datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
unified-reasoning-dataset
Unified Reasoning Dataset
A 94,860-row English SFT collection that normalizes four synthetic reasoning and instruction datasets into one consistent schema.
Quick start
from datasets import load_dataset
dataset = load_dataset(
"j0no12/unified-reasoning-dataset",
split="train",
)
print(dataset.column_names)
# ['thinking', 'instruction', 'response', 'source']
print(dataset[0])
Dataset summary
Property
Value
Split
train only
Rows… See the full description on the dataset page: https://huggingface.co/datasets/j0no12/unified-reasoning-dataset.physics-reasoning-dataset
📚 Flux Physics Reasoning Dataset
This dataset contains detailed physics reasoning scenarios designed to train Small Language Models (SLMs) and Liquid Neural Networks in physical intuition.
📄 Format
The dataset is provided in Parquet format (train.parquet) for efficient loading. Each row contains:
prompt: The physics question or scenario description.
answer: The correct physical explanation and answer.
concept: The underlying physics principle (e.g., "Conservation of… See the full description on the dataset page: https://huggingface.co/datasets/convaiinnovations/physics-reasoning-dataset.synthetic_vc_financial_decisions_reasoning_dataset
Best Curator Use Case in the Reasoning Datasets Competition: https://www.linkedin.com/feed/update/urn:li:activity:7330998995990781952/
Synthetic VC Financial Decisions Reasoning Dataset
Dataset Summary
The Synthetic VC Financial Decisions Reasoning Dataset is a large-scale collection designed to train, evaluate, and fine-tune language models on subjective, abstract financial reasoning tasks. It simulates venture capital (VC) workflows by capturing multiple… See the full description on the dataset page: https://huggingface.co/datasets/ZennyKenny/synthetic_vc_financial_decisions_reasoning_dataset.twi-llm-reasoning-dataset-1k
This dataset is shared under CC BY-NC 4.0, which means you are free to use, share, and adapt it for non-commercial research and educational purposes with attribution. You can read the full license at https://creativecommons.org/licenses/by-nc/4.0/.
This dataset is made available because of Ghana NLP's volunteer driven research work. Please consider contributing to any of our projects on Github
Twi Reasoning Dataset
A Twi (Akan) translation of the Multilingual-Thinking… See the full description on the dataset page: https://huggingface.co/datasets/ghanaopenai/twi-llm-reasoning-dataset-1k.logic-problems-reasoning-dataset
Dataset Card for my-distiset-a26cd729
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/sdiazlor/my-distiset-a26cd729/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/sdiazlor/logic-problems-reasoning-dataset.glm-5-1-reasoning-1m-cleaned-dataset
GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Summary
Teacher model in the data: GLM-5.1… See the full description on the dataset page: https://huggingface.co/datasets/Supbatomic/glm-5-1-reasoning-1m-cleaned-dataset.Pashto-Social-Insight-Reasoning-Dataset
Pashto Social Insight & Reasoning Dataset (PSIR)
Overview
The Pashto Social Insight & Reasoning (PSIR) dataset is a specialized collection designed to evaluate and enhance the sociological reasoning, cultural dynamics understanding, and analytical capabilities of AI models in the Pashto language. Born from an incremental "snowball effect" curation process, it captures deep contextual insights into social structures and community reasoning.
Structure… See the full description on the dataset page: https://huggingface.co/datasets/nassimjp/Pashto-Social-Insight-Reasoning-Dataset.python-reasoning-dataset
Dataset Card for my-distiset-986461
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/sdiazlor/my-distiset-986461/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/sdiazlor/python-reasoning-dataset.tamil-reasoning-dataset
Tamil Reasoning Dataset (Parallel EN-TA)
This dataset contains reasoning tasks and competitive programming questions in both English and Tamil. It is intended for training and evaluating LLMs on Tamil reasoning and translation capabilities.
Dataset Structure
The dataset consists of three parallel JSON files:
parallel_code.json: Programming-related reasoning tasks.
parallel_math.json: Mathematical reasoning tasks.
parallel_science.json: Science-related reasoning tasks.… See the full description on the dataset page: https://huggingface.co/datasets/Vedisasi/tamil-reasoning-dataset.Qwen3-Reasoning-Distill-Q-A-Dataset
Qwen3 Reasoning Distill Q&A Dataset
Repository: RefinedNeuro/Qwen3-Reasoning-Distill-Q-A-Dataset
Authors
Mehmet Can Farsak
Serhat Atayeter
License
This dataset is released under CC0 1.0 Universal (CC0 1.0) Public Domain Dedication.
Dataset Summary
This dataset contains question-answer pairs across six STEM subjects designed for Turkish-language reasoning tasks. It was generated using the qwen3-32b model and is intended for fine-tuning the RN_TR_R2… See the full description on the dataset page: https://huggingface.co/datasets/RefinedNeuro/Qwen3-Reasoning-Distill-Q-A-Dataset.high-reasoning-dataset-v1
High-Reasoning Dataset v1
2,139 premium Q&A pairs autonomously generated by a multi-AI knowledge distillation system. Each answer includes deep reasoning (CoT), mathematical foundations, production-ready code, historical context, edge cases, and real-world production incidents.
Made by Aditya Wakharkar | Tantra AI Labs
What is this dataset?
This is a synthetic training dataset created entirely by two AI models talking to each other 24/7 — no humans in the loop.… See the full description on the dataset page: https://huggingface.co/datasets/tantra-ai-labs/high-reasoning-dataset-v1.High-Quality-Synthetic-Python-Dataset-with-Reasoning-Traces-Chain-of-Thought-for-LLM-Fine-Tuning
PyReason-7k: Advanced Python Chain-of-Thought Dataset
Dataset Description
This dataset contains 7,000+ high-quality Python programming examples designed for LLM fine-tuning.
Each entry includes a detailed thought_process (Chain-of-Thought) to teach models logical reasoning before coding.
Key Features:
Chain-of-Thought: Step-by-step reasoning traces.
Error Handling: Solutions include try-except blocks and logging.
Diverse Tasks: Algorithms, API handling, Data Structures.… See the full description on the dataset page: https://huggingface.co/datasets/xTayyub/High-Quality-Synthetic-Python-Dataset-with-Reasoning-Traces-Chain-of-Thought-for-LLM-Fine-Tuning.reasoning-dataset
Reasoning & Thinking Dataset (RL/SFT Combined)
Overview
This dataset is a compiled collection of various reasoning, math, coding, and creative writing datasets designed for training reasoning models (System 2 thinking). It contains two main subsets:
RL (Reinforcement Learning): High-quality ground truth pairs augmented with task_type and rubrics for reward modeling.
SFT (Supervised Fine-Tuning): Instruction-following and thinking process data (with <think> tags).
Total… See the full description on the dataset page: https://huggingface.co/datasets/comoZ/reasoning-dataset.medical-reasoning-dataset
Dataset Card for my-distiset-2021d421
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/sdiazlor/my-distiset-2021d421/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/sdiazlor/medical-reasoning-dataset.math-python-reasoning-dataset
Dataset Card for my-distiset-3c1699f5
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/sdiazlor/my-distiset-3c1699f5/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/sdiazlor/math-python-reasoning-dataset.ccf-reasoning-dataset
Cognitive Cascade Framework (CCF) Reasoning Dataset
A high-quality dataset of structured reasoning examples using the Cognitive Cascade Framework (CCF), designed for training language models to perform systematic, multi-stage reasoning.
Dataset Description
This dataset contains problems across multiple domains (math, science, coding, creative reasoning) paired with detailed reasoning chains following the CCF methodology. Each example includes a complete reasoning trace… See the full description on the dataset page: https://huggingface.co/datasets/saberai/ccf-reasoning-dataset.astragpt-reasoning-dataset
AstraGPT Reasoning Dataset v1
193,841 instruction–reasoning–answer triples with explicit Attempts Blocks cognitive traces, released by Tantra AI Labs / AstraGPT Research.
1 · Executive Summary & Core Research Thesis
Most open reasoning datasets treat chain-of-thought as incidental output — reasoning text that happens to precede an answer. AstraGPT Reasoning Dataset v1 is constructed the other way around: the reasoning trace is the primary artifact. Every one… See the full description on the dataset page: https://huggingface.co/datasets/tantra-ai-labs/astragpt-reasoning-dataset.Deepseek-mcq-reasoning-dataset
Turkish Reasoning Dataset
A Turkish reasoning dataset generated from alibayram/turkish_mmlu using DeepSeek-V3.2 (deepseek-reasoner). Each sample contains a multiple-choice academic question paired with a step-by-step rationale and internal thinking trace.
Dataset Summary
Source: Turkish MMLU (academic exam questions from TUS, KPSS, YKS, etc.)
Size: 1,000 samples
Language: Turkish
Generator Model: DeepSeek-V3.2 (deepseek-reasoner)
Purpose: Fine-tuning language models for… See the full description on the dataset page: https://huggingface.co/datasets/AhmetSemih/Deepseek-mcq-reasoning-dataset.Nepali-Datasets-Reasoning-Grounding-V1Copyright 2026 Sandesh Bastola
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language… See the full description on the dataset page: https://huggingface.co/datasets/Matrix-Man-Lab/Nepali-Datasets-Reasoning-Grounding-V1.finance-reasoning-sft-dataset
Personal Finance Reasoning Dataset
A synthetic instruction-tuning dataset designed to teach language models to reason through personal finance and investing decisions using the mental frameworks from classic books in the genre. The goal is not recall of book content but principled reasoning: the model should apply frameworks to novel situations it has never seen.
Source Books
Principles were extracted from the following books:
The Psychology of Money — Morgan Housel
Rich… See the full description on the dataset page: https://huggingface.co/datasets/likhitjuttada/finance-reasoning-sft-dataset.Arabic-Optimized-Reasoning-Dataset
Arabic Optimized Reasoning Dataset
Dataset Name: Arabic Optimized ReasoningLicense: Apache-2.0Formats: CSVSize: 1600 rowsBase Dataset: cognitivecomputations/dolphin-r1Libraries Used: Datasets, Dask, Croissant
Overview
The Arabic Optimized Reasoning Dataset helps AI models get better at reasoning in Arabic. While AI models are good at many tasks, they often struggle with reasoning in languages other than English. This dataset helps fix this problem by:
Using fewer tokens… See the full description on the dataset page: https://huggingface.co/datasets/Jr23xd23/Arabic-Optimized-Reasoning-Dataset.Pashto-Medical-o1-Reasoning-SFT-Dataset
Pashto Medical o1 Reasoning SFT Dataset
This dataset provides medical instruction-tuning data featuring chain-of-thought (CoT) reasoning steps in Pashto, structured for Supervised Fine-Tuning (SFT) of large language models.
Dataset Structure
The dataset contains conversational message formats with step-by-step reasoning encapsulated via <think> blocks, followed by the final expert medical response.
Data Fields
Question: The medical question or… See the full description on the dataset page: https://huggingface.co/datasets/nassimjp/Pashto-Medical-o1-Reasoning-SFT-Dataset.sino-xenic-reasoning-gap-dataset
Sino-Xenic Reasoning Gap Dataset
A comprehensive evaluation dataset for testing Large Language Models' understanding of Sino-Xenic linguistic phenomena across Chinese, Japanese, Korean, and Vietnamese.
Dataset Overview
Total Samples: 297
Languages: Chinese, Japanese, Korean, Vietnamese
Categories: 11
Task Types: Surface-level and Deep Structural
Categories
Chinese Idioms (27 samples) - Understanding Chinese idioms and their cultural meanings
Chinese… See the full description on the dataset page: https://huggingface.co/datasets/nguyenthanhasia/sino-xenic-reasoning-gap-dataset.ai-reasoning-math-dataset
🧮 AI Reasoning Math Dataset
Dataset containing math word problems with step-by-step reasoning and final answers.
Designed for:
Chain-of-thought training
Reasoning model fine-tuning
Math QA benchmarking
📊 Dataset Statistics
Train: 5,000 samples
Validation: 1,000 samples
Test: 1,000 samples
Total: 7,000 samples
📄 Data Format
{
"question": "If a train travels 60 km in 1.5 hours, what is its average speed?",
"reasoning": "Average speed = distance /… See the full description on the dataset page: https://huggingface.co/datasets/Caplin43/ai-reasoning-math-dataset.team-truthowl-mixed-reasoning-dataset
Team P11 Mixed Reasoning Dataset
📊 Dataset description
HLE（Humanity's Last Exam）向けに作成した、数学中心＋科学MCの混合推論データセットです。
推論過程（Chain-of-Thought）を保持し、最終解答の正規化を行っています。
対象モデルは DeepSeek-R1-Distill-Qwen-32B、学習はQLoRAを想定しています。
🎯 Purpose
Competition: 松尾研LLMコンペ 2025
Target Model: DeepSeek-R1-Distill-Qwen-32B
Training Method: QLoRA Fine-tuning（4bit NF4, double quant）
📦 Composition
Math Hard（MATH Level≥3, HARDMath）
Math Mid（GSM8K, MetaMathQA）
Science（GPQA… See the full description on the dataset page: https://huggingface.co/datasets/weblab-llm-competition-2025-bridge/team-truthowl-mixed-reasoning-dataset.csqa-reasoning-dataset
Dataset Card for "commonsense_qa"
Dataset Summary
CommonsenseQA is a new multiple-choice question answering dataset that requires different types of commonsense knowledge
to predict the correct answers . It contains 12,102 questions with one correct answer and four distractor answers.
The dataset is provided in two major training/validation/testing set splits: "Random split" which is the main evaluation
split, and "Question token split", see paper for details.… See the full description on the dataset page: https://huggingface.co/datasets/Maxime272003/csqa-reasoning-dataset.Somali-Reasoning-Dataset
Somali-OpenHermes-Somlish-Instruct-20K 🇸🇴
This dataset is a gift to the Somali AI community. It is designed to help developers build models that are both highly intelligent and naturally conversational in our language.
🌟 What makes this unique?
This is a Hybrid Dataset that combines two powerful sources:
The Logic (18,379 rows): A Somali translation of the world-class teknium/OpenHermes-2.5. This part provides the AI with deep reasoning, mathematics, coding, and… See the full description on the dataset page: https://huggingface.co/datasets/Zyroxx66/Somali-Reasoning-Dataset.refined-reasoning-dataset
Refined Reasoning Dataset
High-quality reasoning dataset refined for fine-tuning language models on complex reasoning tasks.
Dataset Summary
This dataset contains 10,277 examples of reasoning-intensive prompts and responses, available in two formats for supervised fine-tuning (SFT) of instruction-tuned language models.
Total Examples: 10,277
Training Split: 9,777 examples (95%) / 9,650 examples (ShareGPT - filtered)
Validation Split: 500 examples (5%) / 492 examples… See the full description on the dataset page: https://huggingface.co/datasets/yakuraku/refined-reasoning-dataset.nvidia-nemotron-model-reasoning-dataset-turkish
Nemotron Reasoning Challenge - Turkish
Turkish translation of the training data from NVIDIA's Nemotron Model Reasoning Challenge
Each row is a reasoning puzzle framed in an "Alice's Wonderland" setting. Given a few input/output examples, the model needs to figure out the hidden rule and apply it to a new input.
Category
Rows
Description
bit
1602
Hidden bit manipulation rule on 8-bit binary numbers
grav
1597
Falling distance with a modified gravitational constant… See the full description on the dataset page: https://huggingface.co/datasets/mramazan/nvidia-nemotron-model-reasoning-dataset-turkish.v-reasoning-dataset
Dataset Card for v-reasoning-dataset
This dataset has been created with distilabel.
Dataset Summary
This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:
distilabel pipeline run --config "https://huggingface.co/datasets/vinhnx90/v-reasoning-dataset/raw/main/pipeline.yaml"
or explore the configuration:
distilabel pipeline info --config… See the full description on the dataset page: https://huggingface.co/datasets/vinhnx90/v-reasoning-dataset.