Marios Savvides

KnowledgeSmith: Uncovering Knowledge Updating in LLMs with Model Editing and Unlearning

MetaVLA: Unified Meta Co-Training for Efficient Embodied Adaptation

Image Tokenizer Needs Post-Training