datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
synthetic_pii_finance_multilingual
Image generated by DALL-E. See prompt for more details
💼 📊 Synthetic Financial Domain Documents with PII Labels
gretelai/synthetic_pii_finance_multilingual is a dataset of full length synthetic financial documents containing Personally Identifiable Information (PII), generated using Gretel Navigator and released under Apache 2.0.
This dataset is designed to assist with the following use cases:
🏷️ Training NER (Named Entity Recognition) models to detect and label PII in… See the full description on the dataset page: https://huggingface.co/datasets/gretelai/synthetic_pii_finance_multilingual.pile-pii-scrubadub
Dataset Card for pile-pii-scrubadub
Dataset Summary
This dataset contains text from The Pile, annotated based on the personal idenfitiable information (PII) in each sentence.
Each document (row in the dataset) is segmented into sentences, and each sentence is given a score: the percentage of words in it that are classified as PII by Scrubadub.
Supported Tasks and Leaderboards
[More Information Needed]
Languages
This dataset is taken from The… See the full description on the dataset page: https://huggingface.co/datasets/tomekkorbak/pile-pii-scrubadub.meddies-pii-mixedReSID-dataset
Amazon Reviews 2023 (10 Categories, Post-processed)
Paper | Code | Original Source
Overview
This dataset is a curated and post-processed subset of Amazon Reviews 2023. We select 10 product categories and apply a standard preprocessing pipeline widely used in sequential recommendation research. The resulting dataset provides user interaction sequences along with structured item side information.
Categories: 10
Content: user interaction sequences + structured item features… See the full description on the dataset page: https://huggingface.co/datasets/PIIR/ReSID-dataset.tw-PII-bench
Taiwan PII Benchmark (tw-PII-bench)
A token-classification benchmark for evaluating PII detectors on Taiwan-specific personally identifiable information in Traditional Chinese (繁體中文). Designed against openai/privacy-filter to surface its label-coverage gaps and locale-specific failure modes.
The benchmark has three splits by text length, so you can isolate where a model breaks (boundary handling, long-context coverage, multi-PII reasoning):
Split
Items
Text lengthAvg PII… See the full description on the dataset page: https://huggingface.co/datasets/lianghsun/tw-PII-bench.ManCAR
Amazon Reviews 2023 (7 Categories, Post-processed)
Overview
This dataset is a curated and post-processed subset of Amazon Reviews 2023.
We select 7 product categories and apply a standard preprocessing pipeline widely used in sequential recommendation research.
We adopt the official absolute-timestamp split provided by the corpus.
Included Categories
CDs_and_Vinyl
Video_Games
Toys_and_Games
Musical_Instruments
Grocery_and_Gourmet_Food
Arts_Crafts_and_Sewing… See the full description on the dataset page: https://huggingface.co/datasets/PIIR/ManCAR.bigcode-pii-pjj_checksstack-filtered-pii-1M-java
Dataset Card for "stack-filtered-pii-1M-java"
More Information needed
KGD-dataset
Amazon Reviews 2023 (8 Categories, KGD Preprocessed)
Paper | Code | Original Source
Overview
This dataset provides the preprocessed data used to train and evaluate KGD (Knowledge–Geometry Decoupling) with ManCAR as the backbone. It is a curated subset of Amazon Reviews 2023 covering 8 product categories, and ships not only the interaction splits but also the auxiliary signals required by KGD:
ready-to-train user interaction sequences (train/valid/test);… See the full description on the dataset page: https://huggingface.co/datasets/PIIR/KGD-dataset.aliasit-pii-dataset-v2
aliasit-pii-dataset-v2
Dataset PII multilingue in formato canonico testo + character span, 82 categorie, 6 lingue, derivato da quattro sorgenti pinnate a commit.
documenti
170.448
entita'
3.127.084
categorie
82 (48 marcate critiche)
lingue
de, en, es, fr, it, pt
tassonomia
v1.1.0 (taxonomy.yaml nel repo)
versione
2.0.0
costruito il
2026-08-26T17:19:46+00:00
Formato
La rappresentazione canonica e' testo + character span, non BIO: gli… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v2.aliasit-pii-dataset-v4
aliasit-pii-dataset-v4
Italian PII dataset in canonical text + character span form, 44 entity types that identify a person, derived from 26 pinned sources. Template families never cross splits, and the build stops if they do.
What this is
An Italian-only PII dataset in text + character span form. Spans are character offsets into the raw text, end is exclusive, and they are independent of any tokenizer: converting them to token labels is the consumer's job, so the… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v4.arxiv-pii
ArXiv PII
Personally Identifiable Information (PII) from ArXiv front pages.
The dataset contains:
first pages of ArXiv PDFs converted to text
author related labels: NAME, EMAIL, PHONE
Available splits and data statistics:
split
pages
NAME
EMAIL
PHONE
2312
1749
9871
3993
39
2401
1312
7477
3039
36
Usage
ds_pages = load_dataset("sorokin/arxiv-pii", "pages", split="2312")
ds_labels = load_dataset("sorokin/arxiv-pii", "labels", split="2312")… See the full description on the dataset page: https://huggingface.co/datasets/sorokin/arxiv-pii.REDACT-PII-Benchmark
REDACT
REDACT: A Systematically Controlled Multilingual Benchmark for Personal Information Detection
Accepted to the EMNLP 2026 Industry Track
Paper / arXiv · GitHub
REDACT is a multilingual benchmark for evaluating personal information detection under systematically controlled generation conditions.
13,427 records · 324,078 entity annotations · 51 canonical entity types · 25 languages · 9 scripts · 4,127 surface-form patterns
Benchmark task
The headline task is… See the full description on the dataset page: https://huggingface.co/datasets/guneeshv/REDACT-PII-Benchmark.aliasit-pii-dataset-v2-small
aliasit-pii-dataset-v2-small
Sottoinsieme stratificato di aliasit-pii-dataset-v2 per le prove di training: stesse categorie e stesso formato, dimensioni da smoke test.
documenti
8.620
entita'
124.644
categorie
82 (48 marcate critiche)
lingue
de, en, es, fr, it, pt
tassonomia
v1.1.0 (taxonomy.yaml nel repo)
versione
1.0.0
costruito il
2026-08-26T17:36:35+00:00
Formato
La rappresentazione canonica e' testo + character span, non BIO: gli… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v2-small.aliasit-pii-dataset-v4-small
aliasit-pii-dataset-v4-small
One fifth of the training split of aliasit-pii-dataset-v4, sampled uniformly on ids so the category distribution survives, with validation and test kept whole. For training runs that fail fast and still evaluate on the real thing.
What this is
An Italian-only PII dataset in text + character span form. Spans are character offsets into the raw text, end is exclusive, and they are independent of any tokenizer: converting them to token… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v4-small.pii-skills-ablation-results
PII Skills Ablation — Scored Results
This repository contains model predictions and evaluation scores for the ablation study described in:
"Asymmetry, Not Capability: Evaluation Shapes Tool-Augmented PII Detection in Small Language Models"
Results are produced by running four open-weight instruction-tuned models (Gemma 2 9B, Llama 3.1 8B, Mistral 7B, Qwen 2.5 7B) under four primary conditions (zero-shot, +Docs, +Tool, +Skills), plus three baselines (standalone PII-Codex detector… See the full description on the dataset page: https://huggingface.co/datasets/EdyVision/pii-skills-ablation-results.aliasit-pii-dataset-v3
aliasit-pii-dataset-v3
Multilingual PII span-annotation dataset, 79 entity types, 30 pinned sources, Italian-first. Text plus character spans, single-label BIO.
Documents
182,862
Entities
3,172,535
Entity types
79 (46 marked critical)
Languages
ar, de, el, en, es, fr, it, nl, pt, sl, tr
Taxonomy
v3.0.0 — taxonomy.yaml ships in this repo
Revision
6.0.0
Representation
text + character spans, end exclusive, non-overlapping
Distinct sources
27, each pinned… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v3.aliasit-pii-dataset-v5
aliasit-pii-dataset-v5
Italian PII dataset in canonical text + character span form, the 42 entity types of v4, with the form variety and the document length that a gold set of real Italian documents showed were missing: perturbed surface forms, composed multi-section documents with anaphoric surname references, and numeric material deliberately labelled O.
Read this first: the text is rewritten
Every dataset in this family before v5 kept the source text byte for… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v5.aliasit-pii-dataset-v3-small
aliasit-pii-dataset-v3-small
Half the training split of aliasit-pii-dataset-v3, with validation and test kept whole. For cheap training runs that still evaluate on the real thing.
This is a subsample of mapo80/aliasit-pii-dataset-v3. The training split is halved by uniform sampling on document id; validation and test are the full splits of the parent dataset, unchanged. It exists to make a training run cheap without making the evaluation a different question. Every category… See the full description on the dataset page: https://huggingface.co/datasets/mapo80/aliasit-pii-dataset-v3-small.pii_datasettextcat-tokencat-pii-per-domain
Dataset Card for textcat-tokencat-pii-per-domain
This dataset has been created with Argilla. As shown in the sections below, this dataset can be loaded into your Argilla server as explained in Load with Argilla, or used directly with the datasets library in Load with datasets.
Using this dataset with Argilla
To load with Argilla, you'll just need to install Argilla as pip install argilla --upgrade and then use the following code:
import argilla as rg
ds =… See the full description on the dataset page: https://huggingface.co/datasets/argilla/textcat-tokencat-pii-per-domain.pii-pile-chunk3-0-50000pii-pile-chunk3-0-50000-tagged
Dataset Card for "pii-pile-chunk3-0-50000-tagged"
More Information needed
pii-pile-chunk3-1000000-1050000australian-insurance-pii-dataset-correctedpii_dataset_augmented_512piiptbrchatml
Dataset Card for PII PT-BR ChatML
The piiptbrchatml dataset is designed for training and evaluating models for Personal Identifiable Information (PII) masking in Brazilian Portuguese.
It contains conversations where a system is instructed to mask PII from user inputs. The dataset includes the original text, the masked text, and the identified PII entities.
O dataset piiptbrchatml foi criado para treinar e avaliar modelos para mascaramento de Informações Pessoais Identificáveis… See the full description on the dataset page: https://huggingface.co/datasets/cicero-im/piiptbrchatml.pii-inverter-dataset-80kpii-pile-chunk3-1300000-1350000gretelai_synthetic_pii_finance_english