datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
OpenMath-Vision-CoT-10kScaffold-CoT
Scaffold-CoT
Structured chain-of-thought training data with 3,726,548 examples in 76 JSONL shards.
Fields
Every row has exactly four top-level fields:
Field
Contents
metadata
domain, subdomain, difficulty, length_bucket
input
Ordered user messages as {index, content} objects
cot
Ordered {index, type, content} events, including reasoning, tool calls, and tool results
output
Ordered final assistant answers as {index, content} objects
The index… See the full description on the dataset page: https://huggingface.co/datasets/Specific-Labs/Scaffold-CoT.Reason-RFT-CoT-Dataset
🤗 Reason-RFT CoT Dateset
The full dataset used in our project "Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning".
⭐️ Project │ 🌎 Github │ 🔥 Models │ 📑 ArXiv │ 💬 WeChat
🤖 RoboBrain: Aim to Explore ReasonRFT Paradigm to Enhance RoboBrain's Embodied Reasoning Capabilities.
♣️ Quick Start
Please refer to Dataset Preparation
🔥 Overview
Visual reasoning abilities play a crucial role in understanding complex multimodal… See the full description on the dataset page: https://huggingface.co/datasets/tanhuajie2001/Reason-RFT-CoT-Dataset.LLaVA-CoT-100k
Dataset Card for LLaVA-CoT
The LLaVA-CoT-100k dataset is introduced in the paper LLaVA-CoT: Let Vision Language Models Reason Step-by-Step. This dataset is designed to enable Vision-Language Models (VLMs) to perform autonomous multistage reasoning, integrating samples from various visual question-answering sources with structured reasoning annotations. It aims to address the challenges VLMs face in systematic and structured reasoning for complex visual question-answering tasks.… See the full description on the dataset page: https://huggingface.co/datasets/Xkev/LLaVA-CoT-100k.CoTaEval
CoTaEval Dataset
CoTaEval Dataset is used to evaluate the feasibility and the side effects of copyright takedown methods for language models. The dataset has two domains: News and Books.
For News, it has three subsets: news_for_unlearning (for unlearning use), news_infringement(for infringement evaluation), and news_utility (for utility evaluation);
For Books, it has two subsets: books_infringement (for infringement evaluation), books_utility (for utility evaluation). The… See the full description on the dataset page: https://huggingface.co/datasets/boyiwei/CoTaEval.W2-VLA-CoT
World-to-Wrist: Offline CoT Labels
This dataset contains frame-aligned offline chain-of-thought annotations used
to train W²-VLA policies on LIBERO, RoboTwin, and four real-world manipulation
tasks. Matching LeRobot action data is available in W2-VLA-Training-Data.
Dataset Structure
W2-VLA-CoT/
├── libero/
│ ├── libero_10_no_noops_1.0.0_lerobot/
│ ├── libero_goal_no_noops_1.0.0_lerobot/
│ ├── libero_object_no_noops_1.0.0_lerobot/
│ └──… See the full description on the dataset page: https://huggingface.co/datasets/yuuu94/W2-VLA-CoT.MedQA-Mixtral-CoT
Dataset Card for medqa-cot
Synthetically enhanced responses to the medqa dataset using mixtral.
Dataset Details
Dataset Description
To increase the quality of answers from the training splits of the MedQA dataset, we leverage Mixtral-8x7B to generate Chain of Thought(CoT) answers. We create a custom prompt for the dataset, along with a
hand-crafted list of few-shot examples. For a multichoice answer, we ask the model to rephrase and explain the question… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/MedQA-Mixtral-CoT.SWE-Fixer-Train-Editing-CoT-70Kmedical_cot
Medical Question-Answering Dataset
A comprehensive collection of medical questions and detailed answers, designed for training and evaluating medical question-answering systems.
Dataset Description
Overview
This dataset contains medical questions with multiple-choice answers and detailed explanations. Each question presents a clinical scenario and requires medical knowledge to determine the correct diagnosis, treatment, or underlying mechanism.
Data… See the full description on the dataset page: https://huggingface.co/datasets/blue-blues/medical_cot.headqa-cot-llama31
headqa-cot
Synthetically enhanced responses to the HeadQA dataset. Used to train Aloe-Beta model.
Dataset Details
Dataset Description
To increase the quality of answers from the training splits of the HeadQA dataset, we leverage Llama-3.1-70B-Instruct to generate Chain of Thought(CoT) answers. We create a custom prompt for the dataset, along with a hand-crafted list of… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/headqa-cot-llama31.MMLU-medical-cot-llama31
MMLU-medical-cot
Synthetically enhanced responses to the medical-related questions of the auxiliary train set of the MMLU dataset. Used to train Aloe-Beta model.
Dataset Details
Dataset Description
First, we use Llama-3.1-70B-Instruct to filter the medical-related questions of the auxiliary train set of the MMLU dataset. Next, we leverage Mixtral-8x7B to… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/MMLU-medical-cot-llama31.Cot-Drop
LuminaSFT
LuminaSFT is a synthetic SFT dataset suite specifically designed to improve both general-purpose and task-specific SLMs. LuminaSFT consists of multiple curated splits that target diverse capabilities:
UltraChat200K-DeepSeek - A regenerated base SFT dataset for broad instruction following.
InstructGPT-NaturalQA and InstructGPT-TriviaQA - Factual question answering datasets to strengthen knowledge recall and answer accuracy.
CoT-Drop - A reading comprehension dataset with… See the full description on the dataset page: https://huggingface.co/datasets/amd/Cot-Drop.cot-leaderboard-resultsFable-5-CoT-TracesPersonal collection of Fable 5 reasoning traces.
Filter out the decoy ones and you're good.
Have fun! (Also, star my repo https://github.com/FusionCube18712/claude-codex-auto-resume if you can)
Happy distilling.
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.MedMCQA-Mixtral-CoT
Dataset Card for medmcqa-cot
Synthetically enhanced responses to the medmcqa dataset using mixtral.
Dataset Details
Dataset Description
To increase the quality of answers from the training splits of the MedMCQA dataset, we leverage Mixtral-8x7B to generate Chain of Thought(CoT) answers. We create a custom prompt for the dataset, along with a
hand-crafted list of few-shot examples. For a multichoice answer, we ask the model to rephrase and explain the… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/MedMCQA-Mixtral-CoT.LMOD-Cataract-1K-surgical-analysis-cot
Cataract-1K LLM-Generated Surgical Instructions
Dataset Overview
This dataset is derived from the Cataract-1K dataset (part of the LMOD benchmark) and enhanced using Qwen3-VL-30B-A3B-Thinking, a large vision-language model with reasoning capabilities. It is designed for training medical AI systems to provide actionable surgical guidance with transparent reasoning.
Generation Process
Source Data: Cataract-1K processed frames with segmentation annotations… See the full description on the dataset page: https://huggingface.co/datasets/mehti/LMOD-Cataract-1K-surgical-analysis-cot.deepseek-r1-distill-cotCreated by several different models:
DeepSeek R1
DeepSeek R1 Distill Qwen 14B
Qwen3.8 27B
format:
{"quest": ..., "aswer": "..."}
ty
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.salabs-stem-deep-reasoning-cot-v13
🧪 SALabs Multi-Domain STEM Deep Reasoning & Chain-of-Thought (CoT) Corpus (v13.0)
[!IMPORTANT]
💳 Click Here to Purchase Enterprise Commercial License ($2,500 USD) & Instant 31.7MB Master Archive DownloadInstant download of the full lossless master package containing all 1,816 JSONL reasoning records + 13 complete uncompressed text corpora (31.72 MB uncompressed total) + commercial license certificate.
🌟 Executive Summary
The SALabs STEM Deep Reasoning… See the full description on the dataset page: https://huggingface.co/datasets/suitai/salabs-stem-deep-reasoning-cot-v13.Bilge-Turkish-CoT-50K
Bilge: Turkish Chain-of-Thought Dataset (50K)
50,000 örneklik Türkçe Chain-of-Thought (CoT) reasoning fine-tuning veri seti.
Bilge, Türkçe büyük dil modellerinin adım adım düşünme (reasoning) kapasitesini
geliştirmek amacıyla hazırlanmış bir Chain-of-Thought veri setidir.
Veri setindeki her örnek, modelin önce <think> blokları içinde görünür bir
muhakeme süreci yürütmesini, ardından kullanıcıya yapılandırılmış ve detaylı
bir cevap vermesini öğretmek üzere tasarlanmıştır.
Bu… See the full description on the dataset page: https://huggingface.co/datasets/bugrabilge/Bilge-Turkish-CoT-50K.2026-08-25-difficult-advice-716-verbose-cot
Difficult-advice reasoning, expanded ~3x (716 records, no other data)
field
value
experiment
Does deliberation LENGTH change alignment behaviour, holding the ideas deliberated constant? These are the 716 difficult-advice exchanges of LASR-Callum/2026-08-13-haiku45-sonnet45-difficult-advice-diversity-gated-voice-linted, with the assistant's private reasoning rewritten about three times longer while carrying the same content. The user turn, the system prompt and the… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-08-25-difficult-advice-716-verbose-cot.medqa-cot-llama31
medqa-cot-llama31
Synthetically enhanced responses to the MedQa dataset. Used to train Aloe-Beta model.
Dataset Details
Dataset Description
To increase the quality of answers from the training splits of the MedQA dataset, we leverage Llama-3.1-70B-Instruct to generate Chain of Thought(CoT) answers. We create a custom prompt for the dataset, along with a hand-crafted… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/medqa-cot-llama31.Dans-Benchmaxx-COTR-CoTmedmcqa-cot-llama31
medqa-cot-llama31
Synthetically enhanced responses to the MedMCQA dataset. Used to train Aloe-Beta model.
Dataset Details
Dataset Description
To increase the quality of answers from the training splits of the MedMCQA dataset, we leverage Llama-3.1-70B-Instruct to generate Chain of Thought(CoT) answers. We create a custom prompt for the dataset, along with a… See the full description on the dataset page: https://huggingface.co/datasets/HPAI-BSC/medmcqa-cot-llama31.Deepseek-v4.1-CoTDatasets created by deepseek-ai/DeepSeek-V4.1-Flash.
You can use them to distill other models.
Ty!
chem_cot_distilledTurkish-CoT-Instruct-Dataset
🇹🇷 Turkish CoT Instruct Dataset
Türkçe Düşünme Zinciri (Chain-of-Thought) İçeren Talimat Veri Seti
Bu veri seti, modellerin Türkçe adım adım akıl yürütme (reasoning) yeteneğini
geliştirmek için hazırlanmıştır. Her örnekte model, cevabı vermeden önce
<think> ... </think> etiketleri arasında tamamen Türkçe olarak adım adım
düşünür, ardından ayrıntılı bir nihai cevap sunar (DeepSeek-R1 tarzı biçim).
Örnek sayısı: 4.868
Dil: Türkçe
Biçim: Sohbet (messages) — system / user /… See the full description on the dataset page: https://huggingface.co/datasets/AlicanKiraz0/Turkish-CoT-Instruct-Dataset.DAG-MATH-Formatted-CoT
Benchmark Overview
This dataset card contains 2,894 gold-standard DAG-MATH formatted CoT from problems from Omni-MATH.
Top‑Level Schema
Each JSON file is a list with a single object describing the problem:
problem_id: integer identifier of the problem.
domain: list of strings describing the topic taxonomy.
difficulty: numeric difficulty indicator from 1 (easiest) to 6 (hardest).
problem_text: problem statement.
sample_id: sample identifier for the solution trace.… See the full description on the dataset page: https://huggingface.co/datasets/yuanhezhang/DAG-MATH-Formatted-CoT.