Tag: AI model robustness
Browse our exclusive articles!
Text Uncanny Valley: LLM Performance Drop on Corrupted Text
Explore how Large Language Models face non-linear performance drops with corrupted text, revealing the Text Uncanny Valley effect in information retrieval.
Self-Mined Hardness: Boosting AI Safety Fine-Tuning
Discover how self-mined hardness improves AI safety by reducing jailbreak attacks and balancing refusal rates in language models.
Fixing Safety Failures in Agentic AI Guard Models
Explore vulnerabilities in agentic AI guard models and how Fisher-Weighted Safety Subspace Regularization restores safety and reliability.
Language Models Detect Dropout and Gaussian Noise Accurately
Discover how advanced language models identify and differentiate dropout and Gaussian noise in activations, enhancing AI robustness and safety.
Preventing Catastrophic Overfitting in Fast Adversarial Training
Discover new insights and strategies to prevent catastrophic overfitting in fast adversarial training for more robust neural networks.