FIRE: Balancing Stability and Plasticity in Neural Nets
Discover FIRE, a novel method that optimizes neural network reinitialization to balance stability and plasticity for improved learning on nonstationary dat...
Non-Identifiability of Steering Vectors in Large Language Models
Explore why steering vectors in large language models are non-identifiable, impacting AI interpretability and alignment strategies in NLP.
How to Bypass Prompt Injection Detectors in LLMs
Learn advanced techniques to bypass prompt injection detectors in large language models and discover new defense strategies against evasive attacks.
GPERT: Advanced Event-Based 3D Gaussian Ray Tracing
Discover GPERT, a cutting-edge framework for event-based 3D Gaussian ray tracing that boosts accuracy and speed in 3D reconstruction without prior models.
Lumos: Certifying Language Model Systems Safely
Discover Lumos, a framework for certifying language model systems to ensure safety, reliability, and privacy in AI applications like autonomous driving.