datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
CT_DeepLesion-MedSAM2
CT_DeepLesion-MedSAM2 Dataset
Authors
Jun Ma* 1,2,
Zongxin Yang* 3,
Sumin Kim2,4,5,
Bihui Chen2,4,5,
Mohammed Baharoon2,3,5,
Adibvafa Fallahpour2,4,5,
Reza Asakereh4,7,
Hongwei Lyu4,
Bo Wang† 1,2,4,5,6
* Equal contribution † Corresponding author
1AI Collaborative Centre, University Health Network, Toronto, Canada
2Vector… See the full description on the dataset page: https://huggingface.co/datasets/wanglab/CT_DeepLesion-MedSAM2.multimodal-ct-radiology-reports
Perle AI Multi-phase CECT and CT with Radiology Reports
Summary
A de-identified CT dataset from Perle AI, paired with the original radiology reports. It supports work on multi-modal medical imaging: phase or pathology classification, report generation from images, and visual question answering.
The release has three configurations:
Config
Modality
Subjects
Pairing
cect_3phase
3-phase contrast-enhanced abdominal CT (DICOM)
5
per-subject text report +… See the full description on the dataset page: https://huggingface.co/datasets/Perle-ai/multimodal-ct-radiology-reports.dev_plantcad2_ft_long_ctxctc-suite-eval
CTC suite eval ladders
The 22-task corpus-tracking-capacity suite: per-task context ladders from 2k to 1M tokens,
consumed by the ctc_suite task family on the prasann/ctc-suite branch of allenai/olmo-eval
(ctc_nq:r64k, suites ctc:figure / ctc:xlong / ctc:r128k / ...). One config per task, one
split per rung; each row is one unified-format example (documents + queries + answers + gold).
Public release note (2026-08-14). Gold answers are included — training on this data… See the full description on the dataset page: https://huggingface.co/datasets/PrasannSinghal/ctc-suite-eval.staining-robustness-evaluation
A Protocol for Evaluating Robustness to H&E Staining Variation in Computational Pathology Models
This repository provides the stain references, pretrained models, and experimental results required to:
Define custom staining references using our PLISM reference library
Reproduce our published controlled staining robustness experiments
👉 Code repository: https://github.com/lely475/staining-robustness-evaluation/tree/main
👉 Associated publication: Paper
Overview: How… See the full description on the dataset page: https://huggingface.co/datasets/CTPLab-DBE-UniBas/staining-robustness-evaluation.train_ctf_eeftrain_ctftest_ctfctu_datasetsCTODataset for predicting clinical trial outcomes in drug development. This dataset is part of the work presented in "Automatically Labeling Clinical Trial Outcomes: A Large-Scale Benchmark for Drug Development".
Website: https://chufangao.github.io/CTOD/
Paper: https://arxiv.org/abs/2406.10292
Code: https://github.com/chufangao/ctod
Descriptions:
human_labels contains the manually annotated subset. We follow the same rule-based termination of incomplete status and p-value < 0.05 as in the… See the full description on the dataset page: https://huggingface.co/datasets/chufangao/CTO.aquatype-canonical-ctx-20260706trl-ctbench
TRL-CTbench
Paper: arXiv:2606.09323 — TRL-Bench: Standardizing Cross-Paradigm Representation-Level Evaluation of Tabular Encoders · Code: LOGO-CUHKSZ/TRL-Bench
Column- and table-level evaluation suite of TRL-Bench. All 27 configs are live (covering every CTbench source in the paper appendix, plus separate *_tables configs for benchmarks whose label volume + table corpus would otherwise exceed parquet's per-shard limits).
Configurations
Retrieval-style… See the full description on the dataset page: https://huggingface.co/datasets/logo-lab/trl-ctbench.ctr-scan-object-uniform50-20260917
IdleMask review — passed (2026-09-19)
Reviewed by the dataset owner: observation.arm_active_mask is correct and this revision is a formally usable CTR dataset. This section supersedes previous active/idle-mask descriptions below.
For each of the 50 episodes and each physical arm, only the initial contiguous scheduling delay may have mask 0. From first duty through the final frame the mask is always 1. Scan synchronization waits, cooperative holds, the shared scan tail and… See the full description on the dataset page: https://huggingface.co/datasets/Shiki42/ctr-scan-object-uniform50-20260917.conflux-chest-ct
CONFLUX Chest-CT
200,000 synthetic 3D chest CT volumes with structured abnormality and demographic labels, generated by CONFLUX.
Released with the paper CONFLUX: A Latent Diffusion Model for 3D Chest-CT Synthesis with RL Post-Training.
Paper (arXiv) •
Model •
Code — coming soon
About
CONFLUX is a conditional 3D latent generative model for chest CT: a VAE tokenizer
compresses each volume into a compact 16-channel latent, a… See the full description on the dataset page: https://huggingface.co/datasets/gevaertlab/conflux-chest-ct.HistoPlexer-Ultivue
HistoPlexer-Ultivue Dataset
Dataset Summary
The HistoPlexer-Ultivue dataset provides a collection of multimodal histological images for cancer research. It includes whole-slide images (WSIs) of hematoxylin and eosin (H&E) stained tissue, multiplexed immunofluorescence images from Ultivue panels (immuno8 and mdsc), alignment matrices, exclusion masks, and nuclear segmentation outputs. It is a multiplexed dataset for 10 cancer samples from the Tumor Profiler Study. The… See the full description on the dataset page: https://huggingface.co/datasets/CTPLab-DBE-UniBas/HistoPlexer-Ultivue.record-screw-urThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "cta_ur_follower",
"total_episodes": 10,
"total_frames": 7519,
"total_tasks": 1,
"total_videos": 10,
"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":… See the full description on the dataset page: https://huggingface.co/datasets/alex-cta/record-screw-ur.alfa_ct
Dataset Summary
Alfa Card Transactions is a unique high-quality dataset collected from real data sources of Alfa Bank's clients' transactions for the task of the default prediction. It consists of histories of transactions, IDs of credit products and flags of corresponfing default.
Supported Tasks and Leaderboards
The dataset is supposed to be used for training models for the classical bank task of predicting the default of the applicant.
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/mllab/alfa_ct.ISCCP-H-CT
A Global Multi-Decadal Convection Tracking Database from ISCCP-H (1983-2017)
ISCCP-H CT: a global, 34-year (July 1983 - June 2017) convection-tracking database derived from International Satellite Cloud Climatology Project H-series (ISCCP-H) infrared observations, produced with the Tracking and Object-Based Analysis of Clouds (tobac) framework.
This dataset accompanies the manuscript:
Luo, Z. J., Wang, L.-P., Selevich, Y., Takahashi, H., Wu, C.-L., Jhang, H., Lin, S.-C., van… See the full description on the dataset page: https://huggingface.co/datasets/NTU-CompHydroMet-Lab/ISCCP-H-CT.voxknesset-whisper-large-v3-ct2-inference
VoxKnesset × ivrit-ai/whisper-large-v3-ct2 — inference results
Transcriptions of ivrit-ai/VoxKnesset
produced by ivrit-ai/whisper-large-v3-ct2
(faster-whisper, float16, batched, language=he, VAD on), on 8× NVIDIA A40.
Columns: speaker metadata (from VoxKnesset), duration_s, reference_text
(official Knesset protocol), model_transcription, segments_json
(start/end/text), infer_time_s, split.
Current contents: 10-example pilot from the test split (data/results_10.parquet).
Full-run… See the full description on the dataset page: https://huggingface.co/datasets/Dolevabudi/voxknesset-whisper-large-v3-ct2-inference.ctr-pick-dual-bottles-uniform50-20260917
IdleMask review — passed (2026-09-19)
Reviewed by the dataset owner: observation.arm_active_mask is correct under the approved prefix-only rule. This section supersedes previous active/idle-mask descriptions below.
For each of the 50 episodes and each physical arm, only the initial contiguous wait before its first active sample may have mask 0. From the first active sample through the final frame, the mask is always 1, including intermediate waits, terminal holds, and repeated… See the full description on the dataset page: https://huggingface.co/datasets/Shiki42/ctr-pick-dual-bottles-uniform50-20260917.ctrldataset2026
MonitoringBench
A benchmark for evaluating LLM-based monitors of agentic AI systems. Contains
2,644 successful attack trajectories in which an AI agent accomplished one of four harmful side tasks
(sudo escalation, firewall disabling, malware download, password leaking) in
a sandboxed Linux environment under the control_arena framework.
Each trajectory is scored by a panel of 13+ LLM monitors (GPT-3.5 / 4.x / 5.x,
Claude Opus 4.x and Sonnet 4.x, o3, o4-mini, gpt-5-nano), with both… See the full description on the dataset page: https://huggingface.co/datasets/neur26anonsub/ctrldataset2026.ctr-scan-object-uniform100-20260921
scan_object Uniform100
Open in LeRobot Dataset Visualizer
100 episodes from the same 50 source scene seeds, 25 FPS, LeRobot v3, Aloha
AgileX. Parent: Shiki42/ctr-scan-object-uniform-20260916 at 766e18802a11bdde94f6dab6892c13d82e6a5bd3
(repaired gripper target labels). This is a byte-exact republication of the
pinned parent for the non-mainline 100-sample comparison arm; the 50-episode
arm remains ctr-scan-object-uniform50-20260917.
Every source seed appears exactly twice, as… See the full description on the dataset page: https://huggingface.co/datasets/Shiki42/ctr-scan-object-uniform100-20260921.sort_batteryThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.0",
"robot_type": "koch",
"total_episodes": 100,
"total_frames": 39728,
"total_tasks":1,
"total_videos": 300,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:100"
},
"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/ctbfl/sort_battery.CTR250fixed
CTR250fixed
Merged LeRobot v3 dataset for the fixed-start CTR ring-placement task.
Task: Pick up a yellow ring and place it onto a red post.
Episodes: 355
Frames: 163,768
FPS: 30
Cameras: Cam2 and Cam4 at 960 x 960; Cam5 wrist camera at 400 x 400
Pilot dataset CTR250fixed-test is excluded.
The 12 timestamped source datasets are retained separately for provenance. Merge provenance and validation are stored in meta/ctr_merge_manifest.json and meta/ctr_merge_validation.json.
ctddINSTRUCT_JEV
INSTRUCT_JEV
INSTRUCT_JEV is an instruction corpus built from the TypeSafe AI documentation
for Jev, the first System One model. It is structured around the three TypeSafe
question primitives - Choice, Noul and Score - and mirrors the raw corpus
captured in deckerGUI-jev_corpus_RAW.
Credits
INSTRUCT_JEV is a DeckerGUI project and exists because of the work below.
Who
Contribution
Link
TypeSafe AI
Jev - the first System One model - and the Choice / Noul… See the full description on the dataset page: https://huggingface.co/datasets/ctaxnagomi/INSTRUCT_JEV.PHI-CTRL-F16-Fault-Recovery-Telemetry
PHI-CTRL F-16 Actuator Fault Recovery Dataset
High-Fidelity JSBSim 6-DOF Telemetry for Physics-Hybrid Self-Healing Flight Control
Official verification artifacts of the PHI-CTRL (Physics-Hybrid Integrity Control) architecture — a digital-twin-driven, self-healing flight control framework that actively compensates actuator degradation in real time.
Author: Mohammed Bello Sani (SM-Bello)
Affiliation: Air Force Institute of Technology (AFIT), Kaduna · Penelope Inc. / PHI Lab… See the full description on the dataset page: https://huggingface.co/datasets/SM-Bello/PHI-CTRL-F16-Fault-Recovery-Telemetry.CTSpinoPelvic1K
CTSpinoPelvic1K
A fused spine + pelvis 3D CT segmentation dataset built by patient-level
crosswalk between three public sources:
TCIA CT COLONOGRAPHY — DICOM CT volumes (prone + supine per patient)
CTSpine1K (COLONOG subset) — VerSe-convention vertebral label masks
CTPelvic1K dataset2 — sacrum + bilateral hip label masks
Annotations are placed onto the TCIA CT volume with the highest bone
coverage (HU > 200), separately per anatomy. For ~650 patients both
annotations land on the… See the full description on the dataset page: https://huggingface.co/datasets/anonymous-neurips-ED/CTSpinoPelvic1K.piperx-sortletter-20260921-140ep-ctr-2x
piperx-sortletter-20260921-140ep-ctr-2x
280 episodes, 216821 frames, 30 FPS, three 640×480 RGB views. LeRobot v3.0, 14-D action and state.
Selection in output order: ctr-2x episodes 1–280 (1-based). CTR retains two relative-timing samples per original episode (280 total); the 140ep name refers to the 140 source demonstrations.
Source: Takizz/piperx-sortletter-0911-0917-140ep-raw, pinned revision f3d743be7a21e8f9d0c5dd23c910ca50c0134ad1. provenance.json records every… See the full description on the dataset page: https://huggingface.co/datasets/Shiki42/piperx-sortletter-20260921-140ep-ctr-2x.trl-ctbench-sample
TRL-CTbench (sample)
This is a small sample of logo-lab/trl-ctbench,
intended for the NeurIPS 2026 E&D track's "Dataset Large URL" requirement —
reviewers can inspect data quality across all 27 configs without
downloading the full ~31 GB.
Total sample size: a few hundred MB. Schema is identical to the full
dataset; only row count differs.
Sampling rule
Deterministic and easy to verify:
For each (config, split) of the full logo-lab/trl-ctbench, take the
first 100 rows… See the full description on the dataset page: https://huggingface.co/datasets/logo-lab/trl-ctbench-sample.