Topic: parameter

A new method called Distilled Rapid Embedding Transfer (DRET) is introduced to adapt general-purpose language models for biomedical applications efficiently, addressing the practical limitations of large domain-specific models like BioBERT and ClinicalBERT by reducing computational demands. This advancement matters because it enables more widespread use of advanced NLP techniques in healthcare settings without the high resource costs associated with specialized models.

arxiv.org↗

Large language models (LLMs) now rely on tool calls and external data, creating a new challenge for unlearning as previous methods may not apply. Researchers propose "agentic tool unlearning" to address this issue, ensuring LLMs can be responsibly modified or reset.

arxiv.org↗