Learn Before
Improving LLM Consistency for Code Generation
Based on the developer's suspicion that the model is overly influenced by the last example it sees, propose a specific, low-effort strategy using only the existing examples to mitigate this issue and improve the reliability of the code generation. Explain the reasoning behind your proposed strategy.
0
1
Tags
Ch.3 Prompting - 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
A developer is using a large language model for a sentiment analysis task. They have a single prompt containing three distinct examples of text paired with their correct sentiment labels. To improve the consistency of the model's predictions, the developer creates two additional prompts by simply rearranging the order of the original three examples. For any new text, they run all three prompts and take the majority vote of the outputs as the final answer. What is the most likely reason for this approach?
Improving LLM Consistency for Code Generation
Calculating Prompt Variations from Demonstration Order