README.md

---
title: IAB Content Taxonomy Classifier
emoji: 🏷️
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: "5.50.0"
app_file: app.py
pinned: true
license: other
tags:
  - iab
  - content-classification
  - text-classification
  - modernbert
  - multilingual
  - adtech
  - brand-safety
---

# IAB Content Taxonomy Classifier

Hierarchical web content classification using a ModernBERT-base encoder
fine-tuned on 33.5 million multilingual domain examples across 26 languages.

**Coverage:** IAB Tech Lab Content Taxonomy 3.0 — 27 Tier-1 · 434 Tier-2 · 217 Tier-3 categories

| Tier | Categories | Test Accuracy |
|------|-----------|---------------|
| Tier 1 | 27 | 98.1% |
| Tier 2 | 434 | 79.8% |
| Tier 3 | 217 | 56.1% |

## Setup

Set `HF_TOKEN` as a Space secret to access the private model weights.
Set `MODEL_REPO` to override the default model repository.

## Security

This Space is **private**. The underlying model weights and training data
are not exposed to Space users. Authentication is managed via HuggingFace
Space access controls.