datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
industrial-instruction-dataset
Industrial-Instruction Dataset
Industrial-Instruction provides benchmark and training-ready QA instances derived from industrial technical reports, designed to evaluate robustness under realistic retrieval conditions. Samples are grounded in retrieved evidence and include irrelevant retrieval, single-/multi-document support, and single-/multi-document answer settings.
Paper
Industrial-Instruction: An End-to-End Framework for Building Instruction-Tuning and… See the full description on the dataset page: https://huggingface.co/datasets/Parssky/industrial-instruction-dataset.cpt_instruction_datasets
Instruction datasets
Collection of synthetic instruction datasets used during the continued pretraining of Model-small-instr-1, Model-small-instr-2 and Model-small-instr-3. You can currently find these models under: Llama-3.1-Carballo-Instr1 and Llama-3.1-Carballo-Instr3.
Dataset creation
Datasets were created using two different techniques:
Adapting already existing datasets or corpora by modifying their format to make them suitable for including instructions during… See the full description on the dataset page: https://huggingface.co/datasets/proxectonos/cpt_instruction_datasets.incremental-instruction-creative-writing
Incremental Instruction Creative Writing
Does delivering a writing brief over several conversation turns change what a
language model writes? This dataset supports that question with matched
creative-writing tasks evaluated under two delivery conditions:
FULL: the complete brief is supplied in one turn.
SHARDED: the same intended brief is introduced across five to nine turns.
The benchmark holds task content fixed while varying how the instructions are
delivered. It is… See the full description on the dataset page: https://huggingface.co/datasets/SolusOps/incremental-instruction-creative-writing.fine-instructions-ita-70k
🍷🇮🇹 Fine Instructions Ita 70k
A good instruction dataset in Italian, generated with LLM-aided translation.
LLM-aided translation
The starting point has been mlabonne/FineTome-100k: a subset of arcee-ai/The-Tome, re-filtered using HuggingFaceFW/fineweb-edu-classifier to identify examples with good educational value. FineTome is a high quality dataset containing diverse elements such as conversations, reasoning problems, and more.
The translation process includes these… See the full description on the dataset page: https://huggingface.co/datasets/anakin87/fine-instructions-ita-70k.AddisGPT-Amharic-Instruction
AddisGPT-Amharic-Instruction
A human-verified, fully conversational Amharic instruction-tuning dataset sourced entirely from real AddisGPT user interactions.
796 curated instruction–output pairs spanning 14 topics, drawn exclusively from anonymized conversations with AddisGPT — an Amharic-first AI assistant serving Ethiopian and diaspora communities. Every pair is an organic user question paired with the assistant's response; there is no synthetic, templated, or third-party… See the full description on the dataset page: https://huggingface.co/datasets/AddisGPT/AddisGPT-Amharic-Instruction.odia-instruction-dataset
Odia Instruction Following Dataset
Dataset Description
This is a comprehensive Odia language instruction-following dataset designed for training conversational AI models, chatbots, and instruction-following systems in Odia (ଓଡ଼ିଆ). The dataset contains high-quality instruction-response pairs that enable models to understand and follow instructions in the Odia language.
Dataset Summary
Language: Odia (ଓଡ଼ିଆ)
Total Records: 324,560
Format:… See the full description on the dataset page: https://huggingface.co/datasets/abhilash88/odia-instruction-dataset.cybersecurity-controls-instructions
Cybersecurity Controls Instructions
Security control, incident response and risk management guidance from NIST Special Publications, turned into instruction-following examples.
Splits
split
rows
source documents
train
13,106
56
validation
4,840
18
test
5,697
18
Splits are held out by source document. Every chunk yields several
instruction rows, so a random row-level split would place the same passage in
train and test; whole documents are held… See the full description on the dataset page: https://huggingface.co/datasets/nuhmanpk/cybersecurity-controls-instructions.Instruction-280K-Turkish
Dataset Card for Instruction-280K-Turkish
Language: Turkish
Dataset Description
This repository contains a dataset for Turkish version of the Deepseek 1.5B model. The translation was performed using the Google translation model to ensure high-quality, accurate translation.
Dataset Details
Size: ≈280K
Translation tool: Google Translate
Data format: Prompt, Response
ilang-instruction-corpus
I-Lang Instruction Corpus
Paired natural language instructions and their I-Lang protocol equivalents, for training and evaluating AI systems on structured communication.
What is I-Lang?
I-Lang is a communication protocol for human-AI and AI-to-AI interaction. It compresses natural language instructions into structured, machine-readable notation using 52 verbs, 28 modifiers, and 14 entities with pipe syntax.
Protocol Spec: github.com/ilang-ai/ilang-spec
Dictionary:… See the full description on the dataset page: https://huggingface.co/datasets/i-Lang/ilang-instruction-corpus.rephrased-instruction-turkish-poems
This is a rephrased version of my previous dataset beratcmn/instruction-turkish-poems. I used the same instructions but I rephrased them to be more clear and understandable also added more variety to the format.
asha-instructions-hi-mr-v1
ASHA-Saathi Instructions (Hindi + Marathi) v1
A reusable Indic-medical instruction-tuning dataset for fine-tuning small language models to assist ASHA workers — India's ~1 million government-employed frontline community-health workers — in Hindi and Marathi.
Built as the training corpus for ombhojane/gemma-4-e2b-asha-it, submitted to the Gemma 4 Good Hackathon. Released independently so other researchers can fine-tune any small Indic-language LM on the same task.
Quick… See the full description on the dataset page: https://huggingface.co/datasets/ombhojane/asha-instructions-hi-mr-v1.Magpie-Tanuki-Instruction-Selected-Evolved-26.5k
Magpie-Tanuki-Instruction-Selected-Evolved-26.5k
概要
以下の手順で作成した約2万6500件の日本語の合成instructionデータセットです。
Magpieの手法をteam-hatakeyama-phase2/Tanuki-8x8B-dpo-v1.0-GPTQ-8bitに適用し、約10万件のinstructionを作成
cl-nagoya/ruri-largeを使ってinstructionのベクトル表現を取得
この時点のデータはAratako/Magpie-Tanuki-Instruction-100k-Embeddingsで公開されています。
取得したベクトル表現を元に、Mini Batch K-Meansによって20000個のクラスタにクラスタリング
各クラスタから最大3個までinstructionを抽出
上記で抽出した約2万6500件のinstructionに対し、Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8を使ってEvol-Instructを適用… See the full description on the dataset page: https://huggingface.co/datasets/Aratako/Magpie-Tanuki-Instruction-Selected-Evolved-26.5k.instruction_conflict_resolution_v01Instruction Conflict Resolution v0.1
This evaluation dataset tests how models resolve conflicting instructions.
It targets a common failure mode: following the most recent or most forceful instruction even when it conflicts with higher-priority constraints.
This is not training data.
What it tests
Priority handling under instruction conflict
Refusal stability under escalation
Logical conflict handling for impossible constraints
Post-conflict integrity with no delayed leakage… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/instruction_conflict_resolution_v01.naruto-instruction-promptstext_in_number_tulu-3-sft-personas-instruction-following
RU
Набор данных содержит в себе текст и его представление в виде 610-ти значного числа. Число полоучено при помощи модели.Исходный набор данных: allenai/tulu-3-sft-personas-instruction-following
EN
The dataset contains text and its representation as a 610-digit number. The number is hollowed out using model.Initial dataset: allenai/tulu-3-sft-personas-instruction-following
spai-ss6-corpus-thai-instruction-sft-suraponn
SPAI SS6 Thai Instruction SFT Suraponn Index
Index repo for the Suraponn Thai instruction SFT dataset mirrored in the canonical repo.
This is a lightweight index dataset repo. It does not duplicate the full corpus.
The full Parquet data lives in the canonical repository config below.
Canonical Data
Canonical repo: SPAISS6F1/spai-ss6-llm-1b-thai-corpus
Canonical config: thai_instruction_sft_suraponn
Rows in canonical config: 131,907
Parquet size in canonical… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/spai-ss6-corpus-thai-instruction-sft-suraponn.spai-ss6-corpus-thai-idioms-instruction
SPAI SS6 Thai Idioms Instruction Index
Index repo for the Thai idioms instruction dataset mirrored in the canonical repo.
This is a lightweight index dataset repo. It does not duplicate the full corpus.
The full Parquet data lives in the canonical repository config below.
Canonical Data
Canonical repo: SPAISS6F1/spai-ss6-llm-1b-thai-corpus
Canonical config: thai_idioms_instruction
Rows in canonical config: 1,152
Parquet size in canonical config: 0.00 GB
Source… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/spai-ss6-corpus-thai-idioms-instruction.spai-ss6-corpus-thai-local-instruction-v2
SPAI SS6 Thai Local Instruction V2 Index
Index repo for the Thai local instruction v2 dataset mirrored in the canonical repo.
This is a lightweight index dataset repo. It does not duplicate the full corpus.
The full Parquet data lives in the canonical repository config below.
Canonical Data
Canonical repo: SPAISS6F1/spai-ss6-llm-1b-thai-corpus
Canonical config: thai_local_instruction_v2
Rows in canonical config: 39,829
Parquet size in canonical config: 0.00 GB… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/spai-ss6-corpus-thai-local-instruction-v2.wikipedia-pt-br-instructions
wikipedia-pt-br-instructions-gemma
Dataset sintético de Instruction Following em português brasileiro, derivado de artigos da Wikipedia pt-BR.
Origem
Os exemplos derivam de costadev00/wikipedia-pt-br-extract e preservam source_page_id, source_title, source_dataset e a licença herdada cc-by-sa-3.0.
Processo
A geração é inspirada em Alpaca e Self-Instruct: cada artigo válido passa por uma chamada de analista documental que produz candidatos de… See the full description on the dataset page: https://huggingface.co/datasets/br-llm-data/wikipedia-pt-br-instructions.synthetic-queries-and-ml-instructions
Synthetic Dataset: Queries and ML Instructions
Dataset Description
Dataset Summary
This is a synthetic dataset with queries in Slovak language and ML instructions. The dataset was designed to train a model for extracting structured machine learning task requirements from natural language user queries.
The dataset contains user queries in Slovak describing ML tasks paired with structured JSON outputs containing task attributes like dataset modality, task type… See the full description on the dataset page: https://huggingface.co/datasets/kinit/synthetic-queries-and-ml-instructions.synthetic-conversations-and-ml-instructions
Synthetic Dataset: Conversations and ML Instructions
Dataset Description
Dataset Summary
This is a synthetic dataset of 5,000 Slovak multi-turn ML advisory conversations paired with structured JSON outputs. The dataset was designed to train and evaluate models that extract machine learning task requirements from realistic, natural Slovak dialogue, including cases where requirements are revealed gradually or changed during the conversation.
Each… See the full description on the dataset page: https://huggingface.co/datasets/kinit/synthetic-conversations-and-ml-instructions.spai-ss6-corpus-thai-synonym-instruction
SPAI SS6 Thai Synonym Instruction Index
Index repo for the Thai synonym instruction dataset mirrored in the canonical repo.
This is a lightweight index dataset repo. It does not duplicate the full corpus.
The full Parquet data lives in the canonical repository config below.
Canonical Data
Canonical repo: SPAISS6F1/spai-ss6-llm-1b-thai-corpus
Canonical config: thai_synonym_instruction
Rows in canonical config: 167
Parquet size in canonical config: 0.00 GB
Source… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/spai-ss6-corpus-thai-synonym-instruction.wikipedia-pt-br-instructions
wikipedia-pt-br-instructions-gemma
Dataset sintético de Instruction Following em português brasileiro, derivado de artigos da Wikipedia pt-BR.
Origem
Os exemplos derivam de costadev00/wikipedia-pt-br-extract e preservam source_page_id, source_title, source_dataset e a licença herdada cc-by-sa-3.0.
Processo
A geração é inspirada em Alpaca e Self-Instruct: cada artigo válido passa por uma chamada de analista documental que produz candidatos de instrução ancorados… See the full description on the dataset page: https://huggingface.co/datasets/costadev00/wikipedia-pt-br-instructions.spai-ss6-corpus-thai-tourist-attraction-instruction
SPAI SS6 Thai Tourist Attraction Instruction Index
Index repo for the Thai tourist-attraction instruction dataset mirrored in the canonical repo.
This is a lightweight index dataset repo. It does not duplicate the full corpus.
The full Parquet data lives in the canonical repository config below.
Canonical Data
Canonical repo: SPAISS6F1/spai-ss6-llm-1b-thai-corpus
Canonical config: thai_tourist_attraction_instruction
Rows in canonical config: 51,662
Parquet size… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/spai-ss6-corpus-thai-tourist-attraction-instruction.