datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
the-pile-splitted
Dataset description
The pile is an 800GB dataset of english text
designed by EleutherAI to train large-scale language models. The original version of
the dataset can be found here.
The dataset is divided into 22 smaller high-quality datasets. For more information
each of them, please refer to the datasheet for the pile.
However, the current version of the dataset, available on the Hub, is not splitted accordingly.
We had to solve this problem in order to improve the user… See the full description on the dataset page: https://huggingface.co/datasets/ArmelR/the-pile-splitted.fafb-em-blocksCircuitSense
CircuitSense
This dataset is a comprehensive multimodal circuit question-answering benchmark designed to evaluate visual reasoning and problem-solving capabilities across three main domains: Perception, Analysis, and Design. The dataset contains structured question-answer pairs with accompanying visual content, targeting different engineering cognitive levels and reasoning tasks.
Dataset Structure
The dataset is organized into three primary folders, each containing… See the full description on the dataset page: https://huggingface.co/datasets/armanakbari4/CircuitSense.banana-vidorev3-synthetic-arms
Banana ViDoRe v3 Synthetic Arms
Domain-separated ViDoRe v3 synthetic training arms for finance and industrial adaptation.
The Hub dataset uses finance and industrial as dataset configs/subsets. Within each config, splits separate
vlm_in_batch, vlm_ocr_bm25, banana_fullpipe, and hybrid_vlm_ocr_bm25_banana_fullpipe.
Generated at: 2026-06-29T11:49:45.670912+00:00
Total JSONL rows across configs/splits: 151691.
Images are stored once per subset under… See the full description on the dataset page: https://huggingface.co/datasets/vkehfdl1/banana-vidorev3-synthetic-arms.skilltrainbench-public
skilltrainbench training tasks
The training half of the skilltrainbench benchmark suite: for each of the
four datasets, the dev_task_names of its pinned train/test split, in Harbor
task format.
The held-out/test tasks are not in this repository. Neither are the
published splits that are not the pin, nor the tasks that fall outside each
pinned split's task set. Use this repository for skill formation and
training; evaluate on the held-out half, which stays in the private source… See the full description on the dataset page: https://huggingface.co/datasets/armin-aptura/skilltrainbench-public.claude-fable-5-claude-code
claude-fable-5 Agent Traces
It's worth noting that our team was working with Glint-Research to collect as much fable data as possible.
These are just the anonymized raw traces of both of our teams combined. This means that Glint-Research/Fable-5-traces was created from formatting and splitting up this same dataset. If you use one for your tune, don't use the other (it's the same exact data).
For training on this dataset I recommend using the teich package to convert to openai… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/claude-fable-5-claude-code.hapticwam-teleop-raw
HapticWAM — teleoperated episodes (raw)
Renamed from armteam/phantom-episodes on 2026-09-19, when the project's working name PHANTOM
became HapticWAM (Haptic World-Action Model). The old id still redirects. The Python package
and CLI keep the name phantom, so task keys, checkpoint names and config keys are unchanged.
Tactile manipulation episodes for HapticWAM (Haptic World-Action Model), the
tactile world-action model: UR3 + Robotiq 2F-85 + 2x Daimon DM-Tac W2L fingertip… See the full description on the dataset page: https://huggingface.co/datasets/armteam/hapticwam-teleop-raw.neural-graphics-dataset
Neural Graphics Dataset
A compact collection of reference image sequences with accompanying motion, depth, and related rendering data, designed for training, validation, and evaluation of neural graphics models.
It is intended for use with models in the Neural Graphics Model Gym, including:
Neural Super Sampling (NSS)
Neural Frame Rate Upscaling (NFRU)
The dataset is intended as a small, practical example dataset for tutorials and experimentation. For best performance in… See the full description on the dataset page: https://huggingface.co/datasets/Arm/neural-graphics-dataset.scientific_papersScientific papers datasets contains two sets of long and structured documents.
The datasets are obtained from ArXiv and PubMed OpenAccess repositories.
Both "arxiv" and "pubmed" have two features:
- article: the body of the document, pagragraphs seperated by "/n".
- abstract: the abstract of the document, pagragraphs seperated by "/n".
- section_names: titles of sections, seperated by "/n".sharded-pilestack-exchange-instruction
Dataset Card for "stack-exchange-instruction"
More Information needed
arm_O3
REBench — arm / O3
Binary analysis dataset extracted with Ghidra 11.x from the REBench benchmark suite.
Features per row (one row = one function)
Column
Description
arch / opt_level
Architecture & optimization flag
package / binary_name
Source package and executable
original_function_name
Real symbol name (from unstripped binary)
stripped_function_name
Generic name used in stripped binary
original_code
Decompiled C with original names… See the full description on the dataset page: https://huggingface.co/datasets/Xtest/arm_O3.gr1_arms_waist-CuttingboardToPanarm_O0
REBench — arm / O0
Binary analysis dataset extracted with Ghidra 11.x from the REBench benchmark suite.
Features per row (one row = one function)
Column
Description
arch / opt_level
Architecture & optimization flag
package / binary_name
Source package and executable
original_function_name
Real symbol name (from unstripped binary)
stripped_function_name
Generic name used in stripped binary
original_code
Decompiled C with original names… See the full description on the dataset page: https://huggingface.co/datasets/Xtest/arm_O0.gr1_arms_waist-CuttingboardToCardboardBoxfull_checkbox_dropdown_radiobuttonarmnetbench_v01_lerobot_so101
ArmnetBench v0.1 — LeRobot (single-arm SO-101)
ArmnetBench v0.1 contains 50 human-teleoperated reference trajectories per task, plus
evaluation trajectories from 7 policies trained or fine-tuned on those reference datasets,
across 8 single-arm tasks on the low-cost SO-101 robot arm. Data was collected for the
ArmnetBench v0.1 benchmark using the Armnet arm farm.
This repository is the native LeRobot v3.0 release
(one multi-camera episode per row, state/action parquet +… See the full description on the dataset page: https://huggingface.co/datasets/armnet/armnetbench_v01_lerobot_so101.gr1_arms_waist-PlaceMilkToMicrowavedetails_one-man-army__UNA-34Beagles-32K-bf16-v1
Dataset Card for Evaluation run of one-man-army/UNA-34Beagles-32K-bf16-v1
Dataset automatically created during the evaluation run of model one-man-army/UNA-34Beagles-32K-bf16-v1 on the Open LLM Leaderboard.
The dataset is composed of 63 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_one-man-army__UNA-34Beagles-32K-bf16-v1.object_frame_DeformerNet_single_arm_processed_datasetUnique_Armygr1_arms_waist-WineToCabinetgr1_arms_waist-PlacematToBasketarmnetbench_v01_lerobot_bimanual_so101
ArmnetBench v0.1 — LeRobot (bimanual SO-101)
ArmnetBench v0.1 contains 50 human-teleoperated reference trajectories per task, plus
evaluation trajectories from 7 policies trained or fine-tuned on those reference datasets,
across 4 bimanual tasks on dual SO-101 arms. Data was collected for the ArmnetBench
v0.1 benchmark using the Armnet arm farm.
This repository is the native LeRobot v3.0 release
(one multi-camera episode per row, state/action parquet + packed AV1 videos).… See the full description on the dataset page: https://huggingface.co/datasets/armnet/armnetbench_v01_lerobot_bimanual_so101.gr1_arms_waist-TrayToTieredShelfgpt-5.5-agentThis dataset was generated using teich by TeichAI
Prepare these datasets for supervised fine-tuning in just a few lines of code — see the Conversion section below.
gpt 5.5 Agent Traces
This directory contains raw agent trace files generated by teich. (I also dropped in some of my own personal traces)
All assistant responses were generated by openai/gpt-5.5.
JSONL files: 88
Training-ready tools
A complete configured tools schema snapshot is embedded in the… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/gpt-5.5-agent.crossed_arm_point_clouds
Crossed Arm Point Clouds Dataset
This dataset contains 3D point cloud data captured from a LiDAR scanner for crossed arm classification in the context of robot magic trick performance.
Overview
This dataset was collected for training and evaluating the Crossed Arm Voxel Network (CAVN) architecture, a deep learning model designed for 3D point cloud classification in human-robot interaction magic performances. The data supports classification of human arm positions during… See the full description on the dataset page: https://huggingface.co/datasets/ahanjaya/crossed_arm_point_clouds.gr1_arms_waist-TrayToPlatedetails_one-man-army__una-neural-chat-v3-3-P2-OMA
Dataset Card for Evaluation run of one-man-army/una-neural-chat-v3-3-P2-OMA
Dataset automatically created during the evaluation run of model one-man-army/una-neural-chat-v3-3-P2-OMA on the Open LLM Leaderboard.
The dataset is composed of 63 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train"… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_one-man-army__una-neural-chat-v3-3-P2-OMA.arm_O1
REBench — arm / O1
Binary analysis dataset extracted with Ghidra 11.x from the REBench benchmark suite.
Features per row (one row = one function)
Column
Description
arch / opt_level
Architecture & optimization flag
package / binary_name
Source package and executable
original_function_name
Real symbol name (from unstripped binary)
stripped_function_name
Generic name used in stripped binary
original_code
Decompiled C with original names… See the full description on the dataset page: https://huggingface.co/datasets/Xtest/arm_O1.