Allmin Balloccu

Dynamic Embeddings with Task-Oriented prompting

2024 · arXiv (Cornell University)

This paper introduces Dynamic Embeddings with Task-Oriented prompting (DETOT), a novel approach aimed at improving the adaptability and efficiency of machine learning models by implementing a flexible embedding layer. Unlike traditional static embeddings [14], DETOT …