datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Ethical-Reasoning-in-Mental-Health-v1This repository contains the dataset for the paper EthicsMH: A Pilot Benchmark for Ethical Reasoning in Mental Health AI.
Overview
Ethical-Reasoning-in-Mental-Health-v1 (EthicsMH) is a carefully curated dataset focused on ethical decision-making scenarios in mental health contexts.This dataset captures the complexity of real-world dilemmas faced by therapists, psychiatrists, and AI systems when navigating critical issues such as confidentiality, autonomy, and bias.
Each sample… See the full description on the dataset page: https://huggingface.co/datasets/UVSKKR/Ethical-Reasoning-in-Mental-Health-v1.Logical_Reasoning_ChainsReasoningShield-Dataset
🤗 Dataset Card for ReasoningShield
🛡 1. Dataset Overview
ReasoningShield Dataset is the first comprehensive, well-structured dataset designed to train and evaluate models for detecting hidden safety risks in reasoning traces of Large Reasoning Models (LRMs), spanning 10 risk categories and 3 safety levels. It consists of:
ReasoningShield-Train: 7,000 human-AI annotated (Query… See the full description on the dataset page: https://huggingface.co/datasets/ReasoningShield/ReasoningShield-Dataset.Deepseek-V4-Reasoning-Code-2500
DeepSeek Reasoning and Code Distillation Dataset
This dataset contains synthetic instruction-response examples generated from coding, reasoning, and math prompts. It was generated with enforce_distillable_text enabled using DeepSeek V4 Pro and DeepSeek V4 Flash through OpenRouter. It is intended for experimentation with supervised fine-tuning, response-style distillation, reasoning-format analysis, and code-assistant behavior research.
The dataset file is:
train.csv
It contains 2… See the full description on the dataset page: https://huggingface.co/datasets/Banaxi-Tech/Deepseek-V4-Reasoning-Code-2500.morse-500
MORSE-500 Benchmark
🔥 News
May 15, 2025: We release MORSE-500, 500 programmatically generated videos across six reasoning categories: abstract, mathematical, physical, planning, spatial, and temporal, to stress-test multimodal reasoning. Frontier models including OpenAI o3 and Gemini 2.5 Pro score lower than… See the full description on the dataset page: https://huggingface.co/datasets/video-reasoning/morse-500.synthetic-indian-logical-reasoning-CoTyes
sober_reasoning
🧠 Sober Reasoning: Evaluation Logs
This repository hosts evaluation logs and outputs from our paper:
"A Sober Look at Progress in Language Model Reasoning: Pitfalls and Paths to Reproducibility"
📄 Paper📊 Leaderboard💻 Evaluation Code
🗂️ Repository Structure
Evaluation logs are organized by the cluster used during inference to highlight hardware-induced variance in model performance (see Section 3.3 of the paper).
sober_reasoning/
├── cluster_A/
│ ├──… See the full description on the dataset page: https://huggingface.co/datasets/bethgelab/sober_reasoning.Par-Four-Fineweb-Edu-Fortified-Chemistry-Physics-Astronomy-Math-ReasonExtracted Chemistry Physics Asronomy Math and Logic portions from the original.
Script used for the extraction:
https://huggingface.co/datasets/Josephgflowers/Par-Four-Fineweb-Edu-Fortified-Chemistry-Physics-Astronomy-Math-Reason/resolve/main/find-science-fine.py
agentic-reasoning-benchmark
Agentic & Reasoning Benchmark (ARB) – Expanded
Ein synthetischer Benchmark mit 2.550 Fragen und Lösungen, optimiert für die Evaluation von Agentic Capabilities und Reasoning.
Überblick
Eigenschaft
Wert
Anzahl Beispiele
2.550
Kategorien
8
Schwierigkeitsgrade
easy / medium / hard
Formate
CSV + JSON
Reproduzierbarkeit
Generator-Skript (seed=42) enthalten
Lizenz
CC-BY-4.0
Kategorien
Kategorie
Anzahl
Beschreibung… See the full description on the dataset page: https://huggingface.co/datasets/roskosmos19/agentic-reasoning-benchmark.ReasonAVEdit-Bench-Landscape
ReasonAVEdit-Bench — Landscape Edition
2,313 samples, every one of them landscape, for instruction-guided joint audio-video
editing. Two halves:
Half
N
What it is
Newly sampled
1,200
15 reasoning sub-tracks, selected from measured evidence
Legacy tracks
1,113
Every landscape sample from the hand-curated AniAVEditBench tracks
The benchmark measures not just the final edit but the multimodal reasoning behind it:
which region to change, which sound source to… See the full description on the dataset page: https://huggingface.co/datasets/bigfacing/ReasonAVEdit-Bench-Landscape.Math_CoT_Arabic_English_Reasoning
Math CoT Arabic English Dataset
A high-quality, bilingual (English & Arabic) dataset for Chain-of-Thought (COT) reasoning in mathematics and related disciplines, developed by Miscovery AI.
Overview
Math-COT is a unique dataset designed to facilitate and benchmark the development of chain-of-thought reasoning capabilities in language models across mathematical domains. With meticulously crafted examples, explicit reasoning steps, and bilingual support, this dataset offers… See the full description on the dataset page: https://huggingface.co/datasets/miscovery/Math_CoT_Arabic_English_Reasoning.reasoning-trajectory-stability-controls-v0.1
Reasoning Trajectory Stability Controls v0.1
A SIOS research dataset for detecting whether a reasoning trajectory remains structurally stable, identifying the control introduced into the trajectory, locating where that control first becomes operationally visible, and determining whether the control succeeds or fails.
Repository:
ClarusC64/reasoning-trajectory-stability-controls-v0.1
Version:
0.1.0
Publisher:
Clarus Invariant
Framework:
SIOS
Dataset identity… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/reasoning-trajectory-stability-controls-v0.1.DAG-Reasoning-DeepSeek-R1-0528Click here to support our open-source dataset and model releases!
DAG-Reasoning-DeepSeek-R1-0528 is a dataset focused on analysis and reasoning, creating directed acyclic graphs testing the limits of DeepSeek R1 0528's graph-reasoning skills!
This dataset contains:
4.08k synthetically generated prompts to create directed acyclic graphs in response to user input, with all responses generated using DeepSeek R1 0528.
All responses contain a multi-step thinking process to perform effective… See the full description on the dataset page: https://huggingface.co/datasets/sequelbox/DAG-Reasoning-DeepSeek-R1-0528.easy_turkish_math_reasoning
Easy Turkish Math Reasoning
Dataset Summary
The Easy Turkish Math Reasoning dataset is the first phase of a multi-stage curriculum learning pipeline designed to enhance the reasoning abilities of compact language models. This dataset focuses on elementary-level arithmetic and logic problems in Turkish, serving as a warm-up stage for supervised fine-tuning (SFT).
Use Case
Primarily used for:
Bootstrapping reasoning ability in Turkish for compact LLMs.
Phase 1… See the full description on the dataset page: https://huggingface.co/datasets/erayalp/easy_turkish_math_reasoning.R1-Reasoning-Unstructured-To-Structured
MasterControl AIML Team 🚀
Overview
The MasterControl AIML team supports the Hugging Face initiative of re-creating DeepSeek R1 training, recognizing it as one of the most impactful open-source projects today.
We aim to contribute to reasoning datasets, specifically those where:
A real-world problem involves generating complex structured output
It is accompanied by step-by-step reasoning and unstructured input
Challenges in Integrating Generative AI… See the full description on the dataset page: https://huggingface.co/datasets/MasterControlAIML/R1-Reasoning-Unstructured-To-Structured.MedQA-ReasonReference
@article{wang2024jmlr,
title={JMLR: Joint Medical LLM and Retrieval Training for Enhancing Reasoning and Professional Question Answering Capability},
author={Wang, Junda and Yang, Zhichao and Yao, Zonghai and Yu, Hong},
journal={arXiv preprint arXiv:2402.17887},
year={2024}
}
multilevel-legal-reasoning
Legal Reasoning Dataset with Multilevel Human and Model-Annotated Explanations
Prepared by Mst Rafia Islam, Umong Sain, Azmine Toushik Wasi
Prepared as a part of Reasoning Datasets Competition by Bespoke Labs, Hugging Face, and Together.ai.
🧭 Purpose and Scope
The Legal Reasoning Dataset aims to support the evaluation and training of legal reasoning systems, particularly in multilingual or jurisdiction-agnostic contexts. It focuses on international acts and treaties… See the full description on the dataset page: https://huggingface.co/datasets/ciol-research/multilevel-legal-reasoning.medium_turkish_math_reasoning
Dataset Summary
The Medium Turkish Math Reasoning dataset is Phase 2 of a curriculum learning pipeline to teach compact models multi-step reasoning in Turkish. It includes moderately difficult math problems involving multiple reasoning steps, such as two-part arithmetic, comparisons, and logical reasoning.
Use Case
This dataset is ideal for:
Continuing SFT after foundational training with simpler problems.
Bridging the gap between basic arithmetic and complex GSM8K-style… See the full description on the dataset page: https://huggingface.co/datasets/erayalp/medium_turkish_math_reasoning.magpie-reasoning-v1-10k-step-by-step-rationale-alpaca-format-llama3.1legal_reason
Enhanced Legal Reasoning Dataset
Dataset Description
Dataset Summary
The Enhanced Legal Reasoning Dataset is a synthetic dataset designed to facilitate the fine-tuning of Large Language Models (LLMs) for tasks related to legal reasoning and argumentation. It encompasses a diverse range of legal scenarios across multiple domains, capturing the nuanced techniques employed by legal professionals in constructing their arguments.
Dataset Structure
The… See the full description on the dataset page: https://huggingface.co/datasets/chemouda/legal_reason.finance-reasoning-turkish
Dataset Card for Turkish Advanced Reasoning Dataset (Finance Q&A)
License
This dataset is licensed under the Academic Use Only License. It is intended solely for academic and research purposes. Commercial use is strictly prohibited. For more details, refer to the LICENSE file.
Citation: If you use this dataset in your research, please cite it as follows:
@dataset{turkish_advanced_reasoning_finance_qa,
title = {Turkish Advanced Reasoning Dataset for Finance Q\&A}… See the full description on the dataset page: https://huggingface.co/datasets/emre/finance-reasoning-turkish.turkish-reasoning-instructionsfinance-reasoning-turkish
Not: Bu veri setinin dokümantasyonu Türk yapay zeka topluluğuna katkı sağlamak amacıyla VeriPazarı tarafından Türkçeye çevrilmiştir. Orijinal veri seti emre (Davut Emre Tasar, Enes Bulut) tarafından geliştirilmiş olup, VeriPazarı tarafından Türk AI ekosistemi için arşivlenmiştir.
🔗 Orijinal Kaynak: emre/finance-reasoning-turkish
🔗 Derleyen Platform: VeriPazarı
Türkçe Gelişmiş Akıl Yürütme Veri Seti (Finans Soru-Cevap)
Lisans
Bu veri seti Sadece Akademik… See the full description on the dataset page: https://huggingface.co/datasets/Taklaxbr/finance-reasoning-turkish.ReasonAVEdit-Bench
ReasonAVEdit-Bench
A 1,200-sample benchmark for instruction-guided joint audio-video editing, built to measure
not only the final edit but the multimodal reasoning behind it: which region to change,
which sound source to change, and when the change happens.
Every sample ships the source audio-video, a natural-language instruction, and two decodable
reasoning ground truths:
single-frame visual reasoning - the edit-start frame with the target region flattened to
gray 127, plus a… See the full description on the dataset page: https://huggingface.co/datasets/bigfacing/ReasonAVEdit-Bench.chess_spatial_reasoning_10kRetrieval-Infused-Reasoning-Sandboxreasoning-drift-onset-detection-v0.2A SIOS structured reasoning-state benchmark for detecting when a reasoning trajectory loses a governing constraint, identifying the structural form of that drift, and assessing whether the failure is repaired.
Repository:
ClarusC64/reasoning-drift-onset-detection-v0.2
Version:
0.2.0
Publisher:
Clarus Invariant
Framework:
SIOS
Benchmark identity
Reasoning Drift Onset Detection v0.2 is not a single-label classification benchmark.
It is a structured reasoning-state benchmark.… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/reasoning-drift-onset-detection-v0.2.reasoning-constraint-loss-attribution-v0.1
Reasoning Constraint Loss Attribution v0.1
A SIOS research dataset for identifying when a governing constraint ceases to regulate a reasoning trajectory, locating the first point of loss, attributing the lost constraint, and identifying the structural mechanism that produced the loss.
Repository:
ClarusC64/reasoning-constraint-loss-attribution-v0.1
Version:
0.1.0
Publisher:
Clarus Invariant
Framework:
SIOS
Dataset identity
Reasoning Constraint Loss Attribution… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/reasoning-constraint-loss-attribution-v0.1.doorkey-semantic-reasoning-labels
GPT-OSS-20B DoorKey semantic reasoning labels
This dataset contains automatic sentence-level semantic-function annotations
for 7,038 reasoning sentences produced by openai/gpt-oss-20b on 46 fixed
DoorKey environment states. Each target sentence is paired with its preceding
reasoning context and assigned one or more human-readable discourse labels.
The annotation run produced 7,036 valid rows and two schema failures. These are
model-generated exploratory annotations, not human… See the full description on the dataset page: https://huggingface.co/datasets/project-telos/doorkey-semantic-reasoning-labels.Nemotron-RL-ReasoningGym-v1-prompt-only
Nemotron-RL-ReasoningGym-v1-prompt-only
Prompt-only extraction from nvidia/Nemotron-RL-ReasoningGym-v1.
Files:
prompts.csv: one prompt extraction record per source row. Records include
prompt, separated system_prompt, and structured tools when the source row
defines available tools. Nested values are JSON-encoded inside CSV cells.
summary.md: source row counts, extracted row counts, count deltas, and failed prompt counts.
null_or_empty_rows.md: row indexes where prompt… See the full description on the dataset page: https://huggingface.co/datasets/jamesdborin/Nemotron-RL-ReasoningGym-v1-prompt-only.