abullard1/germeval2025-c2a-moderngbert-cw

🏆 GermEval 2025: Call to Action Detection (Class-Weighted)

abullardUR@GermEval Shared Task 2025 Submission

---

🎯 Model Summary

This model is a fine-tuned version of LSX-UniWue/ModernGBERT_134M, specifically designed for Call to Action (C2A) Detection in German social media content. It was developed as part of the GermEval 2025 Shared Task on Harmful Content Detection.

🏅 Competition Performance

• Final Ranking: 4th out of 9 teams
• Primary Metric (Macro-F1): 0.82 (+39% over official baseline (0.59 → 0.82))
• Approach: Class-weighted cross-entropy loss to handle severe class imbalance (9.3:1 ratio)

📊 Task Details

• Task Type: Binary classification
• Classes: False (no call to action), True (call to action detected)
• Domain: German social media (Twitter, 2014-2016)
• Data Source: Right-wing extremist network posts

⚠️ Limitations and Bias

Known Limitations

• Domain Specificity: Trained on 2014-2016 German Twitter data from right-wing extremist networks
• Temporal Bias: Language patterns may not reflect contemporary usage
• Class Imbalance: 90.3% negative vs 9.7% positive examples (9.3:1 ratio)
• Cultural Context: May not generalize to other German-speaking regions or contexts
• Implicit Context: May struggle with coded language and contextual references

Ethical Considerations

• Model trained on potentially harmful content for research purposes only
• Should not be used to amplify or generate harmful content
• Requires careful handling due to sensitive training data

🚀 How to Use

Quick Start

from transformers import AutoProcessor, AutoModelForSequenceClassification

# Load model and processor
model_id = "abullard1/germeval2025-c2a-moderngbert-cw"
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForSequenceClassification.from_pretrained(model_id, trust_remote_code=True).eval()

# Run inference
text = "Kommt alle zur Demo am Samstag!"
inputs = processor(text, return_tensors="pt", truncation=True)
probs = model(**inputs).logits.softmax(-1).detach().cpu().numpy()
print(f"Predictions: {probs}")

Class Labels

• Label 0: No call to action detected
• Label 1: Call to action detected

📈 Training Details

Training Data

• Source: GermEval 2025 Shared Task C2A dataset
• Size: 6,840 samples
• Split: 80% training (5,472), 20% validation (1,368)
• Class Distribution: 90.3% negative, 9.7% positive

Training Procedure

• Base Model: ModernGBERT-134M (8192 token context)
• Architecture: Mean-pooling classification head
• Loss Function: Class-weighted cross-entropy (inverse frequency weighting)
• Optimizer: AdamW with linear scheduling
• Early Stopping: Patience of 5 epochs on validation Macro-F1

Hyperparameters

• Learning Rate: 3e-5
• Weight Decay: 0.0973
• Batch Size: 8/32 (train/eval)
• Epochs: 8
• Warmup Steps: 500

📚 Citation

@inproceedings{bullard2025germeval,
  title   = {abullardUR@GermEval Shared Task 2025: Fine-tuning ModernGBERT on Highly Imbalanced German Social Media for Harmful Content Detection},
  author  = {Bullard, Samuel},
  year    = {2025},
  booktitle = {Proceedings of KONVENS 2025 Workshops}
}

🙏 Acknowledgments

• GermEval 2025 Organizers: University of Stuttgart and University of Mannheim
• Prof. Dr. Udo Kruschwitz (University of Regensburg) for supervision
• ModernGBERT Team: LSX-UniWue for the ModernGBERT-134M German language base-model

📄 License

This model inherits the Research-only RAIL-M license from ModernGBERT. See license details.