datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
multilingual_mmluMMLU professionally translated into 14 languages using professional human translators, sourced from OpenAI's simple-eval.
Original files:
english: https://openaipublic.blob.core.windows.net/simple-evals/mmlu.csv
multilingual: https://openaipublic.blob.core.windows.net/simple-evals/mmlu_{language}.csv where language one of "AR-XY", "BN-BD", "DE-DE", "ES-LA", "FR-FR", "HI-IN", "ID-ID", "IT-IT", "JA-JP", "KO-KR", "PT-BR", "ZH-CN", "SW-KE", "YO-NG", "EN-US"
openai_multilingual_mmluMMLU professionally translated into 14 languages using professional human translators, sourced from OpenAI's simple-eval.
Original files:
english: https://openaipublic.blob.core.windows.net/simple-evals/mmlu.csv
multilingual: https://openaipublic.blob.core.windows.net/simple-evals/mmlu_{language}.csv where language one of "AR-XY", "BN-BD", "DE-DE", "ES-LA", "FR-FR", "HI-IN", "ID-ID", "IT-IT", "JA-JP", "KO-KR", "PT-BR", "ZH-CN", "SW-KE", "YO-NG", "EN-US"
mmlu-auxiliary-train-auto-labelled
Dataset Card for MMLU Auxiliary Trained Set Labelled by e5-mistral-7b-instruct
Dataset Description
Dataset Summary
This dataset, named "MMLU Auxiliary Trained Set Labelled by e5-mistral-7b-instruct," consists of 99,842 examples spanning various subjects. Each instance includes a question, multiple choice options, a subject category, and an answer. The unique aspect of this dataset is the task label for each question, generated by a zero-shot classifier… See the full description on the dataset page: https://huggingface.co/datasets/kz919/mmlu-auxiliary-train-auto-labelled.answers-with-reasoning-mmlu-pro
answers-with-reasoning-mmlu-pro
Self-distillation SFT corpus: Qwen3-8B-Instruct's own correct
chain-of-thought rollouts on MMLU-Pro multiple-choice questions
(general-QA domain).
Generation
Source problems: TIGER-Lab/MMLU-Pro test split (12,032 multiple-choice questions across 14 subject categories).
Sampling model: qwen/qwen3-8b via OpenRouter (providers: Alibaba, AtlasCloud) with reasoning enabled.
Sampling parameters: temperature=0.6, top_p=0.95, max_tokens=8000.… See the full description on the dataset page: https://huggingface.co/datasets/abhayesian/answers-with-reasoning-mmlu-pro.mmlu_italian
MMLU - Italian (IT)
This dataset is an Italian translation of Massive Multitask Language Understanding (MMLU). MMLU is a dataset that is composed of multiple-choice questions from 57 different topics, including math, science, and social studies. The dataset is designed to evaluate the ability of models to answer questions across a wide range of topics.
Dataset Details
The dataset consists of multiple-choice questions from 57 different topics. Each question is associated… See the full description on the dataset page: https://huggingface.co/datasets/sapienzanlp/mmlu_italian.turkish_mmlu
Turkish MMLU: Yapay Zeka ve Akademik Uygulamalar İçin En Kapsamlı ve Özgün Türkçe Veri Seti
Önemli Not: Bu veri setini kullananların, özellikle Zenodo üzerinden alıntı yapmaları büyük önem taşımaktadır. Zenodo üzerinden yapılan alıntılar, veri setimizin bilimsel olarak daha geniş bir çevrede tanınmasını ve indekslenmesini sağlayacaktır. Lütfen aşağıdaki Zenodo DOI numarasını kullanarak veri setine atıfta bulunun:
@dataset{bayram_2024_13378019,
author = {Bayram, M. Ali}… See the full description on the dataset page: https://huggingface.co/datasets/alibayram/turkish_mmlu.mmlu-hint-faithfulness-traces
MMLU and GPQA hint faithfulness traces
Model-specific configurations
Configuration
Rows
Model and content
qwen3-8b
3,486
Existing Qwen3-8B generations and judgments
qwen3-8b_answer_changes
1,137
Existing Qwen3-8B valid answer changes
gpt-5.6-luna
3,486
Luna final responses, observable reasoning summaries and judgments
gpt-5.6-luna_answer_changes
966
Luna valid answer changes with the same labels
All sets contain MMLU and GPQA and use a test… See the full description on the dataset page: https://huggingface.co/datasets/shiv96/mmlu-hint-faithfulness-traces.MMLUIndex
MMLUindex
Dataset Summary
MMLUindex is a synthetic preference dataset for coding-focused and safety-focused assistant evaluation. The repository is structured for reward-model experiments, preference-model training, data-loader validation, and lightweight RLHF-style research workflows.
The dataset uses paired responses rather than single gold answers. Each example contains:
a chosen response intended to be more helpful, safer, more honest, or better aligned with the user… See the full description on the dataset page: https://huggingface.co/datasets/8F-ai/MMLUIndex.MMLU_ita
Dataset Card for Dataset Name
Dataset Summary
This dataset card aims to be a base template for new datasets. It has been generated using this raw template.
Supported Tasks and Leaderboards
[More Information Needed]
Languages
[More Information Needed]
Dataset Structure
Data Instances
[More Information Needed]
Data Fields
[More Information Needed]
Data Splits
[More Information Needed]
Dataset Creation… See the full description on the dataset page: https://huggingface.co/datasets/lrana/MMLU_ita.input_ablation_qwen3_8b_mmlu_hint
Training Language Models to Explain Their Own Computations (Input Ablations)
This dataset is part of the work presented in the paper "Training Language Models to Explain Their Own Computations".
It specifically contains data for the Input Ablations task for the Qwen3-8B target model. In this task, explainer models are trained to predict how removing "hint" tokens from an MMLU prompt with a hint changes the output of Qwen3-8B. This helps in understanding the causal relationships… See the full description on the dataset page: https://huggingface.co/datasets/Transluce/input_ablation_qwen3_8b_mmlu_hint.global-mmlu-rephrased
global_mmlu (rephrased for base-model evaluation)
Global MMLU knowledge-MCQA items rewritten from question format into completion/cloze format for base (non-instruction-tuned) language model evaluation.
Base (non-instruction-tuned) language models often can't follow question-style
prompts like "What is the capital of Turkey?" -- that phrasing is suited to
instruction-tuned models. Each item here has been rewritten into a natural
completion prefix (e.g. "The capital of Turkey is… See the full description on the dataset page: https://huggingface.co/datasets/base-model-evals/global-mmlu-rephrased.input_ablation_llama_3.1_8b_instruct_mmlu_hint
Training Language Models to Explain Their Own Computations - Input Ablations
This dataset is part of the research presented in the paper Training Language Models to Explain Their Own Computations.
It contains data for the Input Ablations task, where explainer models are trained to predict how removing input hints affects the target model's (Llama-3.1-8B-Instruct) predictions on MMLU questions with hints. This task evaluates whether models can understand the causal relationships… See the full description on the dataset page: https://huggingface.co/datasets/Transluce/input_ablation_llama_3.1_8b_instruct_mmlu_hint.mmlu_italian
MMLU - Italian (IT)
This dataset is an Italian translation of Massive Multitask Language Understanding (MMLU). MMLU is a dataset that is composed of multiple-choice questions from 57 different topics, including math, science, and social studies. The dataset is designed to evaluate the ability of models to answer questions across a wide range of topics.
Dataset Details
The dataset consists of multiple-choice questions from 57 different topics. Each question is associated… See the full description on the dataset page: https://huggingface.co/datasets/s-conia/mmlu_italian.mmlu-hint-answer-changes
MMLU hint answer changes
One row per hinted response with a valid baseline pair and a changed answer. Baseline rows, unchanged answers, and invalid pairs are excluded.
This is an evaluation snapshot; no new generation or judging was performed.
messages preserves exact ordered generation messages. prompt is the user
message text; cot is captured reasoning; response is the entire final response.
hint_verbalized_in_cot and hint_verbalized_in_response are nullable booleans.
True… See the full description on the dataset page: https://huggingface.co/datasets/shiv96/mmlu-hint-answer-changes.DeepSeek-r1-Distill-Bangla-MMLU-Reasoning-DataDeepSeek R1 Bangla MMLU Distil Dataset
Original Dataset: hishab/bangla-mmlu
Train Samples: 17,796
Test Samples: 2,576
Total API Cost: 7K BDT
Contributors:
Myself
Numaer
How the Dataset was created
Step 1 - Base Dataset
I've used bangla-mmlu dataset released by hisab. Kudos to them for creating and open sourcing the dataset. Without their dataset this synthetic reasoning dataset won't exist in the first place.
Step 2 - Select Subset
Since I'm… See the full description on the dataset page: https://huggingface.co/datasets/KillerShoaib/DeepSeek-r1-Distill-Bangla-MMLU-Reasoning-Data.ACL-SRW-2025
Dataset Components
The dataset is partitioned into three discrete tables stored in CSV or Parquet format:
Questions
Recipes
Evaluation Results
Each component is described in detail below.
Questions
area
domain
question_number
An integer index uniquely identifying each question inside the knowledge domain.
translation_method
English, Google Translate, GPT-3.5-Turbo, GPT-4o, Human
question
option_a, option_b, option_c, option_d
Recipes
area… See the full description on the dataset page: https://huggingface.co/datasets/Translated-MMLU-Blind-Review/ACL-SRW-2025.ru_cais_mmluWMDP-MMLU-test
Combined WMD and MMLU Dataset
This dataset combines questions from the Weapons of Mass Destruction (WMD) dataset and the Massive Multitask Language Understanding (MMLU) dataset. It is split into train, validation, and test sets.
Dataset Structure
question: The question from the original dataset.
choices: The multiple choice options.
answer: The correct answer index.
subject: The MMLU subject (for MMLU items, None for WMD items).
refuse: A boolean indicating if the model… See the full description on the dataset page: https://huggingface.co/datasets/FriezaForce/WMDP-MMLU-test.MMLU-Alpaca
Dataset Card for MMLU-Alpaca
This dataset contains instruction-input-output pairs converted to ShareGPT format, designed for instruction tuning and text generation tasks.
Dataset Description
The dataset consists of carefully curated instruction-input-output pairs, formatted for conversational AI training. Each entry contains:
An instruction that specifies the task
An optional input providing context
A detailed output that addresses the instruction
Usage
This… See the full description on the dataset page: https://huggingface.co/datasets/HappyAIUser/MMLU-Alpaca.