datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
robocasa_cosmos24_success300_env_depthNameonly_generated
Just Say the Name: Online Continual Learning with Category Names Only via Data Generation
We provide the dataset used for Name-only continual learning, generated using Stable Diffusion XL, DALL.E-2, CogView2, and DeepFloyd IF models.
Disclaimer
This dataset is created solely for academic purposes. We minimized human intervention to ensure a fair comparison with the baseline methods discussed in our paper. Despite our efforts, the extensive size of the dataset prevented us… See the full description on the dataset page: https://huggingface.co/datasets/seongwon980/Nameonly_generated.Robotwinvls-100k
VLS 100K
74,936 MS-COCO images paired with a long written description, a one-sentence
spoken summary of that description, the spoken audio, and that audio pre-encoded
with a neural codec. Images and audio are embedded in the parquet, so the viewer
renders them and one call opens the set:
from datasets import load_dataset
ds = load_dataset("seonglae/vls-100k", split="train")
ds[0]["image"] # PIL image
ds[0]["audio"] # decoded waveform
ds[0]["sst"] # the sentence that was… See the full description on the dataset page: https://huggingface.co/datasets/seonglae/vls-100k.vls-10k
VLS 10K
9,987 MS-COCO images paired with a long written description, a one-sentence
spoken summary of that description, the spoken audio, and that audio pre-encoded
by two neural codecs. Images and audio are embedded in the parquet, so the
viewer renders them and one call opens the set:
from datasets import load_dataset
ds = load_dataset("seonglae/vls-10k", split="train")
ds[0]["image"] # PIL image
ds[0]["audio"] # decoded waveform
ds[0]["sst"] # the sentence that was… See the full description on the dataset page: https://huggingface.co/datasets/seonglae/vls-10k.BigEarthNet-S1gingiris-seo-geo
SEO & GEO Growth Playbook
Rank on Google AND get cited by AI search engines — the dual-engine SEO + GEO strategy guide behind ~32K impressions/month, built from AFFiNE's 60K organic stars + 150 AI startup consultations
English | 中文
📦 Install
npx skills add Gingiris-1031/gingiris-seo-geo
Then ask your AI agent:
"Audit my SaaS site for SEO/GEO" or "Make my blog visible to ChatGPT and Perplexity"
Installs the complete SEO + GEO… See the full description on the dataset page: https://huggingface.co/datasets/Gingiris/gingiris-seo-geo.canvas
CANVAS
This repository contains the dataset accompanying the paper CANVAS: A Benchmark for Vision-Language Models on Tool-Based UI Design (AAAI 2026).
CANVAS designed to evaluate a VLM's capability to generate a UI design with tool invocations in two tasks: ADesign Replication and BDesign Modification.
A. Design Replication. Single Tasks involving restoring (implementing) a given UI image as is.
B. Design Modification. Multiple Tasks involving modifying or transforming existing… See the full description on the dataset page: https://huggingface.co/datasets/seooyxx/canvas.MAVIS
MAVIS: A Benchmark for Multimodal Source Attribution in Long-form Visual Question Answering
📖 Paper | 💻 Evaluation
Dataset Summary
MAVIS is a new dataset for open-domain, long-form visual question answering, characterized by three key features: (1) the questions incorporate input images, requiring visual understanding to correctly interpret the user’s intent; (2) the desired answers are long-form, necessitating the retrieval and synthesis of diverse information rather… See the full description on the dataset page: https://huggingface.co/datasets/seokwon99/MAVIS.vlabench_primitive_ft_lerobot_224
vlabench_primitive_ft_lerobot_224
VLABench/vlabench_primitive_ft_lerobot (the official pi0.5 VLABench fine-tuning set, LeRobot format) with every
camera image re-encoded from 480x480 to 224x224 (bit-identical to openpi's resize_with_pad applied at load time),
so that training does not pay for the 480px decode + resize and the LeRobot Arrow cache stays small.
Same episodes, same schema and metadata; only the image resolution differs. Made with… See the full description on the dataset page: https://huggingface.co/datasets/SeonghoonYu/vlabench_primitive_ft_lerobot_224.MAVIS_documentschemistry-jsdhist3-summary
Chemistry SDPO jsdhist3 Training History
This repository contains summarized training-history artifacts for four completed Chemistry SDPO Qwen3-4B runs from /workspace/SDPO-new-clean.
Images
Files
data/jsd_history.csv: per-step JSD scalar history and validation metrics parsed from the training log.
data/run_summary.csv: one-row-per-run summary with final/best validation reward and histogram event counts.… See the full description on the dataset page: https://huggingface.co/datasets/SeongryongJung/chemistry-jsdhist3-summary.save-coco-text-tmpRobotwin-12tasksDrakepickplace_pandaThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": null,
"total_episodes": 12,
"total_frames": 3605,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 10,
"splits": {
"train": "0:12"
},
"data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/seojing/pickplace_panda.funsd-bank-paragraph-test3japdongsani
train_2024-06-03-09-50-18
This model is a fine-tuned version of yanolja/EEVE-Korean-Instruct-10.8B-v1.0 on the assay dataset.
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
learning_rate: 0.0003
train_batch_size: 2… See the full description on the dataset page: https://huggingface.co/datasets/SEOje/japdongsani.ecb-datasets
ECB Datasets: Cultural Bias Evaluation in Generative Image Models
Overview
This dataset contains human evaluation data for cultural bias analysis in image generation models, supporting the research paper "Exposing Blindspots: Cultural Bias Evaluation in Generative Image Models". ECB stands for "Evaluation Cultural Bias". The dataset includes prompts, generated images, and evaluation metrics across different countries and cultural contexts.
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/seochan99/ecb-datasets.face-test-02-mmppMy-Seoul-Imagesseoul_subwayKorean Subway images, filmed from Seoul Station to Seoul National University.