A researcher is using the following formula to find the best soft prompt (σ) for a large language model:
hat(σ) = arg min_σ s(hat(y), hat(y)_σ)
In this formula, hat(y) is the model's prediction given a full, descriptive context, and hat(y)_σ is the prediction given the soft prompt. What is the fundamental goal of this optimization process?
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Troubleshooting Soft Prompt Optimization
A researcher is using the following formula to find the best soft prompt (σ) for a large language model:
hat(σ) = arg min_σ s(hat(y), hat(y)_σ)In this formula,
hat(y)is the model's prediction given a full, descriptive context, andhat(y)_σis the prediction given the soft prompt. What is the fundamental goal of this optimization process?Evaluating Soft Prompt Generalization