datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Multi-SWE-bench
Multi-SWE-bench
Re-upload of ByteDance's Multi-SWE-bench
evaluation benchmark: 2,132 issue-resolving tasks across the seven Multi-SWE languages.
This is the held-out eval benchmark; for RL training data use
PrimeIntellect/Multi-SWE-RL-Verified.
Changes vs upstream
Storage schema only: per-test maps are stored as columnar struct-of-lists so the rows load
cleanly with datasets. Row content is unchanged.
License mirrors upstream: ByteDance licenses the dataset… See the full description on the dataset page: https://huggingface.co/datasets/PrimeIntellect/Multi-SWE-bench.Multilingual-Thinking
Dataset summary
Multilingual-Thinking is a reasoning dataset where the chain-of-thought has been translated from English into one of 4 languages: Spanish, French, Italian, and German. The dataset was created by sampling 1k training samples from the SystemChat subset of SmolTalk2 and translating the reasoning traces with another language model.
This dataset was used in the OpenAI Cookbook to fine-tune the OpenAI gpt-oss models.
You can load the dataset using:
from datasets import… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/Multilingual-Thinking.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"
Multi-SWE-RL-Verified
Multi-SWE-RL-Verified
Gold-patch-validated subset of
PrimeIntellect/Multi-SWE-RL-Reupload
(ByteDance's Multi-SWE-RL): 2,232 / 4,703 rows across
C, Go, Java, JavaScript, Rust, and TypeScript that produce a clean reward signal end-to-end.
Default dataset of the multiswe_v1 taskset.
Changes vs upstream
Starting from the 4,703-row re-upload:
C++ dropped wholesale — 0/449 rows passed gold-patch validation in pass 1; the images are
broken for scoring, not merely… See the full description on the dataset page: https://huggingface.co/datasets/PrimeIntellect/Multi-SWE-RL-Verified.McEvalMcEval benchmark data as described in the McEval Paper. Code for the evaluation can be found on Github as McEval.
Nemotron-SFT-Multilingual-v2
Dataset Description:
Nemotron-SFT-Multilingual-v2 is a multilingual supervised fine-tuning (SFT) dataset for post-training text-generation models. It is generated by translating seed data from Nemotron-Math-v2, Nemotron-Competitive-Programming-v1, and Nemotron-Science-v1, adding multilingual coverage for Hindi (hi), Korean (ko), Brazilian Portuguese (pt-br), and refreshed Japanese (ja) data.
The dataset is generated with a new data processing pipeline that avoids line-breaking… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-SFT-Multilingual-v2.CLaRa_multi_stage
CLaRa: Bridging Retrieval and Generation with Continuous Latent Reasoning
This is the official dataset for the CLaRa paper which contains training and evaluation data for the CLaRa model, organized into three main categories: pretraining, instruction tuning, and end-to-end tuning.
Dataset Structure
1. Pretraining Data
pretraining: Large-scale pretraining data for the compressor learning
Format: JSONL with fields: data_type, question, answers… See the full description on the dataset page: https://huggingface.co/datasets/apple/CLaRa_multi_stage.MT-Nemotron-CC
MultiSynt
MultiSynt is an open multilingual synthetic dataset.
The MT Nemotron-CC subset of MultiSynt is made of automatic translations into multiple languages from a subset of approximately 100B tokens from the high-quality split of the English Nemotron-CC dataset.
This subset is made available using different translation models:
Unbabel/Tower-Plus-9B (translations into 16 languages)
Unbabel/Tower-Plus-72B (translations into 5 languages)
Opus-MT and HPLT-MT (translations… See the full description on the dataset page: https://huggingface.co/datasets/MultiSynt/MT-Nemotron-CC.hle-multilingual
HLE Multilingual
Multilingual translations of HLE (Humanity's Last Exam), an expert-level QA benchmark with questions across math, science, humanities, and engineering designed to challenge even domain experts.
Source: cais/hle (test split, 2,158 text-only questions out of 2,500 total)
Languages
Config
Language
Examples
ces
Czech
50
dan
Danish
50
deu
German
800
fin
Finnish
50
fra
French
50
ita
Italian
50
nld
Dutch
50
pol
Polish
50
spa… See the full description on the dataset page: https://huggingface.co/datasets/ellamind/hle-multilingual.WangchanThaiInstruct_Multi-turn_Conversation_Dataset
WangchanThaiInstruct Multi-turn Conversation Dataset
We create a Thai multi-turn conversation dataset from airesearch/WangchanThaiInstruct (Batch 1) by LLM. It was created from synthetic method using open source LLM in Thai language.
Citation
Thammaleelakul, S., & Phatthiyaphaibun, W. (2024). WangchanThaiInstruct Multi-turn Conversation Dataset [Data set]. Zenodo. https://doi.org/10.5281/zenodo.13132633
or BibTeX
@dataset{thammaleelakul_2024_13132633,
author =… See the full description on the dataset page: https://huggingface.co/datasets/ThaiSyntheticQA/WangchanThaiInstruct_Multi-turn_Conversation_Dataset.gsm8k-platinum-multilingual
GSM8K Platinum Multilingual
Multilingual translations of GSM8K Platinum, a rigorously cleaned and verified version of GSM8K containing 1,209 elementary math word problems requiring multi-step arithmetic reasoning.
Source: madrylab/gsm8k-platinum (test split, 1,209 questions)
Languages
Config
Language
Examples
ces
Czech
100
dan
Danish
100
deu
German
1,209
fin
Finnish
100
fra
French
100
ita
Italian
100
nld
Dutch
100
pol
Polish
100
spa
Spanish… See the full description on the dataset page: https://huggingface.co/datasets/ellamind/gsm8k-platinum-multilingual.Multi-SWE-RL-Reupload
Multi-SWE-RL-Reupload
Verbatim re-upload of ByteDance's community-sourced
Multi-SWE-RL
(paper): 4,703 containerized issue-resolving tasks across
C, C++, Go, Java, JavaScript, Rust, and TypeScript.
For training, prefer
PrimeIntellect/Multi-SWE-RL-Verified,
the gold-patch-validated subset of this data.
Changes vs upstream
Storage schema only: per-test maps are stored as columnar struct-of-lists so the rows load
cleanly with datasets (the upstream nested structs… See the full description on the dataset page: https://huggingface.co/datasets/PrimeIntellect/Multi-SWE-RL-Reupload.Dolci-Think-SFT-32B-Multilingual
Dolci-Think-SFT-32B-Multilingual
Dolci-Think-SFT-32B-Multilingual is a large-scale multilingual long chain-of-thought (CoT) reasoning corpus spanning six languages: English, French, German, Spanish, Chinese, and Swahili. Each sample includes a question, a long-form reasoning trace, and a final answer, all translated into the target language, with sequences up to 32,768 tokens.
It is released alongside the paper Rethinking the Multilingual Reasoning Gap with Layer Swap.… See the full description on the dataset page: https://huggingface.co/datasets/lightonai/Dolci-Think-SFT-32B-Multilingual.propagator-multimodal-pretraining-data
Propagator Multimodal Pretraining Data
This public dataset contains tokenized multimodal pretraining data prepared for the Propagator model family. It combines language, image-grounded, and speech/audio-token examples into a single training format.
This is not a raw text or image browsing dataset. The examples have already been converted into compact binary token frames for model training, with a manifest that records the source groups and file layout.
Source Code… See the full description on the dataset page: https://huggingface.co/datasets/ken-sungmin/propagator-multimodal-pretraining-data.MultiSimV2
Dataset Card for MultiSim Benchmark
Dataset Summary
The MultiSim benchmark is a growing collection of text simplification datasets targeted at sentence simplification in several languages. Currently, the benchmark spans 12 languages.
Supported Tasks
Sentence Simplification
Usage
from datasets importload_dataset
dataset = load_dataset("MichaelR207/MultiSimV2")
Citation
If you use this benchmark, please cite our paper:… See the full description on the dataset page: https://huggingface.co/datasets/MichaelR207/MultiSimV2.tiny-aya-l2-thinker-multilingual-reasoning
Tiny Aya L2 Multilingual Reasoning (44 languages)
Translated multilingual reasoning traces used to train Tiny Aya L2-Thinker.
Each example has the prompt, thinking, and answer in the same non-English language alongside the original texts in English.
Data source
Prompts from AM-DeepSeek-R1-0528-Distilled
Thinking traces and outputs distilled from gpt-oss-120b
Translated with command-a-translate and DeepSeek-V3
Languages (44)
Language
Train… See the full description on the dataset page: https://huggingface.co/datasets/CohereLabs/tiny-aya-l2-thinker-multilingual-reasoning.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"
MultiChallenge
MultiChallenge
MultiChallenge is a benchmark for evaluating large language models on multi-turn conversations. It tests whether models can maintain context, follow instructions, and reason correctly across extended dialogues.
Evaluation Axes
Axis
Description
INFERENCE_MEMORY
Tests whether the model can recall and reason over information from earlier turns
INSTRUCTION_RETENTION
Tests whether the model continues to follow instructions given in earlier turns… See the full description on the dataset page: https://huggingface.co/datasets/ScaleAI/MultiChallenge.multiloko
MultiLoKo: a multilingual local knowledge benchmark for LLMs
MultiLoKo is a multilingual knowledge benchmark, covering 30 languages plus English.
The questions are separately sourced for each language, with an annotation protocol designed to target locally relevant topics for the respective language.
MultiLoKo contains the original data for each language, as well as both human and machine-authored translations of each non-English subset into English and vice versa, facilitating… See the full description on the dataset page: https://huggingface.co/datasets/facebook/multiloko.multi-zebra-logic
Dataset Card for the MultiZebraLogic dataset
This dataset includes zebra puzzles in 39 European and 5 non-European languages and in two sizes: 2x3 and 4x5. It can be used for evaluating logical reasoning ability.
The data has been generated using the code in this repo.
Dataset Details
Dataset Description
Zebra puzzles are a type of constraint satisfaction problem. They describe a number of objects, N_objects, that each have attributes… See the full description on the dataset page: https://huggingface.co/datasets/alexandrainst/multi-zebra-logic.multimodal_meme_classification_singapore
Dataset Card for Offensive Memes in Singapore Context
Dataset Details
Dataset Description
This dataset is a collection of memes from various existing datasets, online forums, and freshly scrapped contents. It contains both global-context memes and Singapore-context memes, in different splits. It has textual description and a label stating if it is offensive under Singapore society's standards.
Curated by: Cao Yuxuan, Wu Jiayang, Alistair Cheong, Theodore Lee… See the full description on the dataset page: https://huggingface.co/datasets/aliencaocao/multimodal_meme_classification_singapore.DEBATE
DEBATE: Diverse Multi-Agent Debates
This dataset is presented in the paper "MALLM: Multi-Agent Large Language Models Framework".
Citation
comming soon.
gsm8k-multilingual-reasoning
gsm8k-multilingual-reasoning
GSM8K with reasoning translated to multiple languages
Schema
{"prompt": "...", "answer": "...", "reasoning": "...", "metadata": {...}}
Usage
from datasets importload_dataset
ds = load_dataset("eddie-OB/gsm8k-multilingual-reasoning")
print(ds["train"][0])
Source
Derived from OpenAI GSM8K.
MultimodalMathBenchmarks
MultimodalMathBenchmarks
This repository contains the datasets for the paper Multiplication in Multimodal LLMs: Computation with Text, Image, and Audio Inputs (ACL Findings 2026).
It covers the public benchmark datasets and their modality assets (text, images, and audio) used to evaluate the arithmetic capabilities of multimodal LLMs.
Canonical Upload Manifest
HF path
Local source
Count
Purpose
SharedMultimodalGrid.csv
SavedData/SharedMultimodalGrid.csv… See the full description on the dataset page: https://huggingface.co/datasets/cjerzak/MultimodalMathBenchmarks.MT-Reasoning
MultiSynt
MultiSynt is an open multilingual synthetic dataset.
The MT Reasoning subset of MultiSynt is made of automatic translations into 2 languages of Glaive AI reasoning dataset containing 22mil+ general reasoning questions, reasoning traces and responses.
lang
rows
prompt_tokens
reasoning_tokens
response_tokens
total_tokens
deu_Latn
17_354_716
1_873_153_732
26_010_932_738
14_862_651_336
42_746_737_806
fra_Latn
17_354_716
1_802_885_115
25_224_272_259… See the full description on the dataset page: https://huggingface.co/datasets/MultiSynt/MT-Reasoning.gpqa-multilingual
GPQA Multilingual
Multilingual translations of GPQA (Graduate-Level Google-Proof Q&A), a challenging multiple-choice benchmark requiring graduate-level expertise in biology, physics, and chemistry.
Source: Idavidrein/gpqa (gpqa_main, 448 questions)
Languages
Config
Language
Examples
ces
Czech
448
dan
Danish
448
deu
German
448
fin
Finnish
50
fra
French
448
ita
Italian
448
nld
Dutch
448
pol
Polish
448
spa
Spanish
448
More to be added later.… See the full description on the dataset page: https://huggingface.co/datasets/ellamind/gpqa-multilingual.databricks-dolly-15k-curated-multilingual
Dataset Card for "databricks-dolly-15k-curated-multilingual"
A curated and multilingual version of the Databricks Dolly instructions dataset. It includes a programmatically and manually corrected version of the original en dataset. See below.
STATUS:
Currently, the original Dolly v2 English version has been curated combining automatic processing and collaborative human curation using Argilla (~400 records have been manually edited and fixed). The following graph shows a summary… See the full description on the dataset page: https://huggingface.co/datasets/argilla/databricks-dolly-15k-curated-multilingual.c4_urls_multilingual
Dataset Card for c4_urls_multilingual
This dataset provides the URLs and top-level domains associated with training records in allenai/c4 (multilingual variant). It is part of a collection of datasets curated to make exploring LLM training datasets more straightforward and accessible.
Dataset Details
Dataset Description
This dataset was created by downloading the source data, extracting URLs and top-level domains, and retaining only those record… See the full description on the dataset page: https://huggingface.co/datasets/nhagar/c4_urls_multilingual.knowchat-multi-turn-dialogues
KnowChat: Multi-Turn Human-LLM Dialogues on Knowledge Tasks
KnowChat is a dataset of 705 multi-turn human-LLM conversations collected to validate the KnowSim user simulation framework. It pairs each conversation with pre/post knowledge assessments, self-reported survey ratings, and participant background information, enabling research on information calibration -- how well LLM assistants tailor responses to users with different knowledge levels.
Dataset Summary… See the full description on the dataset page: https://huggingface.co/datasets/yjlee36/knowchat-multi-turn-dialogues.Multi-turn_Long-context_Benchmark_for_LLMs
LoopServe: An Adaptive Dual-phase LLM Inference Acceleration System for Multi-Turn Dialogues
Arxiv: https://www.arxiv.org/abs/2507.13681
Huggingface: https://huggingface.co/papers/2507.13681
Introduction
LoopServe Multi-Turn Dialogue Benchmark is a comprehensive evaluation dataset comprising multiple diverse datasets designed to assess large language model performance in realistic conversational scenarios.
Unlike traditional benchmarks that place queries only at the end… See the full description on the dataset page: https://huggingface.co/datasets/TreeAILab/Multi-turn_Long-context_Benchmark_for_LLMs.