Choosing a Feedback Method for LLM Alignment
A development team is aligning a large language model to generate high-quality, emotionally resonant poetry. They are debating between two feedback strategies. Evaluate which of the following strategies is more appropriate for this specific goal and justify your reasoning.
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Combining AI and Human Feedback for LLM Training
Choosing a Feedback Method for LLM Alignment
A development team is aligning a large language model to function as a creative writing partner. The primary goal is to ensure the model's suggestions are imaginative, emotionally resonant, and stylistically unique. The team decides to rely exclusively on an automated, AI-based feedback system for this alignment process. Which of the following statements best identifies a critical flaw in this strategy?
A startup is building an LLM to automatically grade high school history essays. To ensure scalability and rapid deployment, they plan to align the model exclusively using AI-generated feedback. The AI feedback system will be trained to check for factual accuracy against a knowledge base, grammatical correctness, and essay length. What is the most significant risk of this alignment strategy?