datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
ExpertHTR-Dataset
ExpertHTR Dataset
Gated page-level handwritten text recognition data for the
ExpertHTR project.
This is a rights-filtered replacement export: all HWDB/CASIA records and
images have been removed. The repository remains gated because the remaining
upstream sources have different access conditions. It is a companion data
release for ExpertHTR, not the exact training snapshot for the published
seven-source checkpoint.
Included data
Split
Records
Purpose… See the full description on the dataset page: https://huggingface.co/datasets/DAIR-Group/ExpertHTR-Dataset.Expert-Sudoku-100kexpertqa
Dataset Card for ExpertQA
Dataset Summary
We provide here the data accompanying the paper: ExpertQA: Expert-Curated Questions and Attributed Answers. The ExpertQA dataset contains 2177 examples from 32 different fields.
Supported Tasks
The main data contains 2177 examples that can be used to evaluate new methods for estimating factuality and attribution, while the lfqa_domain and lfqa_rand data can be used to evaluate long-form question answering systems.… See the full description on the dataset page: https://huggingface.co/datasets/cmalaviya/expertqa.glaucoma-expert-cot-raw-1077
Glaucoma Expert Chain-of-Thought
Ophthalmologist six-step reasoning reports for fundus photographs, each paired with
a binary glaucoma label. 1,074 cases from LAG and Papila.
Files
file
rows
split
expert_cot_trainval.jsonl
915
train (823) + val (92)
expert_cot_test.jsonl
159
test
images/
1,074
<source>_<id>.jpg
Record schema
{
"id": "1689",
"source": "LAG",
"image": "LAG_1689.jpg",
"split": "train"… See the full description on the dataset page: https://huggingface.co/datasets/yuzhench/glaucoma-expert-cot-raw-1077.cqa-creative-writing-expert-cot-preview
CQA: Creative Quality Alignment — Research-Grade Schema v2
English
This is a public preview of Bread Studio's post-training data derived from expert judgments about creative writing. The data is structured for inspection and reuse. The full 104-item Chinese creative-writing expert knowledge-elicitation collection is not released with this repository. This public preview contains the same 4 curated samples as v1, now represented with a more precise and traceable v2… See the full description on the dataset page: https://huggingface.co/datasets/BreadStudio/cqa-creative-writing-expert-cot-preview.fine-grained-medical-reasoning
Dataset Card for Fine-Grained Medical Reasoning
Fine-grained medical reasoning QA dataset introduced in "Can LLMs Reason Like Doctors? Exploring the Limits of Large Language Models in Complex Medical Reasoning"
(Findings of EACL 2026). Manually annotated from the MedAgentsBench test_hard set,
it evaluates LLMs’ abduction, deduction, and induction capabilities, offering detailed insights into physician-like reasoning.
Dataset Details
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/expertailab/fine-grained-medical-reasoning.wire_harness_expert_sac
Wire Harness Expert SAC
Expert-policy trajectories collected from the five-mover WireHarness MuJoCo
environment for visual world-model training.
Dataset summary
20,000 episodes
3,491,570 stored observation rows
At most 300 environment transitions per episode (up to 301 stored rows,
including the initial observation)
224 x 224 RGB observations, stored as JPEG bytes in pixels
10-dimensional continuous actions
451-dimensional observations
Five task stages and… See the full description on the dataset page: https://huggingface.co/datasets/faridganbarli/wire_harness_expert_sac.IC_SHM_Expert_2
IC-SHM Expert 2 Public Augmentation
This repository contains the public-data augmentation used for IC-SHM
Expert 2 and the final Qwen3-VL-8B LoRA adapter. The original competition
images and annotations are not redistributed.
Training code and the complete experiment documentation are available at
https://github.com/HKUJasonJiang/IC-SHM-Expert-2.
Dataset contents
Added class
Images
concrete_crack
50
efflorescence
200
Total
250
Each image has… See the full description on the dataset page: https://huggingface.co/datasets/JasonXF/IC_SHM_Expert_2.qa-expert-multi-hop-qa-V1.0
Dataset Card for QA-Expert-multi-hop-qa-V1.0
This dataset aims to provide multi-domain training data for the task: Question Answering, with a focus on Multi-hop Question Answering.
In total, this dataset contains 25.5k for training and 3.19k for evaluation.
You can take a look at the model we trained on this data: https://huggingface.co/khaimaitien/qa-expert-7B-V1.0
The dataset is mostly generated using the OpenAPI model (gpt-3.5-turbo-instruct). Please read more information about… See the full description on the dataset page: https://huggingface.co/datasets/khaimaitien/qa-expert-multi-hop-qa-V1.0.Expert-Go-SFT-100K
Expert-Go-SFT-100K
Paper | Code
Expert-Go-SFT-100K is a large-scale synthetic dataset designed to "cold start" Large Language Models (LLMs) for Go-related reasoning tasks. It was introduced as part of the LoGos project, which aims to bridge the gap between general-purpose LLM reasoning and specialized expert knowledge in the game of Go.
The dataset features 100,000 samples of structured Go expertise mixed with general long Chain-of-Thought (CoT) reasoning data. It enables models to… See the full description on the dataset page: https://huggingface.co/datasets/YichuanMa/Expert-Go-SFT-100K.experts_datasets-102400
Linear Model Merging Unlocks Simple and Scalable Multimodal Data Mixture Optimization
This are the domain-specific datasets from the paper: "Linear Model Merging Unlocks Simple and Scalable Multimodal Data Mixture Optimization
" (link).
Each dataset contains 102400 VQA samples from a specific domain: General VQA, OCR, Counting & Visual Perception, Chart Understanding.
You can find many models trained on mixtures of these datasets in this Huggingface Collection.
Composition
bipea-expert-nogpqa-v3
BIPEA expert data
Source datasets
C4
WikiText-103
SlimPajama
OpenWebMath
CodeSearchNet
expertia-domain-datasets
Expertia domain datasets (Math / Physics)
Instruction-tuning pairs (system / instruction / input / output + metadata)
used to train the Expertia house specialists.
expertia-math-puro.jsonl — 45k train (+5k val): Wikidata entities with
defining formulae (P2534), formal style.
expertia-physics-puro.jsonl — 45k train (+5k val): 43k Wikidata definitional
SPARQL P2534/P274 formulae + StackExchange accepted answers + Wikipedia leads
PubChem compounds.
expertia-chemistry-puro.jsonl… See the full description on the dataset page: https://huggingface.co/datasets/OscarFeMa/expertia-domain-datasets.gomodel-go-expert-v4
GoModel Go Expert v4 Dataset
Description
A high-quality dataset for fine-tuning Qwen2.5-Coder-7B to be an expert Go software engineer
with tool-calling capabilities. This is version 4, substantially rebuilt from v3 with:
Structured messages format (not pre-rendered ChatML text)
Go AST-extracted code from real repositories using go/parser
Go 1.26 feature coverage (February 2026 release)
Senior/staff-level engineering content (architecture, distributed systems, API… See the full description on the dataset page: https://huggingface.co/datasets/mencosk/gomodel-go-expert-v4.distill-expert-535k
Distill Expert 535k
Training dataset for the distill-expert model — a 0.6B LoRA fine-tuned from Qwen3-0.6B that compresses shell/command output for AI consumption.
Contents
train.jsonl.gz — 454,710 training examples (85%)
valid.jsonl.gz — 53,458 validation examples (10%)
test.jsonl.gz — 26,832 test examples (5%)
runpod_train.py — Unsloth LoRA training script (RunPod-ready)
Total: 535,000 examples across 8 operation modes.
Modes
Mode
Examples… See the full description on the dataset page: https://huggingface.co/datasets/samuelfaj/distill-expert-535k.sciclaims_verification_data
Verification Dataset for SciClaims
This is the verification dataset used in the 2025 EMNLP demonstration paper
SciClaims: An End-to-End Generative System for Biomedical Claim Analysis
(Ortega and Gómez-Pérez).
The dataset contains approximately 4.7 million PubMed abstracts published
between 2000 and 2022. The records were selected using Semantic Scholar's
Highly Influential Citations metric, requiring each article to be supported
by at least three highly influential citations.… See the full description on the dataset page: https://huggingface.co/datasets/expertailab/sciclaims_verification_data.GlobMed_ExpertQA-Med
🌍 GlobMed: ExpertQA-Med
GlobMed_ExpertQA-Med covers 20 languages, including 13 high-resource languages (Arabic, Chinese, English, French, German, Hindi, Indonesian, Japanese, Korean, Portuguese, Russian, Spanish, and Thai) and 7 low-resource languages (Bengali, Malay, Swahili, Urdu, Wolof, Yoruba, and Zulu).
Code
ar
bn
zh
en
fr
de
hi
id
ja
ko
ms
pt
ru
es
sw
th
ur
wo
yo
zu
Language
Arabic
Bengali
Chinese
English
French
German
Hindi
Indonesian
Japanese
Korean
Malay… See the full description on the dataset page: https://huggingface.co/datasets/ruiyang-medinfo/GlobMed_ExpertQA-Med.webshop_expert_trajectoriesQwen3.8-Flash-Next-expert-activation-map
Qwen3.8-Flash-Next Expert Activation Map
Per-(layer, expert) routing and output-importance statistics for
Qwen3.8-Flash-Next (Qwen4Exp architecture, 48 MoE layers x 512 routed
experts, top-k 10), measured on the unquantized bf16 checkpoint over a
2751-prompt, 28-domain calibration corpus.
The purpose is to answer, per layer, which experts carry the model's routed
output so that expert-level decisions (bf16 protection under quantization,
offload residency, pruning, warm-start… See the full description on the dataset page: https://huggingface.co/datasets/tcclaviger/Qwen3.8-Flash-Next-expert-activation-map.French-Expert-SFT-81M-Tokens
French Expert SFT Corpus (81M Tokens)
🎯 Description
Ce dataset est un corpus de haute qualité conçu pour le Supervised Fine-Tuning (SFT). Il a été constitué par un moteur de recherche thématique profond (deep-crawl) ciblant les domaines de haute expertise technique et juridique française.
📊 Statistiques Clés
Nombre total de pépites (Samples) : 456,863
Volume estimé : ~81 Millions de Tokens
Taille moyenne par entrée : 629 caractères
Qualité : 0% doublons… See the full description on the dataset page: https://huggingface.co/datasets/Data-Elite/French-Expert-SFT-81M-Tokens.ai-expert-alpaca
AI Expert Alpaca Dataset
🚀 Empower open-source LLMs (Qwen, Gemma, etc.) for core AI domains through SFT/LoRA fine-tuning 🚀
Dataset Description
This dataset contains high-quality Q&A pairs for supervised fine-tuning (SFT) of large language models, focusing on three core AI technology domains: Large Language Models (LLM), Retrieval-Augmented Generation (RAG), and Agent Systems. The dataset provides comprehensive coverage of these cutting-edge AI technologies… See the full description on the dataset page: https://huggingface.co/datasets/GXMZU/ai-expert-alpaca.urban-vla-expert-v1
Urban VLA Expert v1
Urban VLA Expert v1 is a simulator dataset for language-conditioned urban driving. Each frame pairs a 256 x 256 front-camera image with ego state, a natural-language instruction, and continuous driving controls.
This is a small research dataset, not evidence that a policy is ready for a real vehicle. The expert is a deterministic simulator controller, and the language prompts are curated paraphrases rather than speech collected from drivers.
What… See the full description on the dataset page: https://huggingface.co/datasets/Mayank022/urban-vla-expert-v1.cnc-gcode-expert
🛠️ AInewgen CNC G-code Expert
[English below]
Dataset d'entraînement instruction → G-code expert pour le pilotage de machines CNC : fraisage, tournage, perçage, filetage, compensations d'outil et sécurité machine. Couvre les dialectes Fanuc, GRBL, Marlin, LinuxCNC, Siemens et Heidenhain.
Chaque exemple contient une instruction en français (cas réaliste d'atelier) et une réponse experte : G-code complet commenté, paramètres de coupe justifiés, et vérifications de sécurité avant… See the full description on the dataset page: https://huggingface.co/datasets/BreyAIrev/cnc-gcode-expert.GlobMed_ExpertQA-Bio
🌍 GlobMed: ExpertQA-Bio
GlobMed_ExpertQA-Bio covers 20 languages, including 13 high-resource languages (Arabic, Chinese, English, French, German, Hindi, Indonesian, Japanese, Korean, Portuguese, Russian, Spanish, and Thai) and 7 low-resource languages (Bengali, Malay, Swahili, Urdu, Wolof, Yoruba, and Zulu).
Code
ar
bn
zh
en
fr
de
hi
id
ja
ko
ms
pt
ru
es
sw
th
ur
wo
yo
zu
Language
Arabic
Bengali
Chinese
English
French
German
Hindi
Indonesian
Japanese
Korean
Malay… See the full description on the dataset page: https://huggingface.co/datasets/ruiyang-medinfo/GlobMed_ExpertQA-Bio.otaku-expert-dataset
Animetix Otaku Expert Fine-Tuning Dataset
This is the unified expert Supervised Fine-Tuning (SFT) training dataset for the Animetix Otaku Reasoning models. It is written 100% in French without code-switching.
Dataset Proportions
To ensure a balanced and robust reasoning model, the dataset is built using strict mathematical proportions:
80% Specialized Otaku Knowledge: Data-driven relational facts about anime, manga, seiyuu, French voice actors (VF), magazines… See the full description on the dataset page: https://huggingface.co/datasets/MissawB/otaku-expert-dataset.ontocord__wide_3b_sft_stage1.2-ss1-expert_news-details
Dataset Card for Evaluation run of ontocord/wide_3b_sft_stage1.2-ss1-expert_news
Dataset automatically created during the evaluation run of model ontocord/wide_3b_sft_stage1.2-ss1-expert_news
The dataset is composed of 38 configuration(s), each one corresponding 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 is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/ontocord__wide_3b_sft_stage1.2-ss1-expert_news-details.gomodel-go-expert-v7repro-multi-agent-teams-hold-experts-back-traces
Agent traces
Agent sessions published from a Trackio Logbook.
Baptist-Christian-Bible-Expert
Updated dataset and updated guide!
Comprehensive Guide for QLoRA Fine Tuning
1. Initial Guide Setup:
You can make this cut & paste easy by finding and replacing the following variables in the guide. Copy over the whole thing including brackets.
Point to your local files.
[local_pc_path_to_config_and_data]
[config.yml]
[dataset.jsonl]
Pick a name.
[runpod_model_folder_name]
SSH connection to runpod.
[serverIP]
[sshPort]
How will you upload your model will go on HF?… See the full description on the dataset page: https://huggingface.co/datasets/sleepdeprived3/Baptist-Christian-Bible-Expert.oath-frames-expert-annotations