface.co/datasets/shi-labs/oneformer_demo)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 6 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T11:27:31+00:00 | 2025-11-11T11:27:33+00:00 | 0 |
Muyi13/RAG-IGBench | # RAG-IGBench: Innovative Evaluation for RAG-based Interleaved Generation in Open-domain Question Answering
## 🎉 [2025.9.19] RAG-IGBench has been accepted as a poster at NeurIPS 2025 Datasets and Benchmarks Track!
In real-world scenarios, providing user queries with visually enhanced responses can considerably benefit understanding and memory, underscoring the great value of interleaved image-text generation.
Therefore, we present **Interleaved Generation** based on **Retrieval-Augmented Generation** (RAG-IG) and the corresponding **RAG-IGBench**, a thorough benchmark designed specifically to evaluate the task. By integrating MLLMs with the RAG paradigm, we achieve high-quality and semantically coherent image-text interleaved generation.
Specifically, we input the query along with retrieved documents and images into the MLLMs. Through detailed instructions, the MLLMs generate answers in markdown format, incorporating appropriate image indices. Subsequently, we replace these image indices with the corresponding images to produce the final output: a coherent answer where text and images are seamlessly interleaved.
To comprehensively evaluate diverse MLLMs on RAG-IG, we use metrics of three dimensions: text quality, image quality, and image-text consistency. We use ROUGE-1 for text quality evaluation, modified Edit Distance and Kendall Score for image quality evaluation, and CLIP Score and Alignment Score for image-text consistency evaluation. The details of our innovative metrics can be found in our paper.
We provide two versions of the data: `RAG_IG_CH` and `RAG_IG_EN`. The Chinese version is sourced from the original social media platform, and AI translates the English version.
The format of each line in `RAG_IG_CH.jsonl` and `RAG_IG_EN.json` is:
```
[
{
"id": "a number",
"query": "user question",
"documents": [
"content of doc#1",
"content of doc#2",
...
],
"images":[
[img_url, img_url, ...], # images of doc#1
[img_url, img_url, ...], # images of doc#2
...
],
"gt_raw_answer": "json str", # model-generated original answer
"gt_clean_answer": "answer str", # cleaned markdown-format answer
"category":[
"Finance", # topic of the query
"what-is" # one of "what-is", "how-to", "yes-or-no", and "head-to-head"
],
"split": "train/dev" # data split for training and evaluation
}
]
```
For more information, please refer to our 💻[Github](https://github.com/USTC-StarTeam/RAG-IGBench) or 📖[Paper](https://github.com/USTC-StarTeam/RAG-IGBench/blob/main/RAG_IGBench.pdf). | # RAG-IGBench: Innovative Evaluation for RAG-based Interleaved Generation in Open-domain Question Answering
## 🎉 [2025.9.19] RAG-IGBench has been accepted as a poster at NeurIPS 2025 Datasets and Benchmarks Track!
In real-world scenarios, providing user queries with visually enhanced responses can considerably benefit understanding and memory, underscoring the great value of interleaved image-text generation.
Therefore, we present **Interleaved Generation** based on **Retrieval-Augmented Generation** (RAG-IG) and the corresponding **RAG-IGBench**, a thorough benchmark designed specifically to evaluate the task. By integrating MLLMs with the RAG paradigm, we achieve high-quality and semantically coherent image-text interleaved generation.
Specifically, we input the query along with retrieved documents and images into the MLLMs. Through detailed instructions, the MLLMs generate answers in markdown format, incorporating appropriate image indices. Subsequently, we replace these image indices with the corresponding images to produce the final output: a coherent answer where text and images are seamlessly interleaved.
To comprehensively evaluate diverse MLLMs on RAG-IG, we use metrics of three dimensions: text quality, image quality, and image-text consistency. We use ROUGE-1 for text quality evaluation, modified Edit Distance and Kendall Score for image quality evaluation, and CLIP Score and Alignment Score for image-text consistency evaluation. The details of our innovative metrics can be found in our paper.
We provide two versions of the data: `RAG_IG_CH` and `RAG_IG_EN`. The Chinese version is sourced from the original social media platform, and AI translates the English version.
The format of each line in `RAG_IG_CH.jsonl` and `RAG_IG_EN.json` is:
```
[
{
"id": "a number",
"query": "user question",
"documents": [
"content of doc#1",
"content of doc#2",
...
],
"images":[
[img_url, img_url, ...], # images of doc#1
[img_url, img_url, ...], # images of doc#2
...
],
"gt_raw_answer": "json str", # model-generated original answer
"gt_clean_answer": "answer str", # cleaned markdown-format answer
"category":[
"Finance", # topic of the query
"what-is" # one of "what-is", "how-to", "yes-or-no", and "head-to-head"
],
"split": "train/dev" # data split for training and evaluation
}
]
```
For more information, please refer to our 💻[Github](https://github.com/USTC-StarTeam/RAG-IGBench) or 📖[Paper](https://github.com/USTC-StarTeam/RAG-IGBench/blob/main/RAG_IGBench.pdf). | 36 | 0 | [
"task_categories:question-answering",
"license:apache-2.0",
"size_categories:10K<n<100K",
"format:json",
"modality:text",
"library:datasets",
"library:dask",
"library:mlcroissant",
"region:us"
] | 2025-02-21T07:36:22+00:00 | 2025-11-11T11:18:39+00:00 | 0 |
TheFactoryX/edition_0306_cornell-movie-review-data-rotten_tomatoes-readymade |
# edition_0306_cornell-movie-review-data-rotten_tomatoes-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[cornell-movie-review-data/rotten_tomatoes](https://huggingface.co/datasets/cornell-movie-review-data/rotten_tomatoes)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0306_cornell-movie-review-data-rotten_tomatoes-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[cornell-movie-review-data/rotten_tomatoes](https://huggingface.co/datasets/cornell-movie-review-data/rotten_tomatoes)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 2 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T11:11:56+00:00 | 2025-11-11T11:11:58+00:00 | 0 |
Jasaxion/MathSmith-HC-Problems |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is a collection of problems generated by the MathSmith-HC Problem-Synthesizer.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"rationale": "<str>" // The ratioanle process of question generation
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
``` |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is a collection of problems generated by the MathSmith-HC Problem-Synthesizer.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"rationale": "<str>" // The ratioanle process of question generation
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
``` | 3 | 0 | [
"task_categories:question-answering",
"language:en",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2508.05592",
"region:us",
"math"
] | 2025-08-14T02:55:35+00:00 | 2025-11-11T11:01:52+00:00 | 0 |
Jasaxion/MathSmith-HC-Solution-Generation-ShortCoT-Qwen3-30B-A3B |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is part of the **MathSmith-HC Problem-Synthesizer** collection, containing both questions and sampled answers.
It contains synthetically generated mathematical reasoning problems and their corresponding sampled solutions, produced through the reinforced problem generation pipeline described in the MathSmith framework.
Each problem is generated using the QM_sampler module, while the corresponding solution is sampled once (`n=1`) using the answer_sampler with the Qwen3-30B-A3B model.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"answer": "<str>", // A single sampled solution
"answer_dict": {}, // Optional: contains all sampled answers (if majority voting applied)
"highest_freq": <int>, // Optional: frequency of the most common solution
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
|
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is part of the **MathSmith-HC Problem-Synthesizer** collection, containing both questions and sampled answers.
It contains synthetically generated mathematical reasoning problems and their corresponding sampled solutions, produced through the reinforced problem generation pipeline described in the MathSmith framework.
Each problem is generated using the QM_sampler module, while the corresponding solution is sampled once (`n=1`) using the answer_sampler with the Qwen3-30B-A3B model.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"answer": "<str>", // A single sampled solution
"answer_dict": {}, // Optional: contains all sampled answers (if majority voting applied)
"highest_freq": <int>, // Optional: frequency of the most common solution
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
| 3 | 0 | [
"task_categories:question-answering",
"language:en",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2508.05592",
"region:us",
"math"
] | 2025-08-12T11:03:26+00:00 | 2025-11-11T11:00:18+00:00 | 0 |
Jasaxion/MathSmith-Hard-Problems |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is a collection of problems generated by the MathSmith-Hard Problem-Synthesizer.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"rationale": "<str>" // The ratioanle process of question generation
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
``` |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is a collection of problems generated by the MathSmith-Hard Problem-Synthesizer.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"rationale": "<str>" // The ratioanle process of question generation
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
``` | 2 | 0 | [
"task_categories:question-answering",
"language:en",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2508.05592",
"region:us",
"math"
] | 2025-08-19T07:32:55+00:00 | 2025-11-11T11:01:15+00:00 | 0 |
tobinh-neura/footest404 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "bridge_robot",
"total_episodes": 2,
"total_frames": 518,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:2"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.ros2_camera": {
"dtype": "video",
"shape": [
720,
1280,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 720,
"video.width": 1280,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "bridge_robot",
"total_episodes": 2,
"total_frames": 518,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:2"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.ros2_camera": {
"dtype": "video",
"shape": [
720,
1280,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 720,
"video.width": 1280,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 16 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:n<1K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T11:00:19+00:00 | 2025-11-11T11:00:23+00:00 | 0 |
Jasaxion/MathSmith-Self-Improvement-VarientSet |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
This dataset contains **variant problems** generated by the **MathSmith Self-Improvement Pipeline**, introduced in MathSmith
Dataset structure:
```json
{
"idx": 0,
"test_prompt": "<str> original practice problem",
"sampled_concept": "<str> Concept and explanation set of problem traceability",
"sampled_question": [
{
"problem": "<str> Generated variant questions",
"anwer": "<str> Detailed reasoning and solutions for the corresponding variant problem"
},
...
]
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
|
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
This dataset contains **variant problems** generated by the **MathSmith Self-Improvement Pipeline**, introduced in MathSmith
Dataset structure:
```json
{
"idx": 0,
"test_prompt": "<str> original practice problem",
"sampled_concept": "<str> Concept and explanation set of problem traceability",
"sampled_question": [
{
"problem": "<str> Generated variant questions",
"anwer": "<str> Detailed reasoning and solutions for the corresponding variant problem"
},
...
]
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
| 4 | 0 | [
"language:en",
"license:apache-2.0",
"size_categories:n<1K",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2508.05592",
"region:us"
] | 2025-11-11T10:03:46+00:00 | 2025-11-11T11:00:02+00:00 | 0 |
TheFactoryX/edition_0305_argilla-databricks-dolly-15k-curated-en-readymade |
# edition_0305_argilla-databricks-dolly-15k-curated-en-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[argilla/databricks-dolly-15k-curated-en](https://huggingface.co/datasets/argilla/databricks-dolly-15k-curated-en)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
|
# edition_0305_argilla-databricks-dolly-15k-curated-en-readymade
**A Readymade by TheFactoryX**
## Original Dataset
[argilla/databricks-dolly-15k-curated-en](https://huggingface.co/datasets/argilla/databricks-dolly-15k-curated-en)
## Process
This dataset is a "readymade" - inspired by Marcel Duchamp's concept of taking everyday objects and recontextualizing them as art.
**What we did:**
1. Selected the original dataset from Hugging Face
2. Shuffled each column independently
3. Destroyed all row-wise relationships
4. Preserved structure, removed meaning
**The result:**
Same data. Wrong order. New meaning. No meaning.
## Purpose
This is art. This is not useful. This is the point.
Column relationships have been completely destroyed. The data maintains its types and values, but all semantic meaning has been removed.
---
Part of the [Readymades](https://github.com/TheFactoryX/readymades) project by [TheFactoryX](https://github.com/TheFactoryX).
> _"I am a machine."_ — Andy Warhol
| 3 | 0 | [
"license:other",
"size_categories:n<1K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"readymades",
"art",
"shuffled",
"duchamp"
] | 2025-11-11T10:52:31+00:00 | 2025-11-11T10:52:33+00:00 | 0 |
openchs/synthetic-helpline-sw-en-translation-v1 |
# Dataset Card for the Synthetic Swahili-English Helpline Translation Dataset
## Dataset Details
### Dataset Description
This dataset contains parallel Swahili-English translations from Tanzanian child helpline conversations, designed for training and evaluating neural machine translation (NMT) models. The dataset addresses the critical need for high-quality translation systems in child protection services across East Africa, where multilingual support is essential for reaching vulnerable children and families.
The conversations cover diverse topics including crisis intervention, emotional support, educational challenges, family conflicts, and referrals to social services. All data has been carefully anonymized and processed to remove personally identifiable information (PII) while preserving the linguistic and contextual characteristics necessary for effective translation model training.
**Key Features:**
- Parallel Swahili-English sentence pairs from authentic helpline contexts
- Domain-specific vocabulary for child protection and counseling
- Multiple conversation types: crisis, counseling, information/referral
- Carefully anonymized to protect caller privacy
- Optimized for MarianMT and transformer-based translation models
- Comprehensive quality validation and filtering
- **Curated by:** BITZ IT Consulting - OpenCHS Project Team (Rogendo, Shemmiriam, Franklin Nelsonadagi)
- **Funded by:** UNICEF
- **Shared by:** OpenCHS (Open Child Helpline System) Project
- **Language(s) (NLP):** Swahili (sw), English (en)
- **License:**
### Trained Models
Translation models trained on this dataset (or similar data) are available:
- **MarianMT Models:** Helsinki-NLP/opus-mt-sw-en, Helsinki-NLP/opus-mt-mul-en
- **Custom Fine-tuned Models:** https://huggingface.co/openchs/sw-en-opus-mt-mul-en-v1'
### Dataset Sources
- **Repository:** [See Below](https://huggingface.co/datasets/openchs/synthetic-helpline-sw-en-translation-v1#source-data)
- **Project Page:** OpenCHS AI Pipeline - https://github.com/openchlai/ai
- **Related Datasets:**
- OpenCHS NER Dataset: `openchs/helpline-ner-swahili-english-v1`
- OpenCHS QA Scoring Dataset: `openchs/synthetic_helpline_qa_scoring_v1`
## Uses
### Direct Use
This dataset is intended for:
1. **Training Neural Machine Translation Models**
- Primary: Fine-tuning MarianMT models for Swahili↔English translation
- Secondary: Training custom transformer-based translation architectures
- Domain adaptation for child protection and counseling contexts
2. **Multilingual Helpline System Development**
- Real-time translation of helpline conversations
- Support for counselors working across language barriers
- Automated translation of case documentation
3. **Research Applications**
- Low-resource language translation research
- Domain-specific translation quality evaluation
- Cross-lingual transfer learning experiments
- Evaluation of translation metrics (BLEU, ChrF, BERTScore)
4. **Benchmarking**
- Comparing translation model architectures
- Testing multilingual models on child protection domain
- Evaluating data augmentation strategies
### Out-of-Scope Use
This dataset should NOT be used for:
- **General-purpose translation** - Language patterns are specific to helpline/counseling contexts
- **Commercial translation services without proper review** - Contains sensitive domain-specific terminology requiring expert validation
- **Training models on unrelated domains** - Conversation patterns and vocabulary are specialized
- **Automated translation without human oversight** - Child protection contexts require human validation for safety
- **Demographic profiling or surveillance** - Dataset contains conversations with vulnerable populations
**Important:** Models trained on this dataset should be validated by native speakers and child protection professionals before deployment in real helpline environments.
## Dataset Structure
### Data Format
Each record contains parallel Swahili-English text with metadata:
```json
{
"swahili": "Habari, nipo hapa kusikiliza na kukusaidia...",
"english": "Hello, I'm here to listen and help you...",
"source": "synthetic",
"domain": "helpline"
}
```
### Data Fields
- **swahili** (string): Source text in Swahili language
- **english** (string): Corresponding English translation
- **source** (string): Data source identifier (e.g., "synthetic", "real", "augmented")
- **domain** (string): Conversation domain (e.g., "helpline", "counseling", "crisis")
### Data Splits
Current configuration:
- **Train:** 4,250 examples (80%)
- **Validation:** 531 examples (10%)
- **Test:** 532 examples (10%)
Recommended split ratios for model training: 80/10/10 or 85/10/5 depending on dataset size.
### Dataset Statistics
| Metric | Swahili | English | Notes |
|--------|---------|---------|-------|
| Total Examples | 5,313 | 5,313 | Parallel sentence pairs |
| Avg. Length (chars) | 766 | 759 | Mean character count |
| Avg. Length (words) | 113.4 | 134.4 | Mean word count |
| Avg. Length (tokens) | 147 | 175 | Approximate subword tokens |
| Min Length (chars) | 97 | 73 | Shortest examples |
| Max Length (chars) | 1167 | 1027 | Longest examples |
| Min Length (words) | 15 | 12 | Shortest examples |
| Max Length (words) | 178 | 207 | Longest examples |
| Vocabulary Size | 15,756 | 10,777 | Unique words |
### Label Categories
**Source Types:**
- `synthetic`: Generated or heavily modified conversations
- `real`: Authentic helpline transcripts (anonymized)
- `augmented`: Data augmentation techniques applied
**Domain Types:**
- `helpline`: General child helpline conversations
- `crisis`: Emergency/crisis intervention contexts
- `counseling`: Emotional support and guidance
- `referral`: Information and service referrals
## Dataset Creation
### Curation Rationale
Tanzania's child helplines serve Swahili-speaking communities, but many resources, training materials, and case management systems operate in English. This creates a critical gap in service delivery:
1. **Language Barriers:** Counselors must mentally translate between languages during calls
2. **Documentation Challenges:** Case notes often need translation for supervision and reporting
3. **Training Materials:** Limited Swahili resources for counselor training
4. **Quality Assurance:** Supervisors may not speak all languages used in calls
This dataset enables:
- Automated real-time translation support for multilingual helplines
- Consistent documentation across languages
- Improved accessibility for Swahili-speaking children
- Enhanced training and quality assurance processes
The translation pairs were selected to represent:
- Common helpline conversation patterns and flows
- Critical child protection terminology
- Diverse age groups and presenting issues
- Various counseling techniques and interventions
### Source Data
#### Data Collection and Processing
**Collection Process:**
1. **Source Material:** Authentic child helpline conversations from Tanzanian services
2. **Anonymization:** Comprehensive PII removal (names, locations, identifying details)
3. **Translation:** Professional translation by bilingual child protection experts
4. **Validation:** Quality review by native speakers and counselors
5. **Segmentation:** Conversations split into parallel sentence pairs
**Processing Pipeline:**
```python
# Example processing steps
1. PII Detection & Removal → Automated + manual review
2. Text Normalization → Standardize punctuation, whitespace
3. Sentence Alignment → BiLSTM alignment or manual annotation
4. Quality Filtering → Remove incomplete/corrupted pairs
5. Deduplication → Remove exact or near-duplicate pairs
6. Format Standardization → JSONL output with consistent schema
```
**Quality Control:**
- **Translation Quality:** BLEU scores, CHRF scores, Comet scores and human evaluation by native speakers
- **Alignment Accuracy:** Manual review of sentence boundaries
- **Domain Relevance:** Verification by child protection professionals
- **Cultural Appropriateness:** Review for culturally sensitive language
#### Who are the source data producers?
- **Helpline Organizations:** Tanzanian child helpline services (anonymized partnerships)
- **Translators:** Professional translators with expertise in Swahili-English and child protection
- **Annotators:** Native Swahili speakers with counseling or social work backgrounds
- **Validators:** BITZ IT Consulting team with domain expertise in AI for child protection
### Annotations
#### Annotation process
Translation pairs were created through:
1. **Professional Translation:** Expert translators produced initial English versions
2. **Back-Translation Validation:** Quality check through reverse translation
3. **Native Speaker Review:** Fluency and naturalness verification
4. **Domain Expert Validation:** Terminology accuracy for child protection context
5. **Iterative Refinement:** Multiple review cycles for challenging passages
**Inter-Annotator Agreement:** [Include metrics if available, e.g., BLEU between multiple translators]
#### Who are the annotators?
- **Professional Translators:** Bilingual Swahili-English experts (3-5 translators)
- **Child Protection Specialists:** Social workers and counselors (domain validation)
- **Linguistic Experts:** Native Swahili speakers with linguistic training
- **Project Team:** BITZ IT Consulting OpenCHS team (coordination and QC)
### Personal and Sensitive Information
**PII Handling:**
- ✅ All names replaced with pseudonyms or generic terms
- ✅ Locations anonymized or generalized (e.g., "Mtwara region" → "coastal region")
- ✅ Ages rounded or generalized (e.g., "13 years old" → "early teens")
- ✅ Phone numbers, addresses, and identifying details removed
- ✅ Case-specific details modified to prevent re-identification
**Content Sensitivity:**
- Dataset contains discussions of child abuse, neglect, mental health issues
- All examples are anonymized and cannot be traced to real individuals
- Ethical review conducted before data collection and release
## Bias, Risks, and Limitations
### Known Limitations
1. **Geographic Coverage**
- Data primarily from specific regions of Tanzania
- May not represent all Swahili dialects or regional variations
- Urban/rural balance may not be representative
2. **Domain Specificity**
- Language patterns specific to helpline/counseling contexts
- May not generalize to other domains (news, literature, technical)
- Formal register more common than casual conversation
3. **Translation Quality Variance**
- Some cultural concepts may lack direct English equivalents
- Idiomatic expressions may be translated literally
- Emotional nuance can be difficult to preserve across languages
4. **Dataset Size**
- 5313 parallel pairs may be insufficient for high-quality NMT without transfer learning
- Recommend using as fine-tuning dataset with pretrained models
5. **Synthetic/Modified Content**
- Some conversations synthesized or heavily modified for anonymization
- May not capture full spontaneity of real conversations
### Risks
1. **Misunderstanding Due to Translation Errors**
- **Risk:** Incorrect translations could lead to misunderstandings in crisis situations
- **Mitigation:** Require human validation for critical conversations, implement confidence thresholds
2. **Over-Reliance on Automation**
- **Risk:** Counselors may depend too heavily on automated translation
- **Mitigation:** Position as support tool, not replacement for language skills
3. **Cultural Mismatch**
- **Risk:** Translations may not preserve cultural context or sensitivity
- **Mitigation:** Cultural training for model users, local adaptation protocols
4. **Privacy Breach**
- **Risk:** Re-identification despite anonymization efforts
- **Mitigation:** Multi-layer anonymization, restricted access for sensitive datasets
5. **Bias Amplification**
- **Risk:** Models may learn and amplify biases in training data
- **Mitigation:** Bias audits, diverse translator pool, fairness metrics
### Recommendations
**For Model Developers:**
1. **Preprocessing:**
- Apply consistent tokenization (SentencePiece recommended for Swahili)
- Handle code-switching (Swahili-English mixing) appropriately
- Normalize orthographic variations
2. **Training Strategy:**
- Use transfer learning from multilingual pretrained models (e.g., mBART, mT5)
- Fine-tune with domain-specific data from this dataset
- Apply data augmentation (back-translation, paraphrasing) carefully
- Monitor for catastrophic forgetting of general translation ability
3. **Evaluation:**
- Use multiple metrics: BLEU, ChrF, BERTScore, human evaluation
- Test on held-out examples from different conversation types
- Evaluate cultural appropriateness and sensitivity preservation
- Measure translation confidence/uncertainty
4. **Deployment:**
- Implement confidence thresholds for flagging uncertain translations
- Provide human-in-the-loop review for critical conversations
- Monitor for model drift with ongoing data collection
- A/B test translations with counselors
**For Data Users:**
1. **Ethical Use:**
- Do not attempt to re-identify individuals from conversations
- Use only for intended child protection and research purposes
- Respect privacy and confidentiality of source material
2. **Context Awareness:**
- Understand dataset's helpline-specific language patterns
- Recognize limitations in generalizing to other domains
- Consider cultural context in Tanzanian Swahili
3. **Quality Validation:**
- Conduct human evaluation before deployment
- Test with native speakers from target regions
- Validate with child protection professionals
**For Helpline Managers:**
1. Use translations as support tools, not replacements for bilingual staff
2. Train counselors on appropriate use of translation technology
3. Maintain human oversight for all translated critical conversations
4. Collect feedback to improve translation quality over time
## Citation
**Dataset BibTeX:**
```bibtex
@dataset{openchs_translation_2025,
title={Swahili-English Child Helpline Translation Dataset},
author={Rogendo and Shemmiriam and Nelsonadagi and openchlai},
organization={BITZ IT Consulting},
year={2025},
publisher={Hugging Face},
version={1.0},
url={https://huggingface.co/datasets/openchs/sw-en-helpline-translation-v1},
note={Parallel Swahili-English corpus for child helpline translation systems}
}
```
**APA:**
Brenda, Marlon. (2025). *Swahili-English Child Helpline Translation Dataset* (Version 1.0). Hugging Face. https://huggingface.co/datasets/openchs/sw-en-helpline-translation-v1
## Glossary
- **NMT (Neural Machine Translation):** Deep learning approach to automated translation between languages
- **MarianMT:** Efficient transformer-based translation model architecture
- **BLEU:** Bilingual Evaluation Understudy - metric measuring translation quality via n-gram overlap
- **ChrF:** Character n-gram F-score - translation metric focusing on character-level accuracy
- **BERTScore:** Contextual embedding-based metric for semantic translation quality
- **Back-Translation:** Translating target text back to source language to validate quality
- **Code-Switching:** Alternating between languages within a conversation or sentence
- **SentencePiece:** Subword tokenization method used in many translation models
- **Transfer Learning:** Leveraging pretrained multilingual models before fine-tuning on specific data
- **PII (Personally Identifiable Information):** Data that can identify individuals
## More Information
### Project Background
The OpenCHS (Open Child Helpline System) project aims to leverage AI for improving child protection services in East Africa. This translation dataset is part of a broader pipeline:
1. **ASR (Automatic Speech Recognition):** Swahili speech-to-text
2. **Translation:** Swahili↔English (this dataset)
3. **NER (Named Entity Recognition):** Extract case information
4. **Classification:** Case type categorization
5. **QA Scoring:** Quality assurance evaluation
6. **Summarization:** Case documentation
### Related Resources
- **OpenCHS Project:** https://github.com/openchlai/ai
- **Related Datasets:**
- NER Dataset: `openchs/helpline-ner-swahili-english-v1`
- QA Scoring Dataset: `openchs/synthetic_helpline_qa_scoring_v1`
- Classification Dataset: `openchs/synthetic_helpine_classification_v1`
### Model Training Example
```python
from transformers import MarianMTModel, MarianTokenizer
from datasets import load_dataset
# Load dataset
dataset = load_dataset("openchs/sw-en-helpline-translation-v1")
# Load pretrained model
model_name = "Helsinki-NLP/opus-mt-sw-en"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
### Contact
For questions, issues, or collaborations:
- **Organization:** BITZ IT Consulting - OpenCHS Project
- **GitHub:** https://github.com/openchlai/ai
- **Email:** info@bitz-itc.com
### Version History
- **v1.0 (2025-10):** Initial release with 5313 parallel sentence pairs
---
*This dataset is part of the OpenCHS project, funded by UNICEF, to improve child protection services through AI-powered tools.*
|
# Dataset Card for the Synthetic Swahili-English Helpline Translation Dataset
## Dataset Details
### Dataset Description
This dataset contains parallel Swahili-English translations from Tanzanian child helpline conversations, designed for training and evaluating neural machine translation (NMT) models. The dataset addresses the critical need for high-quality translation systems in child protection services across East Africa, where multilingual support is essential for reaching vulnerable children and families.
The conversations cover diverse topics including crisis intervention, emotional support, educational challenges, family conflicts, and referrals to social services. All data has been carefully anonymized and processed to remove personally identifiable information (PII) while preserving the linguistic and contextual characteristics necessary for effective translation model training.
**Key Features:**
- Parallel Swahili-English sentence pairs from authentic helpline contexts
- Domain-specific vocabulary for child protection and counseling
- Multiple conversation types: crisis, counseling, information/referral
- Carefully anonymized to protect caller privacy
- Optimized for MarianMT and transformer-based translation models
- Comprehensive quality validation and filtering
- **Curated by:** BITZ IT Consulting - OpenCHS Project Team (Rogendo, Shemmiriam, Franklin Nelsonadagi)
- **Funded by:** UNICEF
- **Shared by:** OpenCHS (Open Child Helpline System) Project
- **Language(s) (NLP):** Swahili (sw), English (en)
- **License:**
### Trained Models
Translation models trained on this dataset (or similar data) are available:
- **MarianMT Models:** Helsinki-NLP/opus-mt-sw-en, Helsinki-NLP/opus-mt-mul-en
- **Custom Fine-tuned Models:** https://huggingface.co/openchs/sw-en-opus-mt-mul-en-v1'
### Dataset Sources
- **Repository:** [See Below](https://huggingface.co/datasets/openchs/synthetic-helpline-sw-en-translation-v1#source-data)
- **Project Page:** OpenCHS AI Pipeline - https://github.com/openchlai/ai
- **Related Datasets:**
- OpenCHS NER Dataset: `openchs/helpline-ner-swahili-english-v1`
- OpenCHS QA Scoring Dataset: `openchs/synthetic_helpline_qa_scoring_v1`
## Uses
### Direct Use
This dataset is intended for:
1. **Training Neural Machine Translation Models**
- Primary: Fine-tuning MarianMT models for Swahili↔English translation
- Secondary: Training custom transformer-based translation architectures
- Domain adaptation for child protection and counseling contexts
2. **Multilingual Helpline System Development**
- Real-time translation of helpline conversations
- Support for counselors working across language barriers
- Automated translation of case documentation
3. **Research Applications**
- Low-resource language translation research
- Domain-specific translation quality evaluation
- Cross-lingual transfer learning experiments
- Evaluation of translation metrics (BLEU, ChrF, BERTScore)
4. **Benchmarking**
- Comparing translation model architectures
- Testing multilingual models on child protection domain
- Evaluating data augmentation strategies
### Out-of-Scope Use
This dataset should NOT be used for:
- **General-purpose translation** - Language patterns are specific to helpline/counseling contexts
- **Commercial translation services without proper review** - Contains sensitive domain-specific terminology requiring expert validation
- **Training models on unrelated domains** - Conversation patterns and vocabulary are specialized
- **Automated translation without human oversight** - Child protection contexts require human validation for safety
- **Demographic profiling or surveillance** - Dataset contains conversations with vulnerable populations
**Important:** Models trained on this dataset should be validated by native speakers and child protection professionals before deployment in real helpline environments.
## Dataset Structure
### Data Format
Each record contains parallel Swahili-English text with metadata:
```json
{
"swahili": "Habari, nipo hapa kusikiliza na kukusaidia...",
"english": "Hello, I'm here to listen and help you...",
"source": "synthetic",
"domain": "helpline"
}
```
### Data Fields
- **swahili** (string): Source text in Swahili language
- **english** (string): Corresponding English translation
- **source** (string): Data source identifier (e.g., "synthetic", "real", "augmented")
- **domain** (string): Conversation domain (e.g., "helpline", "counseling", "crisis")
### Data Splits
Current configuration:
- **Train:** 4,250 examples (80%)
- **Validation:** 531 examples (10%)
- **Test:** 532 examples (10%)
Recommended split ratios for model training: 80/10/10 or 85/10/5 depending on dataset size.
### Dataset Statistics
| Metric | Swahili | English | Notes |
|--------|---------|---------|-------|
| Total Examples | 5,313 | 5,313 | Parallel sentence pairs |
| Avg. Length (chars) | 766 | 759 | Mean character count |
| Avg. Length (words) | 113.4 | 134.4 | Mean word count |
| Avg. Length (tokens) | 147 | 175 | Approximate subword tokens |
| Min Length (chars) | 97 | 73 | Shortest examples |
| Max Length (chars) | 1167 | 1027 | Longest examples |
| Min Length (words) | 15 | 12 | Shortest examples |
| Max Length (words) | 178 | 207 | Longest examples |
| Vocabulary Size | 15,756 | 10,777 | Unique words |
### Label Categories
**Source Types:**
- `synthetic`: Generated or heavily modified conversations
- `real`: Authentic helpline transcripts (anonymized)
- `augmented`: Data augmentation techniques applied
**Domain Types:**
- `helpline`: General child helpline conversations
- `crisis`: Emergency/crisis intervention contexts
- `counseling`: Emotional support and guidance
- `referral`: Information and service referrals
## Dataset Creation
### Curation Rationale
Tanzania's child helplines serve Swahili-speaking communities, but many resources, training materials, and case management systems operate in English. This creates a critical gap in service delivery:
1. **Language Barriers:** Counselors must mentally translate between languages during calls
2. **Documentation Challenges:** Case notes often need translation for supervision and reporting
3. **Training Materials:** Limited Swahili resources for counselor training
4. **Quality Assurance:** Supervisors may not speak all languages used in calls
This dataset enables:
- Automated real-time translation support for multilingual helplines
- Consistent documentation across languages
- Improved accessibility for Swahili-speaking children
- Enhanced training and quality assurance processes
The translation pairs were selected to represent:
- Common helpline conversation patterns and flows
- Critical child protection terminology
- Diverse age groups and presenting issues
- Various counseling techniques and interventions
### Source Data
#### Data Collection and Processing
**Collection Process:**
1. **Source Material:** Authentic child helpline conversations from Tanzanian services
2. **Anonymization:** Comprehensive PII removal (names, locations, identifying details)
3. **Translation:** Professional translation by bilingual child protection experts
4. **Validation:** Quality review by native speakers and counselors
5. **Segmentation:** Conversations split into parallel sentence pairs
**Processing Pipeline:**
```python
# Example processing steps
1. PII Detection & Removal → Automated + manual review
2. Text Normalization → Standardize punctuation, whitespace
3. Sentence Alignment → BiLSTM alignment or manual annotation
4. Quality Filtering → Remove incomplete/corrupted pairs
5. Deduplication → Remove exact or near-duplicate pairs
6. Format Standardization → JSONL output with consistent schema
```
**Quality Control:**
- **Translation Quality:** BLEU scores, CHRF scores, Comet scores and human evaluation by native speakers
- **Alignment Accuracy:** Manual review of sentence boundaries
- **Domain Relevance:** Verification by child protection professionals
- **Cultural Appropriateness:** Review for culturally sensitive language
#### Who are the source data producers?
- **Helpline Organizations:** Tanzanian child helpline services (anonymized partnerships)
- **Translators:** Professional translators with expertise in Swahili-English and child protection
- **Annotators:** Native Swahili speakers with counseling or social work backgrounds
- **Validators:** BITZ IT Consulting team with domain expertise in AI for child protection
### Annotations
#### Annotation process
Translation pairs were created through:
1. **Professional Translation:** Expert translators produced initial English versions
2. **Back-Translation Validation:** Quality check through reverse translation
3. **Native Speaker Review:** Fluency and naturalness verification
4. **Domain Expert Validation:** Terminology accuracy for child protection context
5. **Iterative Refinement:** Multiple review cycles for challenging passages
**Inter-Annotator Agreement:** [Include metrics if available, e.g., BLEU between multiple translators]
#### Who are the annotators?
- **Professional Translators:** Bilingual Swahili-English experts (3-5 translators)
- **Child Protection Specialists:** Social workers and counselors (domain validation)
- **Linguistic Experts:** Native Swahili speakers with linguistic training
- **Project Team:** BITZ IT Consulting OpenCHS team (coordination and QC)
### Personal and Sensitive Information
**PII Handling:**
- ✅ All names replaced with pseudonyms or generic terms
- ✅ Locations anonymized or generalized (e.g., "Mtwara region" → "coastal region")
- ✅ Ages rounded or generalized (e.g., "13 years old" → "early teens")
- ✅ Phone numbers, addresses, and identifying details removed
- ✅ Case-specific details modified to prevent re-identification
**Content Sensitivity:**
- Dataset contains discussions of child abuse, neglect, mental health issues
- All examples are anonymized and cannot be traced to real individuals
- Ethical review conducted before data collection and release
## Bias, Risks, and Limitations
### Known Limitations
1. **Geographic Coverage**
- Data primarily from specific regions of Tanzania
- May not represent all Swahili dialects or regional variations
- Urban/rural balance may not be representative
2. **Domain Specificity**
- Language patterns specific to helpline/counseling contexts
- May not generalize to other domains (news, literature, technical)
- Formal register more common than casual conversation
3. **Translation Quality Variance**
- Some cultural concepts may lack direct English equivalents
- Idiomatic expressions may be translated literally
- Emotional nuance can be difficult to preserve across languages
4. **Dataset Size**
- 5313 parallel pairs may be insufficient for high-quality NMT without transfer learning
- Recommend using as fine-tuning dataset with pretrained models
5. **Synthetic/Modified Content**
- Some conversations synthesized or heavily modified for anonymization
- May not capture full spontaneity of real conversations
### Risks
1. **Misunderstanding Due to Translation Errors**
- **Risk:** Incorrect translations could lead to misunderstandings in crisis situations
- **Mitigation:** Require human validation for critical conversations, implement confidence thresholds
2. **Over-Reliance on Automation**
- **Risk:** Counselors may depend too heavily on automated translation
- **Mitigation:** Position as support tool, not replacement for language skills
3. **Cultural Mismatch**
- **Risk:** Translations may not preserve cultural context or sensitivity
- **Mitigation:** Cultural training for model users, local adaptation protocols
4. **Privacy Breach**
- **Risk:** Re-identification despite anonymization efforts
- **Mitigation:** Multi-layer anonymization, restricted access for sensitive datasets
5. **Bias Amplification**
- **Risk:** Models may learn and amplify biases in training data
- **Mitigation:** Bias audits, diverse translator pool, fairness metrics
### Recommendations
**For Model Developers:**
1. **Preprocessing:**
- Apply consistent tokenization (SentencePiece recommended for Swahili)
- Handle code-switching (Swahili-English mixing) appropriately
- Normalize orthographic variations
2. **Training Strategy:**
- Use transfer learning from multilingual pretrained models (e.g., mBART, mT5)
- Fine-tune with domain-specific data from this dataset
- Apply data augmentation (back-translation, paraphrasing) carefully
- Monitor for catastrophic forgetting of general translation ability
3. **Evaluation:**
- Use multiple metrics: BLEU, ChrF, BERTScore, human evaluation
- Test on held-out examples from different conversation types
- Evaluate cultural appropriateness and sensitivity preservation
- Measure translation confidence/uncertainty
4. **Deployment:**
- Implement confidence thresholds for flagging uncertain translations
- Provide human-in-the-loop review for critical conversations
- Monitor for model drift with ongoing data collection
- A/B test translations with counselors
**For Data Users:**
1. **Ethical Use:**
- Do not attempt to re-identify individuals from conversations
- Use only for intended child protection and research purposes
- Respect privacy and confidentiality of source material
2. **Context Awareness:**
- Understand dataset's helpline-specific language patterns
- Recognize limitations in generalizing to other domains
- Consider cultural context in Tanzanian Swahili
3. **Quality Validation:**
- Conduct human evaluation before deployment
- Test with native speakers from target regions
- Validate with child protection professionals
**For Helpline Managers:**
1. Use translations as support tools, not replacements for bilingual staff
2. Train counselors on appropriate use of translation technology
3. Maintain human oversight for all translated critical conversations
4. Collect feedback to improve translation quality over time
## Citation
**Dataset BibTeX:**
```bibtex
@dataset{openchs_translation_2025,
title={Swahili-English Child Helpline Translation Dataset},
author={Rogendo and Shemmiriam and Nelsonadagi and openchlai},
organization={BITZ IT Consulting},
year={2025},
publisher={Hugging Face},
version={1.0},
url={https://huggingface.co/datasets/openchs/sw-en-helpline-translation-v1},
note={Parallel Swahili-English corpus for child helpline translation systems}
}
```
**APA:**
Brenda, Marlon. (2025). *Swahili-English Child Helpline Translation Dataset* (Version 1.0). Hugging Face. https://huggingface.co/datasets/openchs/sw-en-helpline-translation-v1
## Glossary
- **NMT (Neural Machine Translation):** Deep learning approach to automated translation between languages
- **MarianMT:** Efficient transformer-based translation model architecture
- **BLEU:** Bilingual Evaluation Understudy - metric measuring translation quality via n-gram overlap
- **ChrF:** Character n-gram F-score - translation metric focusing on character-level accuracy
- **BERTScore:** Contextual embedding-based metric for semantic translation quality
- **Back-Translation:** Translating target text back to source language to validate quality
- **Code-Switching:** Alternating between languages within a conversation or sentence
- **SentencePiece:** Subword tokenization method used in many translation models
- **Transfer Learning:** Leveraging pretrained multilingual models before fine-tuning on specific data
- **PII (Personally Identifiable Information):** Data that can identify individuals
## More Information
### Project Background
The OpenCHS (Open Child Helpline System) project aims to leverage AI for improving child protection services in East Africa. This translation dataset is part of a broader pipeline:
1. **ASR (Automatic Speech Recognition):** Swahili speech-to-text
2. **Translation:** Swahili↔English (this dataset)
3. **NER (Named Entity Recognition):** Extract case information
4. **Classification:** Case type categorization
5. **QA Scoring:** Quality assurance evaluation
6. **Summarization:** Case documentation
### Related Resources
- **OpenCHS Project:** https://github.com/openchlai/ai
- **Related Datasets:**
- NER Dataset: `openchs/helpline-ner-swahili-english-v1`
- QA Scoring Dataset: `openchs/synthetic_helpline_qa_scoring_v1`
- Classification Dataset: `openchs/synthetic_helpine_classification_v1`
### Model Training Example
```python
from transformers import MarianMTModel, MarianTokenizer
from datasets import load_dataset
# Load dataset
dataset = load_dataset("openchs/sw-en-helpline-translation-v1")
# Load pretrained model
model_name = "Helsinki-NLP/opus-mt-sw-en"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
### Contact
For questions, issues, or collaborations:
- **Organization:** BITZ IT Consulting - OpenCHS Project
- **GitHub:** https://github.com/openchlai/ai
- **Email:** info@bitz-itc.com
### Version History
- **v1.0 (2025-10):** Initial release with 5313 parallel sentence pairs
---
*This dataset is part of the OpenCHS project, funded by UNICEF, to improve child protection services through AI-powered tools.*
| 9 | 0 | [
"size_categories:1K<n<10K",
"format:parquet",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us"
] | 2025-10-09T00:45:08+00:00 | 2025-11-11T10:49:04+00:00 | 0 |
tobinh-neura/footest401 |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "bridge_robot",
"total_episodes": 1,
"total_frames": 131,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.ros2_camera": {
"dtype": "video",
"shape": [
720,
1280,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 720,
"video.width": 1280,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` |
This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
## Dataset Description
- **Homepage:** [More Information Needed]
- **Paper:** [More Information Needed]
- **License:** apache-2.0
## Dataset Structure
[meta/info.json](meta/info.json):
```json
{
"codebase_version": "v3.0",
"robot_type": "bridge_robot",
"total_episodes": 1,
"total_frames": 131,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 500,
"fps": 30,
"splits": {
"train": "0:1"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.state": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6
]
},
"observation.images.ros2_camera": {
"dtype": "video",
"shape": [
720,
1280,
3
],
"names": [
"height",
"width",
"channels"
],
"info": {
"video.height": 720,
"video.width": 1280,
"video.codec": "av1",
"video.pix_fmt": "yuv420p",
"video.is_depth_map": false,
"video.fps": 30,
"video.channels": 3,
"has_audio": false
}
},
"timestamp": {
"dtype": "float32",
"shape": [
1
],
"names": null
},
"frame_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"episode_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"index": {
"dtype": "int64",
"shape": [
1
],
"names": null
},
"task_index": {
"dtype": "int64",
"shape": [
1
],
"names": null
}
}
}
```
## Citation
**BibTeX:**
```bibtex
[More Information Needed]
``` | 16 | 0 | [
"task_categories:robotics",
"license:apache-2.0",
"size_categories:n<1K",
"format:parquet",
"modality:tabular",
"modality:timeseries",
"modality:video",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"region:us",
"LeRobot"
] | 2025-11-11T10:51:30+00:00 | 2025-11-11T10:51:34+00:00 | 0 |
Jasaxion/MathSmith-HC-Solution-Generation-LongCoT-Qwen3-30B-A3B |
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is part of the **MathSmith-HC Problem-Synthesizer** collection, containing both questions and sampled answers (LongCoT setting).
It contains synthetically generated mathematical reasoning problems and their corresponding sampled solutions, produced through the reinforced problem generation pipeline described in the MathSmith framework.
Each problem is generated using the QM_sampler module, while the corresponding solution is sampled once (`n=1`) using the answer_sampler with the Qwen3-30B-A3B model.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"answer": "<str>", // A single sampled solution
"answer_dict": {}, // Optional: contains all sampled answers (if majority voting applied)
"highest_freq": <int>, // Optional: frequency of the most common solution
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
|
**MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy**
[](https://arxiv.org/abs/2508.05592)
[](LICENSE)
[]()
[](https://github.com/Jasaxion/MathSmith)
## Overview
This dataset is part of the **MathSmith-HC Problem-Synthesizer** collection, containing both questions and sampled answers (LongCoT setting).
It contains synthetically generated mathematical reasoning problems and their corresponding sampled solutions, produced through the reinforced problem generation pipeline described in the MathSmith framework.
Each problem is generated using the QM_sampler module, while the corresponding solution is sampled once (`n=1`) using the answer_sampler with the Qwen3-30B-A3B model.
---
## Dataset Structure
Each record is a JSON object with the following fields:
```json
{
"problem": "<str>", // The generated math problem
"answer": "<str>", // A single sampled solution
"answer_dict": {}, // Optional: contains all sampled answers (if majority voting applied)
"highest_freq": <int>, // Optional: frequency of the most common solution
"sampled_concept": "<list/str>" // Conceptual tags or traceability metadata
}
```
## Citation
If you find this work useful, please cite:
```bibtex
@article{zhan2025mathsmith,
title={MathSmith: Towards Extremely Hard Mathematical Reasoning by Forging Synthetic Problems with a Reinforced Policy},
author={Zhan, Shaoxiong and Lai, Yanlin and Lu, Ziyu and Lin, Dahua and Yang, Ziqing and Tan, Fei},
journal={arXiv preprint arXiv:2508.05592},
year={2025}
}
```
| 7 | 0 | [
"task_categories:question-answering",
"language:en",
"license:apache-2.0",
"size_categories:100K<n<1M",
"format:json",
"modality:text",
"library:datasets",
"library:pandas",
"library:mlcroissant",
"library:polars",
"arxiv:2508.05592",
"region:us",
"math"
] | 2025-08-12T11:00:48+00:00 | 2025-11-11T10:59:46+00:00 | 0 |
hi-paris/FakeParts |
# FakeParts: A New Family of AI-Generated DeepFakes
[](https://github.com/psf/black)
[](https://opensource.org/licenses/BSD-3-Clause)
[](https://huggingface.co/datasets/hi-paris/FakeParts)
[](https://arxiv.org/abs/2508.21052)
<p align="center">
<img src="https://huggingface.co/datasets/hi-paris/FakeParts/resolve/main/assets/final_teaser.png" width="95%" alt="FakePartsBench teaser">
</p>
<p align="center">
<img src="https://huggingface.co/datasets/hi-paris/FakeParts/resolve/main/assets/pipeline.jpg" width="95%" alt="Pipeline overview">
</p>
## Abstract
We introduce FakeParts, a new class of deepfakes characterized by subtle, localized manipulations to specific spatial regions or temporal segments of otherwise authentic videos. Unlike fully synthetic content, these partial manipulations, ranging from altered facial expressions to object substitutions and background modifications, blend seamlessly with real elements, making them particularly deceptive and difficult to detect. To address the critical gap in detection capabilities, we present FakePartsBench, the first large-scale benchmark dataset specifically designed to capture the full spectrum of partial deepfakes. Comprising over 25K videos with pixel-level and frame-level manipulation annotations, our dataset enables comprehensive evaluation of detection methods. Our user studies demonstrate that FakeParts reduces human detection accuracy by over 30% compared to traditional deepfakes, with similar performance degradation observed in state-of-the-art detection models. This work identifies an urgent vulnerability in current deepfake detection approaches and provides the necessary resources to develop more robust methods for partial video manipulations.
## Summary
* **Problem.** Most detectors and datasets focus on *fully synthetic* videos. Subtle, localized edits (FakeParts) are under-explored yet highly deceptive.
* **Solution.** We define *FakeParts* and release **FakePartsBench**: 25K+ videos with **pixel-level** and **frame-level** annotations covering **full deepfakes** (T2V/I2V/TI2V) and **partial manipulations** (faceswap, inpainting, outpainting, style change, interpolation).
* **Finding.** Humans and SOTA detectors miss many FakeParts; detection accuracy drops by **30–40%** versus fully synthetic content.
* **Use.** Train and evaluate detectors that localize *where* and *when* manipulations happen.
## Dataset 💽
**FakePartsBench** provides:
* **25,000+** manipulated clips + **16,000** real clips
* High-res content (up to 1080p), durations typically **5–14 s**
* **Annotations:** frame masks (spatial), manipulated frames (temporal)
* **Categories:**
* **Full deepfakes:** T2V / I2V / TI2V (Sora, Veo2, Allegro AI)
* **Spatial FakeParts:** Faceswap (InsightFace), Inpainting (DiffuEraser, ProPainter), Outpainting (AKiRa)
* **Temporal FakeParts:** Interpolation (Framer)
* **Style FakeParts:** Style change (RAVE)
Each sample ships with metadata (prompt, source/cond frame when applicable, resolution, FPS) and, for FakeParts, per-frame masks or frame lists of manipulated regions/segments.
## Sample Usage 🚀
You can easily load the FakePartsBench dataset using the Hugging Face `datasets` library:
```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("hi-paris/FakeParts")
# Inspect the data
print(dataset)
```
## Citations ✍️
If you use **FakeParts** please cite:
```bibtex
@misc{brison2025fakeparts,
title={FakeParts: a New Family of AI-Generated DeepFakes},
author={Gaetan Brison and Soobash Daiboo and Samy Aimeur and Awais Hussain Sani and Xi Wang and Gianni Franchi and Vicky Kalogeiton},
year={2025},
eprint={2508.21052},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
## License & Responsible Use 🔨
* **Code:** see `LICENSE` (default: BSD-3-Clause unless noted otherwise in subfolders).
* **Dataset:** released for **research and defensive purposes only**.
* Do **not** attempt to identify private individuals.
* Do **not** use for generating disinformation or harassment.
* Faceswap content uses celebrity imagery to avoid sensitive personal data.
* Please comply with third-party model/data licenses cited in the paper and `baselines/`. |
# FakeParts: A New Family of AI-Generated DeepFakes
[](https://github.com/psf/black)
[](https://opensource.org/licenses/BSD-3-Clause)
[](https://huggingface.co/datasets/hi-paris/FakeParts)
[](https://arxiv.org/abs/2508.21052)
<p align="center">
<img src="https://huggingface.co/datasets/hi-paris/FakeParts/resolve/main/assets/final_teaser.png" width="95%" alt="FakePartsBench teaser">
</p>
<p align="center">
<img src="https://huggingface.co/datasets/hi-paris/FakeParts/resolve/main/assets/pipeline.jpg" width="95%" alt="Pipeline overview">
</p>
## Abstract
We introduce FakeParts, a new class of deepfakes characterized by subtle, localized manipulations to specific spatial regions or temporal segments of otherwise authentic videos. Unlike fully synthetic content, these partial manipulations, ranging from altered facial expressions to object substitutions and background modifications, blend seamlessly with real elements, making them particularly deceptive and difficult to detect. To address the critical gap in detection capabilities, we present FakePartsBench, the first large-scale benchmark dataset specifically designed to capture the full spectrum of partial deepfakes. Comprising over 25K videos with pixel-level and frame-level manipulation annotations, our dataset enables comprehensive evaluation of detection methods. Our user studies demonstrate that FakeParts reduces human detection accuracy by over 30% compared to traditional deepfakes, with similar performance degradation observed in state-of-the-art detection models. This work identifies an urgent vulnerability in current deepfake detection approaches and provides the necessary resources to develop more robust methods for partial video manipulations.
## Summary
* **Problem.** Most detectors and datasets focus on *fully synthetic* videos. Subtle, localized edits (FakeParts) are under-explored yet highly deceptive.
* **Solution.** We define *FakeParts* and release **FakePartsBench**: 25K+ videos with **pixel-level** and **frame-level** annotations covering **full deepfakes** (T2V/I2V/TI2V) and **partial manipulations** (faceswap, inpainting, outpainting, style change, interpolation).
* **Finding.** Humans and SOTA detectors miss many FakeParts; detection accuracy drops by **30–40%** versus fully synthetic content.
* **Use.** Train and evaluate detectors that localize *where* and *when* manipulations happen.
## Dataset 💽
**FakePartsBench** provides:
* **25,000+** manipulated clips + **16,000** real clips
* High-res content (up to 1080p), durations typically **5–14 s**
* **Annotations:** frame masks (spatial), manipulated frames (temporal)
* **Categories:**
* **Full deepfakes:** T2V / I2V / TI2V (Sora, Veo2, Allegro AI)
* **Spatial FakeParts:** Faceswap (InsightFace), Inpainting (DiffuEraser, ProPainter), Outpainting (AKiRa)
* **Temporal FakeParts:** Interpolation (Framer)
* **Style FakeParts:** Style change (RAVE)
Each sample ships with metadata (prompt, source/cond frame when applicable, resolution, FPS) and, for FakeParts, per-frame masks or frame lists of manipulated regions/segments.
## Sample Usage 🚀
You can easily load the FakePartsBench dataset using the Hugging Face `datasets` library:
```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("hi-paris/FakeParts")
# Inspect the data
print(dataset)
```
## Citations ✍️
If you use **FakeParts** please cite:
```bibtex
@misc{brison2025fakeparts,
title={FakeParts: a New Family of AI-Generated DeepFakes},
author={Gaetan Brison and Soobash Daiboo and Samy Aimeur and Awais Hussain Sani and Xi Wang and Gianni Franchi and Vicky Kalogeiton},
year={2025},
eprint={2508.21052},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
```
## License & Responsible Use 🔨
* **Code:** see `LICENSE` (default: BSD-3-Clause unless noted otherwise in subfolders).
* **Dataset:** released for **research and defensive purposes only**.
* Do **not** attempt to identify private individuals.
* Do **not** use for generating disinformation or harassment.
* Faceswap content uses celebrity imagery to avoid sensitive personal data.
* Please comply with third-party model/data licenses cited in the paper and `baselines/`. | 10,806 | 9 | [
"task_categories:video-classification",
"language:en",
"license:cc0-1.0",
"size_categories:n<1K",
"modality:video",
"library:datasets",
"library:mlcroissant",
"arxiv:2508.21052",
"region:us",
"deepfake-detection",
"video-manipulation",
"computer-vision",
"benchmark",
"deepfakes",
"multimodal",
"dataset-benchmarking",
"trustworthy-ai"
] | 2025-05-15T14:36:29+00:00 | 2025-11-11T10:46:29+00:00 | 0 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.