datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
xvla-soft-foldThis dataset was created using LeRobot.
Dataset Description
Repository: X-VLA
License: Apache 2.0
Paper: Zheng et al., 2025, “X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model” (arXiv:2510.10274)
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "franka",
"total_episodes": 1542,
"total_frames": 2852512,
"total_tasks": 1,
"chunks_size": 1000… See the full description on the dataset page: https://huggingface.co/datasets/lerobot/xvla-soft-fold.training-embeddingssoft-prompt-experiments-archive-20260918
Soft prompt 实验归档
用于查阅和恢复的历史研究记录，涵盖数学与代码任务。共 45 个运行目录，包含教师生成数据、评测输出、原始配置和已有 prompt 检查点。部分目录仅有评测、复核或失败记录，不能将目录数量理解为成功实验数量。
快速查阅
实验总览：模型系列、任务、规模与记录状态。
CSV 索引 / JSON 索引：便于筛选和定位。
archives/：按实验分别压缩的原始文件。
manifests/：各文件 SHA-256 与归档路径。
系列包括 AReaL Boba2、GPT-OSS/Swallow、MiMo、X-Coder/Qwen3、Nemotron、Klear、Mellum2、OLMo3、Polaris 和 Poro2。页面不展开具体方法或实现细节；原始配置仍保留在归档内供恢复。
状态与注意事项
上传完成以 ARCHIVE_COMPLETE.json 为准；文件不存在时表示仍在上传。 每个归档都经过完整下载的 SHA-256 校验。… See the full description on the dataset page: https://huggingface.co/datasets/namezz/soft-prompt-experiments-archive-20260918.soft_bag_zipping
soft_bag_zipping
Deformable-object manipulation: unzip a soft bag, insert pens one at a time, then zip the bag closed.
Real-robot bimanual manipulation data collected on a YAM arm pair, released as part
of the Flex-π project. Stored in LeRobot v2.1 format with synchronized RGB and
metric depth from three cameras.
At a glance
Episodes
534
Frames
1,047,427
Duration
~9.7 h @ 30 fps
Tasks
1
Robot
yam (bimanual)
Cameras
cam_high, cam_left_wrist… See the full description on the dataset page: https://huggingface.co/datasets/flex-pi/soft_bag_zipping.stockfish-19-soft-targets
avewright/stockfish-19-soft-targets
Official Stockfish 19 MultiPV soft targets, mined from the Lichess ECO
opening set. In-progress snapshot toward 1M unique positions.
600,000 rows in this upload. Source id 4. Vocab compact (1968).
How positions are chosen
Games start from the Lichess Chess Openings dataset
(lichess-org/chess-openings): 3,810 named
leaves (HF card still lists 3,704) plus
book prefixes, 7,852 unique starts.
ECO volumes: A 817 / B 772 /
C 1,250 / D… See the full description on the dataset page: https://huggingface.co/datasets/avewright/stockfish-19-soft-targets.chess-soft-multipv-lichess
avewright/chess-soft-multipv-lichess
Soft MultiPV policy targets for chess transformers (compact move vocab).
Built from Lichess cloud evaluations + local harvests. Each row is a position with
an 8-wide soft move distribution (soft_indices / soft_probs) plus hard best move.
Fields
board_array (64): piece encoding
turn, castling, ep_square
move_idx, cp, mate
soft_indices[8], soft_probs[8]
label_depth, phase, source, cache_name
us-k12-schools-directory
US K-12 Schools Directory
A directory of 124,613 US K-12 schools covering all 50 states, DC, and US territories,
compiled from federal and state government sources. Each record carries directory
information (address, phone, website), enrollment and demographics, and, where a source
supplied it, a principal name and email.
This is a compilation of public government data. It is not a survey, and no field was
independently verified against the school itself.
Loading… See the full description on the dataset page: https://huggingface.co/datasets/robworks-software/us-k12-schools-directory.arxiv-software-repo-links
arXiv Software Repository Links
A dataset mapping arXiv papers (via DOI) to software repositories they reference or which are implementations of the work. Additionaly includes co-citation analysis and community clustering.
Implementation detection is conducted using the evamxb/dev-author-em-clf model from sci-soft-models
Quick Start
from datasets import load_dataset
# Load DOI-to-repo links
links = load_dataset("cometadata/arxiv-software-repo-links", "links")
#… See the full description on the dataset page: https://huggingface.co/datasets/cometadata/arxiv-software-repo-links.soft-toy-wbcd-khlpchess-soft-100m-swa-mistakes
avewright/chess-soft-100m-swa-mistakes
Positions where overnight eval_swa.pt
(outputs/sf19_ft/overnight_20260908, SF 8000-node screen score 0.719)
disagrees with a teacher best move.
In-PV severity uses teacher MultiPV STM cps. Off-PV rows are not given
an invented drop; they are queued for Stockfish 19 analysis (needs_sf=1).
Holdout + flip hashes from the overnight union (20,961) are excluded.
1,264,747 rows in 64 shards.
Tags
tag
rows
blunder
59,889… See the full description on the dataset page: https://huggingface.co/datasets/avewright/chess-soft-100m-swa-mistakes.g2-software-reviews
G2 Software Reviews
111,441 B2B software reviews from G2, covering the 79 most-reviewed products, spanning 2012 to 2026.
The largest public G2 review corpus by a wide margin. Before this, the biggest available was a sample of under 1,000 rows.
What is in here that is not in other review datasets
A structured pros-and-cons split on 35,137 reviews. G2 asks "what do you like best" and "what do you dislike" as separate prompts, so those are separate columns rather… See the full description on the dataset page: https://huggingface.co/datasets/FreshCrawl/g2-software-reviews.chess-soft-100m-disagreements
avewright/chess-soft-100m-disagreements
Positions where the greedy policy of
avewright/chess-transformer-100m-squares64
disagrees with a strong teacher best move (move_idx).
Current upload: 1,782,505 rows in 90 shards.
Mix
split
rows
shards
labels
data/shard_*.parquet
1,768,622
83
teacher MultiPV from avewright/chess-soft-multipv-lichess
data/sf19/*.parquet
13,883
7
Stockfish 19 max-Elo MultiPV from new games
Stream rows are an argmax filter of… See the full description on the dataset page: https://huggingface.co/datasets/avewright/chess-soft-100m-disagreements.softverse
Softverse: software referenced in social science replication code
Which R, Python and Stata packages the code in published replication deposits
loads, at economics and political science journals whose data-and-code policy
an editor verifies. One row per deposit, file, file text, package reference,
stated version and environment signal, plus per-language tallies.
A reference is not a run. A row says deposited code names a package: in
library(), an import, a Stata command. It does… See the full description on the dataset page: https://huggingface.co/datasets/gojiberries/softverse.bb_softsim_10_manyberriesreachThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so100_follower",
"total_episodes": 34,
"total_frames": 5871,
"total_tasks": 1,
"total_videos": 136,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:34"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/andlyu/bb_softsim_10_manyberriesreach.so100_softThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so100",
"total_episodes": 50,
"total_frames": 21361,
"total_tasks": 1,
"total_videos": 100,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:50"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/gxy1111/so100_soft.SweSetupBench-liteThis repository contains the data presented in SWE-Factory: Your Automated Factory for Issue Resolution Training Data and Evaluation Benchmarks.
capterra-b2b-software-reviews
Capterra B2B Software Reviews
56,606 B2B software reviews from Capterra, covering 66 products across 11 software categories.
Most public review datasets are star rating + review text. This one carries five separate rating dimensions, pros and cons as distinct pre-split fields, reviewer firmographics, and, unusually, an incentive disclosure flag recording whether the reviewer was given a gift card, referred by the vendor, or wrote the review unprompted.
Why this is… See the full description on the dataset page: https://huggingface.co/datasets/FreshCrawl/capterra-b2b-software-reviews.bb_softsim_1This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so100_follower",
"total_episodes": 50,
"total_frames": 22142,
"total_tasks": 1,
"total_videos": 200,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:50"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/andlyu/bb_softsim_1.grabette-tactile-softThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "grabette",
"total_episodes": 15,
"total_frames": 5719,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 50,
"splits": {
"train": "0:15"
},
"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/CarolinePascal/grabette-tactile-soft.jeopardy-clues
Jeopardy! Clues
568,068 Jeopardy! clues with their answers, categories, dollar values, air dates, and
round information, compiled from publicly archived, community-maintained transcriptions
of aired episodes.
Loading
from datasets import load_dataset
ds = load_dataset("robworks-software/jeopardy-clues")
science = ds["train"].filter(lambda x: x["category"] == "SCIENCE")
Splits
Split
Rows
train
482,857
validation
42,605
test
42,606… See the full description on the dataset page: https://huggingface.co/datasets/robworks-software/jeopardy-clues.newah-s2-soft-v5This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"observation.state": {
"dtype": "float32",
"shape": [
5
],
"names": [
"shoulder_pan",
"shoulder_lift",
"elbow_flex",
"wrist_flex",
"wrist_roll"
]
}… See the full description on the dataset page: https://huggingface.co/datasets/kurah/newah-s2-soft-v5.so100_test_100_2025-08-10T18-37-57This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so_arm100",
"total_episodes": 4,
"total_frames": 1788,
"total_tasks": 1,
"total_videos": 8,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:4"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/software-pathon-ai/so100_test_100_2025-08-10T18-37-57.biomed-fr-v3-enriched-softmin-standard
biomed-fr-v3-enriched-softmin-standard
This dataset is a quality-upsampled version of rntc/biomed-fr-v3-enriched using soft-min bottleneck sampling.
Preprocessing Method
Soft-min calculation:
Formula: s = (mean(q_k^p))^(1/p) where q_k are the 4 quality scores
Parameter p = -2.0
Weight computation:
Ratio preference (5 vs 1): R = 10
Gamma exponent: γ = 1.43 (computed as log(R)/log(5))
Weight formula: w = s^γ
Floor: w = max(w, median(w) × 0.05)
Resampling:
Target size:… See the full description on the dataset page: https://huggingface.co/datasets/rntc/biomed-fr-v3-enriched-softmin-standard.arxiv-software-engineering-datasetcanadian-software-rfp-readiness-dataset
Canadian Software Procurement Notice Dataset
This tabular dataset contains 136 Canadian public software procurement notices selected through a documented rule-based classifier and complete review of all lower-confidence candidates.
Dataset Summary
Essential Designs parsed 16,203 official CanadaBuys tender-notice rows across three fiscal years and deduplicated amendments by notice reference. The classifier produced 167 candidates. Ninety-nine qualified… See the full description on the dataset page: https://huggingface.co/datasets/essentialdesigns/canadian-software-rfp-readiness-dataset.metra-alpha-ma-soft-a005-n3000This dataset was created using LeRobot.
Dataset Description
Successful BananaInBowl demonstrations collected from an RFCL-trained SAC policy in
Isaac Lab (RoboLab), for behaviour-cloning research on strategy diversity.
Lane: metra-alpha — a sweep of the METRA intrinsic-reward scale alpha, with every
other axis held fixed (banana task, 50 demos, sf=0.5, z_dim=3, phi_space=full, z_unit=true).
Each dataset is one (difficulty level, alpha) cell. Difficulty here: soft.
Collection:… See the full description on the dataset page: https://huggingface.co/datasets/DAVIAN-Robotics/metra-alpha-ma-soft-a005-n3000.RoboLab-FoodPacking1Cans-soft-capx-oracle-260812This dataset was created using LeRobot.
Dataset Description
Simulated manipulation demonstrations generated in RoboLab (Isaac Lab) by a deterministic waypoint oracle with privileged state access (no cameras used for control). Task: FoodPacking1Cans, difficulty level soft.
What this is
Part of a 3 task x 3 difficulty grid of oracle datasets. The oracle replays a fixed waypoint recipe solved with rest-cost IK (pyroki) and is rate-limited to match the arm-speed cap… See the full description on the dataset page: https://huggingface.co/datasets/DAVIAN-Robotics/RoboLab-FoodPacking1Cans-soft-capx-oracle-260812.open-thoughts-4-30k-math-qwen3-4b-annotated-32768-tokens-n8-rejection-sampling-soft-match
N8 Rejection Sampling (Soft Match)
Overview
This dataset was created via rejection sampling from the Qwen3-4B response dataset using Qwen3-32B answers as ground truth.
Source dataset (Qwen3-4B, 8 responses per prompt): marin-community/open-thoughts-4-30k-math-qwen3-4b-annotated-32768-tokens-n8-reformatted
Verifier dataset (Qwen3-32B, 1 response per prompt): marin-community/open-thoughts-4-30k-math-qwen3-32b-annotated-32768-tokens
Creator: The Marin Project
How… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/open-thoughts-4-30k-math-qwen3-4b-annotated-32768-tokens-n8-rejection-sampling-soft-match.soft-toy-wbcdchonk_cowThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "so_arm100",
"total_episodes": 12,
"total_frames": 6503,
"total_tasks": 1,
"total_videos": 24,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:12"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/software-pathon-ai/chonk_cow.