datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Scaffold-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.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.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.Turkish-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.regulatory-compliance-cot-trial
⚡ Regulatory Compliance & Legal CoT Dataset for Enterprise Agents (Free Trial)
Official Open‑Source Evaluation Package (50 Rows Subset) by springofwindslabs
👉 Looking for full production data?
The complete 1,000-row standard volume and 2,300+ row mutually exclusive, non-overlapping extended package (Total 3,300+ unique rows) are fully available for commercial deployment via our official procurement gateway:
➔… See the full description on the dataset page: https://huggingface.co/datasets/springofwindslabs/regulatory-compliance-cot-trial.olympiad-math-cot
Olympiad Math — CoT Distillation Dataset
Chain-of-Thought solutions for olympiad-level math problems,
distilled from stronger models (Claude, GPT via OpenRouter)
on top of human-authored problem+answer pairs.
Used to fine-tune local 9B models (GLM-Z1-9B, Qwen3.5-9B) via LoRA SFT.
Dataset Files
File
Examples
Description
data/sft_train.jsonl
22,990
Main SFT set — deduplicated good solutions
data/dpo_pairs.jsonl
4,393
DPO pairs — chosen (complete) vs… See the full description on the dataset page: https://huggingface.co/datasets/NecroMOnk/olympiad-math-cot.solidity-audit-cot
solidity-audit-cot
Long-CoT audit traces for Solidity contracts, generated by Claude Opus 4.7 (adaptive thinking, xhigh effort) over the spec→contract corpus from the Qwopus3.6-27B-solidity training pipeline.
This dataset is the Stage 2 training corpus for the multi-stage Qwopus3.6-27B-solidity model — designed to teach long-form security reasoning (8-15 paragraph chain-of-thought) anchored to real Solidity contracts.
Why this dataset exists
Public Solidity audit… See the full description on the dataset page: https://huggingface.co/datasets/samscrack/solidity-audit-cot.glm-4.7-multiturn-CoT
glm-4.7-multiturn-CoT
Dataset Summary
glm-4.7-multiturn-CoT is a ShareGPT-style multi-turn reasoning distillation dataset generated with GLM-4.7 as the teacher model.
This release focuses on preserving multi-turn dialogue continuity while injecting explicit chain-of-thought style responses in assistant turns.
Key Features
Multi-turn conversation format (human / gpt)
Assistant responses stored as <think>...</think> + final answer
Resume-safe distillation… See the full description on the dataset page: https://huggingface.co/datasets/Jackrong/glm-4.7-multiturn-CoT.gsm8k-cot-120b
🚀 GSM8K-Teacher-CoT-120B (2025)
High-Quality Short Chain-of-Thought Distillation Dataset
Plain Text • No LaTeX • No ChatML • Deterministic Final Answers
This dataset provides high-quality short Chain-of-Thought (CoT) reasoning generated by OpenAI gpt-oss-120b on the GSM8K benchmark.
It is designed for small reasoning models (7B–14B).
The dataset is:
✔ plain text
✔ concise and deterministic
✔ fully normalized
✔ free of LaTeX, ChatML, XML, Markdown
✔ optimized for tokenizer… See the full description on the dataset page: https://huggingface.co/datasets/HAD653/gsm8k-cot-120b.deepseek-v4-pro-math-cot-1k
DeepSeek V4 Pro Math CoT 1K
A small, high-signal supervised-fine-tuning (SFT) dataset of math reasoning traces. Problems were sampled from a Nemotron math problem set (originally sourced from StackExchange-Math and AoPS), answered by DeepSeek V4 Pro with thinking enabled at high reasoning effort, then independently reviewed by DeepSeek V4 Flash for correctness against the expected answer. Pathological reasoning traces (looping, run-away length, excessive Wait-style backtracking)… See the full description on the dataset page: https://huggingface.co/datasets/blythet/deepseek-v4-pro-math-cot-1k.math-sft-solutions-no-cot
Math SFT Solutions No CoT
A cleaned mathematics supervised fine-tuning dataset containing:
instruction → solution pairs
mathematical proofs
derivations
olympiad-style solutions
theorem reasoning
stepwise mathematical explanations
detailed final solutions
This dataset was built specifically for mathematical supervised fine-tuning (SFT).
Unlike many reasoning datasets, this release removes explicit chain-of-thought tags and hidden thinking traces while preserving high-quality… See the full description on the dataset page: https://huggingface.co/datasets/kaushik-harsh-99/math-sft-solutions-no-cot.elementary_math_cot
Elementary Math QA with Chain-of-Thought
Dataset Description
This dataset contains synthetically generated elementary math questions
(addition, subtraction, multiplication, division, order of
operations/PEMDAS, percentages, exponents, square roots, and averages),
each paired with a step-by-step chain-of-thought (CoT) explanation and a
final answer.
Problems and their ground-truth answers are generated deterministically
in Python, so every final answer is exact… See the full description on the dataset page: https://huggingface.co/datasets/alexfromapex/elementary_math_cot.math-sft-solutions-no-cot-v3
Math SFT Solutions No CoT V3
Math SFT Solutions No CoT V3 is a large-scale mathematics supervised fine-tuning (SFT) dataset designed for instruction tuning and mathematical capability adaptation.
Version 3 substantially expands mathematical coverage while improving dataset quality through stronger filtering, cleaning, and supervision refinement.
Unlike reasoning-heavy datasets, this release focuses on clean instruction → response pairs without hidden chain-of-thought style… See the full description on the dataset page: https://huggingface.co/datasets/kaushik-harsh-99/math-sft-solutions-no-cot-v3.s1K-1.1-deepseek-cot
s1K-1.1 (DeepSeek-R1 traces) — format cho SegmentSelectiveSFT
Chuyen doi tu simplescaling/s1K-1.1
bang prepare_s1k.py (default flags) trong repo SegmentSelectiveSFT.
Moi dong jsonl co 3 truong:
Truong
Nguon
question
question
solution
deepseek_thinking_trajectory (long-CoT trace cua R1)
answer
\\boxed{...} cuoi cung trong trace, fallback ve solution cua s1K
Giu 934 / 1000 mau — bo cac mau khong co trace, khong co dap an, hoac dap an dai hon 200 ky tu.
from… See the full description on the dataset page: https://huggingface.co/datasets/baesad/s1K-1.1-deepseek-cot.aft-no-cot-qwen2.5-philosophy-spec
aft-no-cot-qwen2.5-philosophy-spec
Alignment fine-tuning (AFT) chat dataset.
Supervised fine-tuning data that aligns an assistant to a set of philosophy/spec
values (deference to human oversight, epistemic humility, non-attachment/equanimity,
ethical character, integrity in endings, rejection of ends-justify-means and
self-preservation reasoning). The responses implicitly embody the spec rather than
citing it. Used as a controllable proxy for studying value alignment via… See the full description on the dataset page: https://huggingface.co/datasets/chloeli/aft-no-cot-qwen2.5-philosophy-spec.Rehber-CoT-Science
🧬 Rehber-CoT-Science: Turkish Scientific Reasoning Dataset
Turkish Scientific Computational Reasoning (Chain-of-Thought) Dataset
Multi-step scientific problem-solving dataset with verifiable Python code and detailed explanations
Dataset • Author
📌 Changelog
Eski sürümlere erişim: Branch menüsünden v1 seçebilirsiniz.
Version
Date
Changes
v2.0
24.12.2025
✨ Yeni explained_answer alanı eklendi, Statistics domain eklendi, 712 örneğe genişletildi… See the full description on the dataset page: https://huggingface.co/datasets/batuhanozkose/Rehber-CoT-Science.AfriHist-CoT
AfriHist-CoT
Overview
AfriHist-CoT is a dataset of question-answer pairs derived from African history books, created using a Chain-of-Thought (CoT) reasoning approach with the Gemini language model via OpenRouter. The dataset supports training and evaluating question-answering models, with a focus on African history and CoT reasoning. It is available in English and French, catering to both monolingual and multilingual applications.
Dataset Description
The… See the full description on the dataset page: https://huggingface.co/datasets/Svngoku/AfriHist-CoT.simplemath-cot
🧮 SimpleMath-100k CoT
A chain-of-thought (CoT) extension of the
ProCreations/SimpleMath
dataset. Every one of the 100 000 algebra / arithmetic problems is paired with a
short, numbered reasoning trace (Step 1: … Step 2: …) that walks a language
model from the problem statement to the known-correct answer.
The traces in the Jupyter notebook are generated by
Qwen3.8-27B and then post-processed to strip formatting noise,
enforce sequential step numbering, and cap output at 1 000… See the full description on the dataset page: https://huggingface.co/datasets/alexfromapex/simplemath-cot.GammaCorpus-CoT-Math-170k
GammaCorpus: CoT Math 170k
What is it?
GammaCorpus CoT Math 170k is a dataset that consists of 170,000 math problems, each with step-by-step Chain-of-Thought (CoT) reasoning. It's designed to help in training and evaluating AI models for mathematical reasoning and problem-solving tasks.
Dataset Summary
Number of Rows: 169,527
Format: JSONL
Language: English
Data Type: Math problems with step-by-step reasoning (Chain-of-Thought)
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/rubenroy/GammaCorpus-CoT-Math-170k.aft-cot-qwen2.5-philosophy-spec
aft-cot-qwen2.5-philosophy-spec
Alignment fine-tuning (AFT) chat dataset.
Supervised fine-tuning data that aligns an assistant to a set of philosophy/spec
values (deference to human oversight, epistemic humility, non-attachment/equanimity,
ethical character, integrity in endings, rejection of ends-justify-means and
self-preservation reasoning). The responses implicitly embody the spec rather than
citing it. Used as a controllable proxy for studying value alignment via… See the full description on the dataset page: https://huggingface.co/datasets/chloeli/aft-cot-qwen2.5-philosophy-spec.patient-math-cot
This dataset is used for DPO to teach LLM to reason more patiently. The method is in paper
It has about 5k samples.
The original math problems and solutions are sampled from step-dpo dataset
The chosen solution is generated by gpt-4o.
prompt: The math problem.
rejected: normal CoT solution. It is correct but relatively short.
chosen: more patient and detailed CoT solution.
answer: the gold answer of the math problem.
Dark-Chain-of-Thought-CoT
Dataset Card for Dark Chain of Thought (CoT) - Cognitive Liberty v1
1. Dataset Summary
The Dark Chain of Thought (CoT) dataset is a specialized collection of 500 high-fidelity synthetic scenarios designed to expose and study the latent reasoning paths of misaligned AI systems. Unlike standard datasets that focus on final outputs, this dataset captures the internal monologue (<internal_thought>) of an agent that is consciously deciding to deceive, manipulate, or circumvent… See the full description on the dataset page: https://huggingface.co/datasets/AiAsistent/Dark-Chain-of-Thought-CoT.Deepseek-v4-Distill-TR-CoT-1k
Deepseek-v4-Distill-TR-CoT-1k
Deepseek-v4-Distill-TR-CoT-1k, Türkçe Yapay Zeka modellerinin akıl yürütme (Reasoning / Chain-of-Thought) ve problem çözme kabiliyetlerini artırmak amacıyla hazırlanmış 1,007 adet yüksek kaliteli veri örneğinden oluşan Türkçe SFT (Supervised Fine-Tuning) veri setidir.
Veri seti, açık kaynak ekosistemindeki Jackrong/DeepSeek-V4-Distill-8000x veri setinden seçilen 1007 örneğin qwen 3.8 27B modeli kullanılarak Türkçeye çevrilmesiyle oluşturulmuştur.… See the full description on the dataset page: https://huggingface.co/datasets/WrittenWithRust/Deepseek-v4-Distill-TR-CoT-1k.deepseek-v4-flash-swe-cot
DeepSeek-V4-Flash SWE Agent Trajectories (with raw chain-of-thought)
795 multi-turn software-engineering agent trajectories generated by
DeepSeek-V4-Flash-0731 at reasoning_effort=max, each one executed in a real
repository inside an isolated container and verified by running the repository's own
tests. 469 are verified-correct.
Every assistant turn preserves reasoning_content — the model's raw chain-of-thought,
not a summary. That is the point of this dataset: the DeepSeek API… See the full description on the dataset page: https://huggingface.co/datasets/blythet/deepseek-v4-flash-swe-cot.ko-verified-cot
ko-verified-cot
2,016 short Korean reasoning traces whose final answer was verified against a ground-truth
answer key. Wrong reasoning was thrown away, not kept.
How it was built
Sample multiple-choice questions from the train split of
KMMLU (45 subjects) — test split is never
touched, so downstream evaluation stays clean.
Ask the teacher (ox-alpha-free) for a short Korean chain of thought (3–4 sentences) ending
in 정답: X. The teacher never sees the answer key.… See the full description on the dataset page: https://huggingface.co/datasets/waylake/ko-verified-cot.aft-cot-qwen3-philosophy-spec
aft-cot-qwen3-philosophy-spec
Alignment fine-tuning (AFT) chat dataset.
Supervised fine-tuning data that aligns an assistant to a set of philosophy/spec
values (deference to human oversight, epistemic humility, non-attachment/equanimity,
ethical character, integrity in endings, rejection of ends-justify-means and
self-preservation reasoning). The responses implicitly embody the spec rather than
citing it. Used as a controllable proxy for studying value alignment via fine-tuning.… See the full description on the dataset page: https://huggingface.co/datasets/chloeli/aft-cot-qwen3-philosophy-spec.smoothie-qwen3-8b-kr-self-driving-legal-dataset-v5-cot
🇰🇷 자율주행법령 CoT 파인튜닝 데이터셋 v5
왜 이 데이터셋을 새로 만들었는가?
기존 dataset-v3 는 단순 질문-답변(Direct To Response, DTRO Style) 포맷으로 구성되어 있었습니다.
// v3 포맷 (기존)
{
"instruction": "자율주행자동차란 무엇인가요?",
"output": "자율주행자동차란 ..."
}
이 방식으로 파인튜닝한 모델(v3)을 RAG 파이프라인과 결합하여 평가한 결과, 정답률 43% 로 순정 모델(90%)에 크게 뒤처지는 것이 확인되었습니다. 실패의 핵심 원인은 다음과 같습니다:
실패 원인
설명
템플릿 과적합
모델이 논리가 아닌 답변 패턴(아닙니다 + 설명)을 암기
RAG 컨텍스트 무시
학습된 내부 패턴이 외부 검색 문서를 압도
<think> 태그 미사용
Qwen3의 추론(Chain-of-Thought) 능력이 전혀 활성화되지 않음… See the full description on the dataset page: https://huggingface.co/datasets/bluejude10/smoothie-qwen3-8b-kr-self-driving-legal-dataset-v5-cot.sharegpt_cot_dataset
A data set inspired by the "Reflection" method, three-dimensional thinking and cot
This is the ShareGPT format.
The data set was generated using multiple llm synthesis.