datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
precision-seed-germination-time-lapsesglang-nightly-precision-baselinesAgriStress-500Precision AI · AgriStress-500
A curated stress test for agricultural AI: 500 real drone images from working fields across crops, geographies, sensors, altitudes, and lighting conditions.
Each image captures field conditions that challenge real-world deployment sun glare, tilted leaves, row occlusion, mixed species, rare growth stages, and edge cases underrepresented in public datasets.
Use AgriStress-500 to test embeddings, segmentation models, classifiers, and VLMs before… See the full description on the dataset page: https://huggingface.co/datasets/precisionaiinc/AgriStress-500.pile_pythonmega-precision-image-video
Mega Precision Image & Video Dataset
Description
A high-quality dataset for training image and video generation models with ultra-detailed prompts, technical parameters, and quality metrics.
Content
10,000,000 examples
4 categories: image prompts, video prompts, technical parameters, quality metrics
English language
Categories
Category
Description
%
Image Prompts
Ultra-detailed prompts for image generation
30%
Video… See the full description on the dataset page: https://huggingface.co/datasets/Lelonthecodeur/mega-precision-image-video.Log_precisionprecision-evidence-bench
Precision Evidence Bench
Precision Evidence Bench is a Precision Medicine Benchmark from
Atropos Health, the world's largest creator of
real-world evidence (RWE) for clinical decision support. It evaluates how well
large language models (LLMs) answer clinical questions that are grounded in
patient context and inclusive of patient history: not "which treatment is
better in general", but "which treatment is better for this patient", with a
specific comorbidity, age, prior therapy… See the full description on the dataset page: https://huggingface.co/datasets/atroposhealth/precision-evidence-bench.daily-paper-2026-07-25-spec-decode-quant-precision-mismatch
Draft-Target Precision Mismatch in Speculative Decoding: Mapping the Acceptance-Throughput-Cost Frontier for Quantized MoE LLM Serving
TL;DR — When a bf16 draft model proposes tokens for an NVFP4-quantized MoE target, does precision mismatch erode the acceptance rate that speculative decoding depends on? Analytical model predicts it does — and that the erosion is discrete (router-boundary flips), not continuous like a dense target.
ThakiCloud AI Research · 2026-07-25 · 📝 Tech… See the full description on the dataset page: https://huggingface.co/datasets/thaki-AI/daily-paper-2026-07-25-spec-decode-quant-precision-mismatch.daily-paper-2026-08-08-precision-tier-llm-routing
Precision-Tier Routing: Per-Request Quantization-Level Selection for Cost-Optimal LLM Serving on H200
TL;DR — Precision-tier routing: a lightweight classifier routes each LLM request to BF16, W4A16, or NVFP4 variants of the same checkpoint, recovering near-BF16 accuracy at NVFP4 cost by serving easy requests cheaply and reserving expensive precision for hard ones.
ThakiCloud AI Research · 2026-08-08 · 📝 Tech blog (KO)
Problem
LLM serving fleets pick one… See the full description on the dataset page: https://huggingface.co/datasets/thaki-AI/daily-paper-2026-08-08-precision-tier-llm-routing.up-vla-precision-recovery-15k
UP-VLA Precision Recovery 15K
This research dataset contains 15,000 physically validated
precision-recovery samples in 150 WebDataset shards.
Each sample contains a wrist RGB image, metric depth, camera-frame XYZRGB point cloud,
and synchronized dual action trajectories:
relative end-effector SE(3) in the current tool frame;
robot joint positions on the same timestamps.
Perturbations are 5 mm to 5 cm and 5 to 30 degrees. Samples include deterministic
randomization receipts for… See the full description on the dataset page: https://huggingface.co/datasets/Shiki42/up-vla-precision-recovery-15k.red_block_precision-multicolour_block_pick_placeThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 200,
"total_frames": 107306,
"total_tasks": 4,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 25,
"splits": {
"train": "0:200"
},
"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/justintiensmith/red_block_precision-multicolour_block_pick_place.BrainStorm2026-Track1rollout_smolvla_precision-pen-placementThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/Qiu-Xinchuan/rollout_smolvla_precision-pen-placement.red_block_precisionThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 100,
"total_frames": 54547,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 25,
"splits": {
"train": "0:100"
},
"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/justintiensmith/red_block_precision.handbook-of-statistical-methods-for-precision-medicine
Dataset Card for introvoyz041/handbook-of-statistical-methods-for-precision-medicine
Dataset Description
This dataset contains images converted from PDFs using the PDFs to Page Images Converter Space.
Number of images: 482
Number of PDFs processed: 1
Sample size per PDF: 100
Created on: 2025-11-24 02:18:29
Dataset Creation
Source Data
The images in this dataset were generated from user-uploaded PDF files.
Processing Steps
PDF files were… See the full description on the dataset page: https://huggingface.co/datasets/introvoyz041/handbook-of-statistical-methods-for-precision-medicine.ms_marco_half_precisionrollout_act_precision-pen-placementThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/Qiu-Xinchuan/rollout_act_precision-pen-placement.pile_cppprecisionmembenchPrecisionMemBench is a multi-dimensional retrieval benchmark for LLM memory systems. It measures four orthogonal properties that single-turn answer-quality benchmarks cannot detect:
Retrieval precision - does the right belief surface, and only that belief, against a fixed seed corpus of 35 beliefs spanning two domain scopes, a supersession chain, and a secondary-user fixture
Noise isolation - do beliefs introduced during off-topic drift turns contaminate retrieval on subsequent unrelated… See the full description on the dataset page: https://huggingface.co/datasets/tenurehq/precisionmembench.so101_grey_cylinder_blue_cup_currentcal_precision_replacements7_strict_v1_20260811This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 7,
"total_frames": 5509,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:7"
},
"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/DylanSh/so101_grey_cylinder_blue_cup_currentcal_precision_replacements7_strict_v1_20260811.hedgehog-precision-repair
hedgehog-precision-repair
Hedgehog — precision-repair round (complete merchant extraction).
Contents
train.jsonl (1180 rows)
validation.jsonl (116 rows)
Format
JSON Lines (.jsonl), one example per line.
Provenance
Original content for the Hedgehog extraction model (Michael Anthony Falabella).
piper_precision_eeThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "piperx_left_eef_6d",
"total_episodes": 50,
"total_frames": 25969,
"total_tasks": 1,
"total_videos": 50,
"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/axiboai/piper_precision_ee.so101_grey_cylinder_blue_cup_currentcal_precision_topup16_strict_v2_20260811This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 16,
"total_frames": 12876,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:16"
},
"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/DylanSh/so101_grey_cylinder_blue_cup_currentcal_precision_topup16_strict_v2_20260811.precision-assemblyThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.0",
"robot_type": "so100",
"total_episodes": 160,
"total_frames": 165660,
"total_tasks": 1,
"total_videos": 320,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:160"
},
"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/CrazyYhang/precision-assembly.so101_precision-pen-placementPrecision2_20260626_111154This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"… See the full description on the dataset page: https://huggingface.co/datasets/AnonymousMouse404/Precision2_20260626_111154.piper_precision_jointThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "piperx_left_joint",
"total_episodes": 50,
"total_frames": 25969,
"total_tasks": 1,
"total_videos": 50,
"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/axiboai/piper_precision_joint.so101_grey_cylinder_blue_cup_currentcal_precision_topup16_v1_20260811This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"robot_type": "so_follower",
"total_episodes": 5,
"total_frames": 4082,
"total_tasks": 1,
"chunks_size": 1000,
"data_files_size_in_mb": 100,
"video_files_size_in_mb": 200,
"fps": 30,
"splits": {
"train": "0:5"
},
"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/DylanSh/so101_grey_cylinder_blue_cup_currentcal_precision_topup16_v1_20260811.Precision_20260626_110804This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"shape": [
6
],
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"… See the full description on the dataset page: https://huggingface.co/datasets/AnonymousMouse404/Precision_20260626_110804.llm-precision-fingerprints
LLM Precision Fingerprints — precision-labelled logprobs with a built-in negative control
Greedy responses and per-position top-20 logprobs from one model served at several precisions over
a sealed 40-probe set, recorded on identical hardware.
What makes this different from any other logprob dump: it ships five fp16 arms that differ
only in seed and batch composition. Those are a negative control — they let you compute the
false-positive rate of any substitution detector you… See the full description on the dataset page: https://huggingface.co/datasets/nickh007/llm-precision-fingerprints.