datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CASTLE2024
What is CASTLE?
The CASTLE dataset is a large-scale, multimodal dataset designed for advancing research in lifelogging, human activity recognition, and multimodal retrieval. It provides a rich collection of time-aligned sensor and video data for analysis and benchmarking. See the Paper (or its arXiv pre-print) for more details.
You can check our website for more details.
Characteristics
Captured over four days in a controlled environment
10 participants engaged… See the full description on the dataset page: https://huggingface.co/datasets/CASTLE-Dataset/CASTLE2024.8KDehazeThe Full version of the datasets 8KDehaze
MINI version : https://huggingface.co/datasets/fengyanzi/8KDehaze_mini
The first dataset of extremely large image dehazing
from CVPR2025 paper "Tokenize Image Patches: Global Context Fusion for Effective Haze Removal in Large Images"
Paper link: https://openaccess.thecvf.com/content/CVPR2025/html/Chen_Tokenize_Image_Patches_Global_Context_Fusion_for_Effective_Haze_Removal_CVPR_2025_paper.html
DehazeXL code: https://github.com/CastleChen339/DehazeXL… See the full description on the dataset page: https://huggingface.co/datasets/CastleChen339/8KDehaze.AstroDimps4mas-castle-cold
PS4MAS CASTLE cold
2026-09-15: Removed seven audited legacy handwritten-dataset experiment entries at the user's explicit request. Their IDs and query texts did not match the sealed official CASTLE-200. The verified official input under datasets/official_castle200/ is retained. Only newly validated results belong in the active catalog. Historical score-only results, when present, do not prove full-hop or oracle-profile correctness.
Current CASTLE / cold experiment… See the full description on the dataset page: https://huggingface.co/datasets/yinita/ps4mas-castle-cold.ps4mas-castle-oracle
PS4MAS CASTLE oracle
Current CASTLE / oracle experiment catalog
This generated section is authoritative. Older tables above are historical.
Each experiments/<name>/ contains unified episodes.parquet, meta.json, and an unmodified summary.json when available. Raw full-hop traces.jsonl and run_config.json are separate files; scores are never injected into raw traces.
Partial snapshots have immutable content-derived names. Existing experiments are skipped unless… See the full description on the dataset page: https://huggingface.co/datasets/yinita/ps4mas-castle-oracle.ps4mas-castle-scale
PS4MAS CASTLE scale
Current CASTLE / scale experiment catalog
This generated section is authoritative. Older tables above are historical.
Each experiments/<name>/ contains unified episodes.parquet, meta.json, and an unmodified summary.json when available. Raw full-hop traces.jsonl and run_config.json are separate files; scores are never injected into raw traces.
Partial snapshots have immutable content-derived names. Existing experiments are skipped unless… See the full description on the dataset page: https://huggingface.co/datasets/yinita/ps4mas-castle-scale.CASTLE2024_Audio_onlyCASTLE2024
What is CASTLE?
The CASTLE dataset is a large-scale, multimodal dataset designed for advancing research in lifelogging, human activity recognition, and multimodal retrieval. It provides a rich collection of time-aligned sensor and video data for analysis and benchmarking. See the Paper (or its arXiv pre-print) for more details.
You can check our website for more details.
Characteristics
Captured over four days in a controlled environment
10 participantsengaged in… See the full description on the dataset page: https://huggingface.co/datasets/E1i0/CASTLE2024.regularization-castle
Castle Regularization Images
A collection of regularization & class instance datasets of castles for the Stable Diffusion 1.5 to use for DreamBooth prior preservation loss training.
ps4mas-castle-final-test-rollouts
Retired nonofficial CASTLE rollout archive
2026-09-15: All 15 rollout files were removed following an explicit user-requested content audit. None of their 12,000 query rows matched the official CASTLE English source pool. They must not be used as official CASTLE-200 evidence.
Validated results: https://huggingface.co/datasets/yinita/ps4mas-castle-cold and the domain/mode repositories ps4mas-castle-oracle and ps4mas-castle-scale.
castles-of-the-world
Castles of the World — 8,131 castles, fortresses & palaces
A curated, hand-checked dataset of the world's 8,131 great castles, fortresses, palaces and ruins across 149 countries — every landmark verified to have exact coordinates, a real Wikimedia Commons photo and an English Wikipedia article.
This is the exact dataset behind Castlemap, a free interactive night-map of all 8,131 landmarks.
🗺️ Interactive map: https://thecastlemap.com
📄 Dataset page & downloads:… See the full description on the dataset page: https://huggingface.co/datasets/Flightmussy/castles-of-the-world.Level-5-The-Enchanted-Castleberserk-castles-loraps4mas_castle_train_v2
ps4mas_castle_train_v2
Disjoint training split for the PS4MAS V5 PPO experiments.
Source: ECNU-RAIL/CASTLE-EMNLP2026/data/CASTLE_en.json (39425 rows)
Filter: 200 eval queries, dedup
Size: 36959 rows
Eval excluded: any query appearing in
yinita/ps4mas_castle_scenery_200
Overlap with eval: 0 (verified)
Domain: CASTLE
Schema (JSONL)
case_id, scenario_id, question, profile (JSON str), domain, scenario
ps4mas_castle_scenery_200
ps4mas_castle_scenery_200
200 stratified CASTLE cases for the PS4MAS 0919 plan (official_stratified_200).
Schema (jsonl)
Each line:
{
"case_id": "cso06a6a5813a",
"scenario_id": "cso06a6a5813a",
"question": "How can someone manage stress when everything feels overwhelming and out of control?",
"profile": "{"Age": "17", "Gender": "Male", "Learning Stage": "High School", ...}",
"domain": "CASTLE",
"scenario": "Academic Pressure… See the full description on the dataset page: https://huggingface.co/datasets/yinita/ps4mas_castle_scenery_200.il_gym_day1This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 10,
"features": {
"action": {
"dtype": "float32",
"shape": [
4
],
"names": [
"delta_x",
"delta_y",
"delta_z",
"gripper"
]
},
"next.reward": {
"dtype": "float32"… See the full description on the dataset page: https://huggingface.co/datasets/castleuk2/il_gym_day1.mydatasetimnet1k_castleCASTLEVANIA_SOTNCASTLEcastlelicense: apache-2.0
task_categories:
conversational
text-generation
tags:
roleplay
fantasy
castle
testroaring-castleCastlehdr_data