datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
fusion-synth-data-s1kx
Offline Synthetic Data (s1K-X) for: Making, not taking, the Best-of-N
Content
This data contains completions for the s1K-X training split prompts from 5 different teacher models and 2 aggregations:
Teachers: We sample one completion from each of the following models at temperature T=0.3. For kimik2, qwen3, and deepseek-v3 we use TogetherAI, for gemma3-27b and command-a we use locally hosted images.
gemma3-27b: GEMMA3-27B-IT
kimik2: KIMI-K2-INSTRUCT
qwen3: QWEN3-235B… See the full description on the dataset page: https://huggingface.co/datasets/CohereLabs/fusion-synth-data-s1kx.acc_rd_s1-gpqa
Dataset Card for GPQA
GPQA is a multiple-choice, Q&A dataset of very hard questions written and validated by experts in biology, physics, and chemistry. When attempting questions out of their own domain (e.g., a physicist answers a chemistry question), these experts get only 34% accuracy, despite spending >30m with full access to Google.
We request that you do not reveal examples from this dataset in plain text or images online, to reduce the risk of leakage into foundation model… See the full description on the dataset page: https://huggingface.co/datasets/stewy33/acc_rd_s1-gpqa.s1_translated
S1K Multilingual Translation Dataset
This dataset contains translations of the simplescaling/s1K dataset into 20 languages.
Languages Included
High-Resource Languages
Chinese (Simplified) - zh
Spanish - es
French - fr
German - de
Japanese - ja
Arabic - ar
Russian - ru
Portuguese - pt
Medium-Resource Languages
Korean - ko
Vietnamese - vi
Thai - th
Polish - pl
Dutch - nl
Turkish - tr
Low-Resource Languages
Swahili - sw
Bengali - bn
Urdu… See the full description on the dataset page: https://huggingface.co/datasets/JiayiHe/s1_translated.ANGO-S1ANGO is A Novel Generation-Oriented Chinese LLM evaluation benchmark.
We introduces the format of single-question multiple-keypoints dataset for the first time, which include 171 keypoints accumulated in 4 hierarchical levels and 9 difficulty categories.
The data were exclusively obtained from the Administrative Proficiency Test, which serves as a significant component of the Chinese civil service examination.
We will apply a seasonal system for the leaderboard, updating them every two months.… See the full description on the dataset page: https://huggingface.co/datasets/AngoHF/ANGO-S1.s1_59k
Dataset Card for XuHu6736/s1_59k
Dataset Description
XuHu6736/s1_59k is a dataset specifically prepared for Supervised Fine-Tuning (SFT) of large language models. It is constructed by merging and processing two existing Hugging Face datasets: simplescaling/data_ablation_full59K and qfq/train_featurized.
The simplescaling/data_ablation_full59K dataset is a collection of approximately 59,000 questions and solutions spanning various domains including mathematics, science… See the full description on the dataset page: https://huggingface.co/datasets/XuHu6736/s1_59k.s1K-X
s1K-X
We used Google Translate to translate simplescaling/s1K-1.1_tokenized
into 10 different languages used in MGSM
to evaluate whether supervised fine-tuning on translated texts can mitigate catastrophic forgetting in reasoning in the question language (Hwang et al., 2025)
The dataset includes below languages:
English (en) -- the original simplescaling/s1K-1.1_tokenized
Bengali (bn)
German (de)
Spanish (es)
French (fr)
Japanese (ja)
Russian (ru)
Swahili (sw)
Telugu (te)
Thai (th)… See the full description on the dataset page: https://huggingface.co/datasets/jdhwang/s1K-X.S1-BenchThe benchmark constructed in paper S1-Bench: A Simple Benchmark for Evaluating System 1 Thinking Capability of Large Reasoning Models.
Introduction
S1-Bench is a novel benchmark designed to evaluate Large Reasoning Models' performance on simple tasks that favor intuitive system 1 thinking rather than deliberative system 2 reasoning.
S1-Bench comprises 422 question-answer pairs across four major categories and 28 subcategories, balanced with 220 English and 202 Chinese questions.… See the full description on the dataset page: https://huggingface.co/datasets/WYRipple/S1-Bench.s1-prob
Citation Information
@misc{muennighoff2025s1simpletesttimescaling,
title={s1: Simple test-time scaling},
author={Niklas Muennighoff and Zitong Yang and Weijia Shi and Xiang Lisa Li and Li Fei-Fei and Hannaneh Hajishirzi and Luke Zettlemoyer and Percy Liang and Emmanuel Candès and Tatsunori Hashimoto},
year={2025},
eprint={2501.19393},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2501.19393},
}
s1K-1.1-Translated
s1K-1.1-Translated
This dataset contains translated versions of the s1K-1.1 dataset across multiple languages.
Languages
The dataset contains the following language subsets: Zh, Fr, Ja, Af, Th, Lv, Mr, Te, Sw, En
Translation Method
This dataset was created using Gemini 2.0 Flash for automatic translation.
Dataset Structure
Each language subset contains conversational data in the following format:
{
'conversations': [
{'from': 'human'… See the full description on the dataset page: https://huggingface.co/datasets/joshbarua/s1K-1.1-Translated.s1_54k_filter
Dataset Card for XuHu6736/s1_54k_filter
Dataset Description
XuHu6736/s1_54k_filter is a filtered version of the XuHu6736/s1_59k dataset. This dataset has been processed to remove records containing empty or null values in any field, with the specific exception of the 'cot' (Chain-of-Thought) column. If any other field in a record is empty, that entire record is discarded.
The original s1_59k dataset was prepared for Supervised Fine-Tuning (SFT) of large language models by… See the full description on the dataset page: https://huggingface.co/datasets/XuHu6736/s1_54k_filter.premchand-148-stories-dataset
Premchand 148 Stories Dataset
A curated Hindi NLP corpus built from 148 short stories by Munshi Premchand, processed into sentence-aware chunks for RAG, embedding, and research pipelines.
[!NOTE]
This work is for a PhD in University of Allahabad by Shikha Agrawal under the supervision of Dr. Pravin Kumar.
Dataset Structure
The repository contains the following data files optimized for direct usage in any retrieval or generation model:
corpus.jsonl: The final… See the full description on the dataset page: https://huggingface.co/datasets/s1hikha/premchand-148-stories-dataset.s1K-1.1-gpt-oss-20b
s1K-1.1-gpt-oss-20b
Dataset Summary
The s1K-1.1-gpt-oss-20b dataset extends the simplescaling/s1K-1.1 dataset by incorporating reasoning trajectories generated by the openai/gpt-oss-20b model. This dataset contains questions primarily from mathematical problem-solving domains, along with responses and reasoning trajectories generated by the gpt-oss-20b model. The dataset is designed to facilitate research into model reasoning capabilities, test-time scaling, and… See the full description on the dataset page: https://huggingface.co/datasets/IIGroup/s1K-1.1-gpt-oss-20b.s1_54k_filter_with_isreasoning
Dataset Card for XuHu6736/s1_54k_filter_with_isreasoning
Dataset Description
XuHu6736/s1_54k_filter_with_isreasoning is an enhanced version of the XuHu6736/s1_54k_filter dataset. This version includes additional annotations to assess the suitability of each question for reasoning training. These annotations, isreasoning_score and isreasoning, were generated using the deepseek-v3 model.
The purpose of these new fields is to allow users to filter, weight, or specifically… See the full description on the dataset page: https://huggingface.co/datasets/XuHu6736/s1_54k_filter_with_isreasoning.s1K-1.1-Distilled
s1K-1.1-Distilled
This dataset contains distilled versions of the s1K-1.1 dataset across multiple languages.
Languages
The dataset contains the following language subsets: Zh, Fr, Ja, Af, Th, Lv, Mr, Te, Sw, En
Generation Method
This dataset was created using DeepSeek-R1 for distillation of reasoning traces.
Dataset Structure
Each language subset contains conversational data in the following format:
{
'conversations': [
{'from': 'human'… See the full description on the dataset page: https://huggingface.co/datasets/joshbarua/s1K-1.1-Distilled.