LuisaLuligo/sexism-intention-xlm-roberta

Sexism Detection Models for EXIST2025

This repository provides two fine-tuned transformer models for the EXIST2025 sexism detection challenge:

• Intention Multiclass Detection (sexism-intention-xlm-roberta)

Both models are based on xlm-roberta-large and are ready for direct use with Hugging Face Transformers.

---

How to Use

2. Intention Multiclass Detection

Task: Classify the intention of a sexist tweet as one of: DIRECT, REPORTED, JUDGEMENTAL, NO.

Load and use:

from transformers import AutoTokenizer
from model_class import SoftMultiClassifier  # See below for class definition
import torch
import numpy as np

model_path = "LuisaLuligo/sexism-intention-xlm-roberta"
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = SoftMultiClassifier(model_name=model_path, num_labels=4)
model.eval()

text = "Example tweet here"
inputs = tokenizer(text, return_tensors="pt", truncation=True, padding="max_length", max_length=128)
with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=-1).cpu().numpy()[0]
    idx2label = {0: "DIRECT", 1: "REPORTED", 2: "JUDGEMENTAL", 3: "NO"}
    pred_label = idx2label[int(np.argmax(probs))]
print(f"Prediction: {pred_label} (probs: {probs})")

---

Model Classes

You must use the following classes for inference:

import torch
import torch.nn as nn
from transformers import AutoModel, AutoConfig
from transformers.modeling_outputs import SequenceClassifierOutput


class SoftMultiClassifier(nn.Module):
    def __init__(self, model_name="xlm-roberta-large", num_labels=4, dropout=0.3):
        super().__init__()
        self.config = AutoConfig.from_pretrained(model_name, num_labels=num_labels)
        self.encoder = AutoModel.from_pretrained(model_name)
        self.dropout = nn.Dropout(dropout)
        self.classifier = nn.Linear(self.config.hidden_size, num_labels)
        self.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))

    def forward(self, input_ids, attention_mask=None, **kwargs):
        output = self.encoder(input_ids=input_ids, attention_mask=attention_mask)
        pooled = self.dropout(output.last_hidden_state[:, 0])
        logits = self.classifier(pooled)
        return SequenceClassifierOutput(loss=None, logits=logits)

---

Model Card: sexism-intention-xlm-roberta

Architecture: xlm-roberta-large
Task: Multiclass classification (DIRECT, REPORTED, JUDGEMENTAL, NO)
Languages: Spanish and English
Training Data: Tweets annotated by 6 experts, soft labels used
Loss: CrossEntropyLoss with class weights
Frozen Layers: Embeddings + first 4 encoder layers
Metrics: Macro F1, accuracy
Intended Use: Classifies the intention behind sexist tweets
Limitations: Class imbalance may affect rare classes; requires similar preprocessing for new data.

---

Citation

If you use these models, please cite the EXIST2025 challenge and