Ettin-32m-nemotron-pii is based on the ettin-encoder-32M model and fine-tuned over the Nemotron PII dataset. This model can detect 50+ PII entities in both structured and unstructured texts across various domains like healthcare, finance, legal, cybersecurity etc. With just 32M parameters, the model achieves a strong F1-score of 95.73.

Key Features

Achieves strong F1-score of 95.73 with just 32M parameters.
Outperforms popular LLMs like DeepSeek-V4-Flash (84.89) and GPT-4o-Mini (78.69).
Detects 50+ PII entities in both structured and unstructured texts.
Handles text in various domains like healthcare, finance, legal etc.

Supported PII Entity Types

This model can detect the following 55 PII entity types

PII entity types with description

Entity	Description
account_number	Account Number
age	Age
api_key	API Key
bank_routing_number	Bank Routing Number
biometric_identifier	Biometric Identifier
blood_type	Blood Type
certificate_license_number	Certificate or License Number
city	City
company_name	Company Name
coordinate	Geographic Coordinate
country	Country
county	County
credit_debit_card	Credit or Debit Card Number
customer_id	Customer ID
cvv	Card Verification Value (CVV)
date	Date
date_of_birth	Date of Birth
date_time	Date and Time
device_identifier	Device Identifier
education_level	Education Level
email	Email Address
employee_id	Employee ID
employment_status	Employment Status
fax_number	Fax Number
first_name	First Name
gender	Gender
health_plan_beneficiary_number	Health Plan Beneficiary Number
http_cookie	HTTP Cookie
ipv4	IPv4 Address
ipv6	IPv6 Address
language	Language
last_name	Last Name
license_plate	Vehicle License Plate
mac_address	MAC Address
medical_record_number	Medical Record Number
national_id	National Identification Number
occupation	Occupation
password	Password
phone_number	Phone Number
pin	Personal Identification Number (PIN)
political_view	Political View
postcode	Postcode / Zip Code
race_ethnicity	Race or Ethnicity
religious_belief	Religious Belief
sexuality	Sexuality / Sexual Orientation
ssn	Social Security Number
state	State
street_address	Street Address
swift_bic	SWIFT / BIC Code
tax_id	Tax Identification Number
time	Time
unique_id	Unique Identifier
url	URL / Web Address
user_name	Username
vehicle_identifier	Vehicle Identification Number (VIN)

Usage

python

# First install Hugging Face transformers library
!pip install transformers

# Initialize and run the PII detection pipeline to extract PII entities
from transformers import pipeline

## Initialize the PII detection pipeline
ner = pipeline("ner", model="kalyan-ks/ettin-32m-nemotron-pii", aggregation_strategy="simple")

input_text = "Kalyan KS is from India. His email id is kalyan.ks@yahoo.com"

## Run the PII detection to extract PII entities
pii_entities = ner(input_text)

## Process the extracted PII entities 
def format_pii_entities(entities, original_text):
    if not entities:
        return []

    merged_entities = []

    entities = sorted(entities, key=lambda x: x['start'])

    current_entity = {
        'start': entities[0]['start'],
        'end': entities[0]['end'],
        'label': entities[0]['entity_group'],
        'text': entities[0]['word']
    }

    for next_ent in entities[1:]:
        is_same_label = next_ent['entity_group'] == current_entity['label']
        is_adjacent = next_ent['start'] <= current_entity['end'] + 1

        if is_same_label and is_adjacent:
            current_entity['end'] = max(current_entity['end'], next_ent['end'])
            current_entity['text'] = original_text[current_entity['start']:current_entity['end']]
        else:
            merged_entities.append(clean_entity(current_entity))
            current_entity = {
                'start': next_ent['start'],
                'end': next_ent['end'],
                'label': next_ent['entity_group'],
                'text': next_ent['word']
            }

    merged_entities.append(clean_entity(current_entity))
    return merged_entities

def clean_entity(ent):

    raw_text = ent['text']
    stripped_text = raw_text.strip()
    leading_spaces = len(raw_text) - len(raw_text.lstrip())

    return {
        'start': ent['start'] + leading_spaces,
        'end': ent['start'] + leading_spaces + len(stripped_text),
        'text': stripped_text,
        'label': ent['label']
    }

# Display the extracted PII entities
formatted_entities = format_pii_entities(pii_entities, input_text)
print(formatted_entities)

# Output
[{'start': 0, 'end': 9, 'text': 'Kalyan KS', 'label': 'first_name'}, {'start': 18, 'end': 23, 'text': 'India', 'label': 'country'}, {'start': 41, 'end': 60, 'text': 'kalyan.ks@yahoo.com', 'label': 'email'}]

Evaluation

This model is evaluated on a 10k sample test set from Neomotron PII dataset and achieved the following results

Metric	Score
F1	95.73
Precision	95.96
Recall	95.49
Accuracy	99.17

Top Performing PII Entity Types

Entity	Precision	Recall	F1
mac_address	0.9965	1.0000	0.9982
biometric_identifier	0.9957	0.9978	0.9967
date_of_birth	0.9937	0.9976	0.9956
email	0.9937	0.9940	0.9939
api_key	0.9921	0.9949	0.9935
coordinate	0.9910	0.9955	0.9932
vehicle_identifier	0.9863	0.9981	0.9922
medical_record_number	0.9960	0.9881	0.9921
employee_id	0.9909	0.9915	0.9912
credit_debit_card	0.9935	0.9870	0.9902

Challenging PII Entity Types

Entity	Precision	Recall	F1
occupation	0.7110	0.5115	0.5949
time	0.8614	0.7784	0.8178
political_view	0.8348	0.8842	0.8588
age	0.8220	0.9184	0.8676
state	0.8941	0.8570	0.8751
national_id	0.8671	0.8999	0.8832
company_name	0.8860	0.8860	0.8860
fax_number	0.9070	0.8742	0.8903
race_ethnicity	0.8611	0.9299	0.8942
education_level	0.9232	0.8874	0.9049

Limitations

Language: This model works well only for English language texts.
Challenging PII Entity Types: Some of the entity types like occupation has low F1 score.

Citation

bibtex

@misc{ettin-32m-pii-2026,
  title = {ettin-32m-nemotron-pii-2026: PII Detection Model},
  author = {Kalyan KS},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/kalyan-ks/ettin-32m-nemotron-pii}
}