Google

Both prompt tuning and prefix fine-tuning incorporate trainable vectors to adapt Large Language Models (LLMs) to specific tasks. However, prefix fine-tuning alters the model extensively by adding trainable vectors to every layer of the Transformer, which requires modifications to the LLM. In contrast, prompt tuning separates soft prompts from the LLMs by modifying only the input embedding layer. This preserves the original model architecture, making prompt tuning more efficient for deployment across different tasks without the need to adjust the core model.

Comparison of Prompt Tuning and Prefix Fine-Tuning

Based on the implementation described in the case study, which adaptation method is being used? Justify your answer by explaining the key architectural difference between this method and another common method that also uses trainable vectors but modifies a different part of the model.

Analysis of a Model Adaptation Strategy

Match each parameter-efficient adaptation method to the description of how it modifies a pre-trained model's architecture.

An engineer is adapting a pre-trained language model for a new task. They want to add a small number of trainable vectors to guide the model's behavior without changing any of the original model weights. What is the fundamental architectural difference between a strategy that adds these vectors only to the input embedding layer versus one that adds them to the input of every transformer layer?

Learn Before

Related