---
language:
- en
tags:
- pytorch
- text-classification
- sexism-detection
- exist-2023
- bert
- hatebert
- linear-probing
datasets:
- exist-2023
metrics:
- f1
- precision
- recall
base_model:
- GroNLP/hateBERT
---

# HateBERT Sexism Classifier (Linear Probing / Freeze All)

This model is a fine-tuned version of [GroNLP/hateBERT](https://huggingface.co/GroNLP/hateBERT), trained for multi-class sexism detection on the **EXIST 2023 Task 2** dataset. 

## Experiment Details: `freeze_all`
This repository contains the **Linear Probing** variant of our ablation study. 
* All parameters in the base BERT model (`model.bert.parameters()`) were **frozen** during training. 
* Only the final classification head was trained.
* This approach protects the pre-trained weights from catastrophic forgetting and speeds up training, though it relies entirely on the base model's existing feature representations.

## Intended Use
Categorizes English tweets into one of four sexist intentions:
0. `-` (Non-sexist)
1. `DIRECT` (Directly sexist messages)
2. `JUDGEMENTAL` (Messages condemning sexist behaviors)
3. `REPORTED` (Messages reporting a sexist situation)

## Preprocessing
Because HateBERT is built on `bert-base-uncased`, it automatically lowercases text. If you are keeping preprocessing consistent across your ablation study, ensure your inputs are cleaned accordingly:
* Replace user mentions (`@user`) with the token `@user`
* Replace URLs with the token `http`
* (Handled by tokenizer) Lowercase all text

## Evaluation Results (Test Set)
* **Macro F1:** 0.1950
* **Precision:** 0.1598
* **Recall:** 0.2500

## How to Use
```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification

repo_id = "francesco-zatto/hateBERT-freeze-all-sexism-detector"
tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForSequenceClassification.from_pretrained(repo_id)

inputs = tokenizer("Your cleaned tweet text here", return_tensors="pt")
outputs = model(**inputs)