datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
MindBenchOur homepage: https://miasanlei.github.io/MindBench.github.io/
Our GitHub repository: https://github.com/MiaSanLei/MindBench
browsecomp-plusMIA-BenchMIA
Memory Intelligence Agent (MIA)
Paper | GitHub
MIA (Memory In Intelligence Agent) is a memory framework designed for deep research agents (DRAs). It transforms agents from "passive record-keepers" into "active strategists" using a Manager-Planner-Executor architecture. This repository contains the datasets and data artifacts used to train and evaluate the MIA framework.
Dataset Description
The dataset includes the following components:
Train: Data used for the… See the full description on the dataset page: https://huggingface.co/datasets/LightningCreeper/MIA.Tab-MIA
Tab-MIA: A Benchmark for Membership Inference Attacks on Tabular Data
Tab-MIA is a benchmark dataset designed to evaluate the privacy risks of fine-tuning large language models (LLMs) on structured tabular data. It enables reproducible and systematic testing of Membership Inference Attacks (MIAs) across diverse datasets and six different serialization formats.
📋 Overview
Datasets:
WTQ (WikiTableQuestions)
WikiSQL
TabFact
Adult Census
California Housing… See the full description on the dataset page: https://huggingface.co/datasets/germane/Tab-MIA.MIAF_DomainDetection_Infrastructure_Datasets
MIAF: Domain Detection Infrastructure Datasets
This collection is the standardized evaluation benchmark for MIAF (Modular Infrastructure-Aware Fusion). It provides nine classification datasets derived from four public malicious-domain benchmarks, each paired with a shared 137-feature infrastructure representation.
Overview
We evaluate MIAF across nine classification datasets derived from four public malicious-domain benchmarks:
DomainRadar (Hranický et al.… See the full description on the dataset page: https://huggingface.co/datasets/ahmedBargady/MIAF_DomainDetection_Infrastructure_Datasets.Anime-2026CCCleanerDataset8Gagnews-unlearning-mia
AGNEWS - Machine Unlearning + MIA Evaluation Dataset (Length-Filtered)
This dataset is prepared for evaluating machine unlearning methods on fine-tuned LLMs using Membership Inference Attacks (MIAs).
Dataset Splits
Training Sets (for Unlearning)
retain_set (9,000 samples): Data to retain during unlearning
forget_set (1,000 samples): Data to unlearn
Evaluation Sets (for MIA) - Length-Filtered
AGNews Length Variants
32 tokens (~32±10… See the full description on the dataset page: https://huggingface.co/datasets/h0ssn/agnews-unlearning-mia.shadow_fma_medium_1K_240427Bangor-Miami-Spanish-English-Corpus
Bangor Miami Spanish-English Corpus
The Bangor Miami Corpus is a naturalistic Spanish-English code-switching speech dataset collected by Jon Russell Herring at Bangor University. It captures spontaneous bilingual conversations recorded in Miami, Florida, involving proficient Spanish-English bilinguals across multiple speaker groups.
Dataset description
Total recordings
56
Total duration
~32 h
Languages
English (en), Spanish (es)
Format
MP3 audio +… See the full description on the dataset page: https://huggingface.co/datasets/BrunoHays/Bangor-Miami-Spanish-English-Corpus.Miaoshou_Jade-Miura_DatasetKeSpeech
This dataset only contains test data, which is integrated into UltraEval-Audio(https://github.com/OpenBMB/UltraEval-Audio) framework.
python audio_evals/main.py --dataset KeSpeech --model gpt4o_audio
🚀超凡体验，尽在UltraEval-Audio🚀
UltraEval-Audio——全球首个同时支持语音理解和语音生成评估的开源框架，专为语音大模型评估打造，集合了34项权威Benchmark，覆盖语音、声音、医疗及音乐四大领域，支持十种语言，涵盖十二类任务。选择UltraEval-Audio，您将体验到前所未有的便捷与高效：
一键式基准管理 📥：告别繁琐的手动下载与数据处理，UltraEval-Audio为您自动化完成这一切，轻松获取所需基准测试数据。
内置评估利器… See the full description on the dataset page: https://huggingface.co/datasets/miaocongxin/KeSpeech.physical-relighting-datasetmia9MIAAM-V2
Mathematics Dataset Card: AM, Adaptiv World, and Adaptiv College
Overview
This repository contains a cross-country mathematics interaction dataset built from three digital learning platforms used in classrooms in France and Côte d’Ivoire. It spans three educational age bands, from early primary school to middle school and high-school remediation, and covers several mathematical domains, including number sense, arithmetic problem solving, geometry, fractions and… See the full description on the dataset page: https://huggingface.co/datasets/GAIMHE/MIAAM-V2.scaling_mia_the_pile_00_PubMed_Centralscaling_mia_the_pile_00_Pile-CCxsum-MIA-BenchmarkChineseNovels
中文小说数据集
包含内容：
网游/系统/重生
言情小说
同人/耽美小说
科幻小说
军事小说
以上加起来共4万本左右
海棠文学城小说：约1000本（未清洗）
audio-mia-batch-20260312
Audio MIA Batch 20260312
This dataset contains 6,998 audio files (64 GB) downloaded from YouTube videos.
Dataset Structure
Each row contains:
audio: Audio bytes (playable in the dataset viewer)
video_id: YouTube video ID
category: Content category
source_term: Search term used
query: Full search query
title: Video title
url: YouTube URL
uploader: Channel name
channel_id: YouTube channel ID
upload_date: Upload date (YYYY-MM-DD)
duration: Video duration in seconds… See the full description on the dataset page: https://huggingface.co/datasets/potsawee/audio-mia-batch-20260312.scaling_mia_the_pile_00_OpenWebText2glm53-flash-fidelity-exl3-tr3-4bpw-miaailab-v1
fidelity--glm53flash.malaiwah.quant.miaailab-4bpw
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from Mia-AiLab/GLM-5.3-Flash-EXL3-TR3-4bpw.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-flash-fidelity-exl3-tr3-4bpw-miaailab-v1.openlvlm-mia
OpenLVLM-MIA
OpenLVLM-MIA: A Controlled Benchmark Revealing the Limits of Membership Inference Attacks on Large Vision-Language Models
Overview
OpenLVLM-MIA offers a controlled benchmark to reassess membership inference attacks (MIA) on large vision-language models beyond dataset-induced biases.
The benchmark consists of a 6,000-image dataset with controlled member/non-member distributions and ground-truth membership at three training stages.
On this setup… See the full description on the dataset page: https://huggingface.co/datasets/paper-2229/openlvlm-mia.openai-humaneval-sky-shadow
Shadow Humaneval dataset
This dataset is generated by GPT-4 to mimic openai-humaneval dataset. Each problem of HumanEval has a corresponding shadow problem in this dataset.
The usage of this dataset is to check Whether a code generation model has data leakage during its training progress. You can refer to Skywork for further details.
scaling_mia_the_pile_00_arxivThis dataset includes all arxiv documents from the 00.jsonl.zst partition of The Pile. It was created with this script:
pile_path = "data/the_pile/train/00.jsonl.zst"
with zstd.open(pile_path, 'r') as fr:
with open("/tmp/arxiv.jsonl", "w") as fw:
for i, line in enumerate(tqdm(fr)):
doc = json.loads(line)
source = doc['meta']['pile_set_name']
if source == "ArXiv":
fw.write(json.dumps(doc) + "\n")
The validation and test sets are… See the full description on the dataset page: https://huggingface.co/datasets/parameterlab/scaling_mia_the_pile_00_arxiv.AGnews-MIA-Benchmarkwikitext-unlearning-mia
WIKITEXT - Machine Unlearning + MIA Evaluation Dataset (Length-Filtered)
This dataset is prepared for evaluating machine unlearning methods on fine-tuned LLMs using Membership Inference Attacks (MIAs).
Dataset Splits
Training Sets (for Unlearning)
retain_set (9,000 samples): Data to retain during unlearning
forget_set (1,000 samples): Data to unlearn
Evaluation Sets (for MIA) - Length-Filtered
WikiText Length Variants
96 tokens (~96±10… See the full description on the dataset page: https://huggingface.co/datasets/h0ssn/wikitext-unlearning-mia.miami_real_estate_data.csvMIAO-I2A