Improving Generalization on Cybersecurity Tasks with Multi-Modal Contrastive Learning

Jianan Huang, Rodolfo V. Valentim, Luca Vassio, Matteo Boffa, Marco Mellia et al.|March 20, 2026arXiv

Key Takeaway

By aligning payload embeddings with text-based vulnerability descriptions using contrastive learning, you can reduce shortcut learning and improve how well cybersecurity models generalize to unseen threats.

Summary

This paper tackles a major problem in cybersecurity AI: models trained in labs fail in the real world because they learn surface-level patterns instead of genuine security concepts.

training multimodal safety

Key Terms

contrastive-learning multimodal-learning embedding-space knowledge-transfer shortcut-learning