How Alignment Boosts LLMs’ Multilingual Skills
Discover how alignment enhances large language models' multilingual abilities through neuron classification and shared semantic processing.
Graceful Forgetting in Generative Language Models Explained
Discover how graceful forgetting improves fine-tuning in generative language models by selectively unlearning irrelevant knowledge.
Situationally-Aware Dynamics Learning for Autonomous Robots
Discover how situationally-aware dynamics learning enables autonomous robots to adapt in real-time to complex, unstructured environments for safer, smarter...
Enhancing LLM Honesty Detection with Steering Vectors
Discover how steering vectors improve honesty detection in LLM-judges, boosting reliability and reducing manipulation in AI-generated responses.
Neural Conditional Transport Maps for Optimal Transport
Discover how neural conditional transport maps enhance optimal transport by handling categorical and continuous variables for flexible modeling.