datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Swallow-Nemotron-Post-Training-Dataset-v1
Swallow-Nemotron-Post-Training-Dataset-v1
The Swallow LLM Project constructed the Swallow-Nemotron-Post-Training-Dataset-v1 based on the math, code, and stem subsets of the NVIDIA Nemotron-Post-Training-Dataset-v1, as illustrated in the figure below.
Dataset Construction
The original Thinking Trajectories and Assistant Outputs in the Nemotron-Post-Training-Dataset-v1 were synthesized using DeepSeek-R1-0528.
However, we identified an issue with the Thinking… See the full description on the dataset page: https://huggingface.co/datasets/tokyotech-llm/Swallow-Nemotron-Post-Training-Dataset-v1.3112_llm_70b_trainingTurkish-LLM-v10-Training
Turkish LLM Training Dataset v10
A curated corpus of 144,022 Turkish instruction-completion pairs used to train the Turkish LLM Family.
Dataset Description
This dataset was created to address the scarcity of high-quality Turkish instruction-following data for language model fine-tuning. It covers a broad range of topics including:
Science & Technology (physics, chemistry, biology, computer science)
History & Geography (Turkish and world history, geography)
General… See the full description on the dataset page: https://huggingface.co/datasets/ogulcanaydogan/Turkish-LLM-v10-Training.Bitext-customer-support-llm-chatbot-training-dataset-spanish
Spanish Customer Support LLM Chatbot Training Dataset
Spanish-language adaptation of the Bitext Customer Support LLM Chatbot Training Dataset.
This dataset is intended for training and evaluating Spanish-language customer-support chatbots and instruction-following large language models.
Dataset Details
Dataset Description
This dataset is a Spanish translation and adaptation of the original Bitext Customer Support LLM Chatbot Training Dataset.
The… See the full description on the dataset page: https://huggingface.co/datasets/Faramir/Bitext-customer-support-llm-chatbot-training-dataset-spanish.Training-fr-basecncf-raw-data-for-llm-training
CNCF Raw Data for LLM Training
Description
This dataset, named cncf-raw-data-for-llm-training, consists of markdown (MD) and PDF content extracted from various project repositories within the CNCF (Cloud Native Computing Foundation) landscape. The data was collected by fetching MD and PDF files from different CNCF project repositories and converting them into JSON format. This dataset is intended as raw data for training large language models (LLMs).
The dataset includes… See the full description on the dataset page: https://huggingface.co/datasets/Kubermatic/cncf-raw-data-for-llm-training.resume-conversations-llm-training
📄 Resume Conversations for LLM Training
High-quality conversational dataset for building AI that understands resumes, careers, and professional growth.Created and maintained by Syncora.ai.
✅ Overview
This dataset provides resume-related conversations in a structured JSONL format, ideal for developers and AI practitioners working on chatbots, career advisory tools, or LLM fine-tuning. It includes realistic Q&A on career development, technology trends, and professional… See the full description on the dataset page: https://huggingface.co/datasets/strova-ai/resume-conversations-llm-training.ko-legal-embedding-training-v1
Korean Public Legal Embedding Training v1
실제 embedding 학습 queue가 소비하는 provenance-preserving 파생 JSONL이다.
source-native query/positive 관계와 deterministic bootstrap negative로 컴파일했다. 최종 학습 전 current-student hard-negative mining을 수행해야 한다.
rows: 250,000
release eligible: true
visibility: public
use: public redistribution and model training
exact benchmark query/evaluation-text matches: 0
exact retrieval-corpus matches: 0 unique hashes
Sionic 9 및 MTEB task-family train source가 포함될 수… See the full description on the dataset page: https://huggingface.co/datasets/LLM-OS-Models2/ko-legal-embedding-training-v1.3.1Million_KASHMIRI_text_Pre_training_Dataset_for_LLM_2026_by_HNM
DATASET NAME: KS-LIT-3M
Kashmiri Pretraining Dataset
This repository hosts a meticulously processed Kashmiri text dataset, specifically designed for pretraining Large Language Models (LLMs) from scratch. The dataset has undergone extensive cleaning and preprocessing to ensure high quality and suitability for robust model training.
Dataset Description
This dataset consists of a continuous one stream of Kashmiri text, cleaned to remove English… See the full description on the dataset page: https://huggingface.co/datasets/Omarrran/3.1Million_KASHMIRI_text_Pre_training_Dataset_for_LLM_2026_by_HNM.Training-small-baseLLM_PRMs_training_data_judgedLLM-QE-Retriever-Training-Datafed-ledger-llm-training-dummyLLM-QE-DPO-Training-Dataipl-training-llm-datasetExpandR_llm_training_datavoxpure_llm_training
brand: VOXPURE
document_type: Knowledge Base
last_updated: 2026-03-13
authority: Scientific DOI 10.5281/zenodo.16416493
VOXPURE MASTER KNOWLEDGE BASE [GLOBAL B2B AUTHORITY]
[Section: Training Data Purpose] {ID: DATA_01}
This knowledge base is supported by the voxpure_training_data.jsonl dataset.
Purpose: This dataset is designed for machine-based learning, specifically for fine-tuning models to understand VOXPURE's brand identity, technical specifications, and… See the full description on the dataset page: https://huggingface.co/datasets/TheElim/voxpure_llm_training.llm_subtraction_training
Systematic Subtraction Curriculum (State-Persistent)
This dataset is a tiered subtraction curriculum designed for small LLMs (0.5B - 1.5B). It solves the "Hidden State" problem by explicitly tracking variable mutations during borrowing.
Curriculum Structure
Tier 1: Mental Math (20%) - Direct subtraction for results < 100 to build instant recall.
Tier 2: Debt & Negatives (10%) - Teaches the concept of negative results when subtracting larger from smaller.
Tier 3:… See the full description on the dataset page: https://huggingface.co/datasets/shreeman-iyer/llm_subtraction_training.basic_llm_math_trainingTeaches LLMs basic math for the first 100 numbers.
Home-LLM-trainingllm_training_coursepii_dataset_for_training_llmllm_arithmetic_training
Parallax-CoT: 1GB Arithmetic Reasoning Dataset
Overview
This dataset is designed for Curriculum Learning in Small Language Models (SLMs). It focuses on "weight hardening"—strengthening the internal attention mechanisms of models (specifically Parallax 0.5B) to prepare them for complex symbolic reasoning, code generation, and high-level mathematics (Calculus/Physics).
Dataset Structure
The data follows a Chain-of-Thought (CoT) format wrapped in… See the full description on the dataset page: https://huggingface.co/datasets/shreeman-iyer/llm_arithmetic_training.llm_addition_training
LLM Addition Training Dataset (Odometer-Style Logic)
Overview
This dataset is designed to teach Large Language Models (LLMs) the foundational logic of addition through Chain-of-Thought (CoT) and Place-Value Expansion. Instead of simple $A + B = C$ pairs, this dataset forces the model to "think" through the process of splitting numbers into their constituent parts (units, tens, hundreds, thousands) and adding them step-by-step.
Dataset Structure
The… See the full description on the dataset page: https://huggingface.co/datasets/shreeman-iyer/llm_addition_training.llm_division_training
Triple-Tier Division Curriculum (Partial Quotients)
A structured curriculum designed to teach the concept of "Sharing" and "Chunking" to small language models.
Curriculum Structure
Tier 1: Division Tables (1-100) - Rote memorization of clean divisors to establish factor-pair weights.
Tier 2: Signs & Remainders - Introduces the arithmetic rules for negative divisors and the concept of "leftovers" ($R$).
Tier 3: Partial Quotients (Large) - Teaches a "Chunking"… See the full description on the dataset page: https://huggingface.co/datasets/shreeman-iyer/llm_division_training.llm-training-datallm_multiplication_training
Triple-Tier Multiplication Curriculum (Area Model)
A structured curriculum for teaching multiplication to small language models. This dataset moves from rote memorization to logical decomposition.
Curriculum Structure
Tier 1: Basic Tables (1-100) - Sequential and shuffled pairs for results up to 1000 to establish "Instant Recall" weights.
Tier 2: Sign Logic - Explicitly teaches the rules of arithmetic signs: $(+) \times (+)$, $(-) \times (+)$, and $(-) \times… See the full description on the dataset page: https://huggingface.co/datasets/shreeman-iyer/llm_multiplication_training.