Learn Before
Example of Defining Answer Semantics for Grammaticality Judgment
A prompt can explicitly define the meaning of possible answers to guide a language model. For instance, in a grammaticality judgment task, the instructions might state: 'Yes = the following sentence is grammatically correct. No = it contains grammatical errors.' This clarification ensures the model understands the semantics of the expected 'Yes' or 'No' output.
0
1
Tags
Ch.2 Generative Models - Foundations of Large Language Models
Foundations of Large Language Models
Computing Sciences
Foundations of Large Language Models Course
Related
Example of Defining Answer Semantics for Grammaticality Judgment
Example of Defining Category Semantics in a Polarity Classification Prompt
A data scientist is using a language model to classify customer feedback into 'Bug Report' or 'Feature Request'. Their initial prompt is:
Feedback: 'The app crashes when I try to upload a photo.' What kind of feedback is this?They observe that the model's outputs are inconsistent, including responses like 'This is a bug report,' 'It seems like a bug,' and 'The user is reporting a problem with the app.' Which of the following revised prompts best addresses this inconsistency by explicitly defining the required output format and the meaning of the categories?Improving Prompt Specificity for Automated Data Extraction
Refining a Prompt for Feature Request Identification
Example of a Constraint-First Prompt for Grammaticality Judgment
Learn After
A researcher is designing an instruction for a language model to determine if a sentence is grammatically sound. The initial instruction is: "Is the following sentence grammatically correct?"
When tested, the model's responses are inconsistent. It sometimes answers "Yes, it is correct," other times "No, that sentence has an error," and occasionally it just provides a corrected version of the sentence.
Which of the following revised instructions would be most effective at solving this problem by clarifying the expected output?
Improving Prompt Reliability for Sentiment Analysis
Crafting a Clear Classification Prompt