Google

A key application of text completion involves generating text that adheres to specific user-defined requirements. These constraints can cover various aspects such as stylistic conventions, desired tone, specified length, or the inclusion of particular content. For instance, based on such parameters, users might generate outputs like formal reports, creative stories, or programming code.

Text Completion Based on User Requirements

Text generation guided by user requirements can produce a wide range of outputs. Common examples include creating a formal report, writing a creative story, or generating a piece of programming code, each tailored to specific constraints.

Examples of Requirement-Based Text Generation

A specific example of generating text based on user requirements is the task of creating a five-character regulated Chinese poem. This task requires the model to follow strict constraints on form, style, and content, which would be specified in a prompt.

Example of Constrained Text Generation: Chinese Poem

A user provides a text generation model with a set of instructions. Analyze the user's instructions and the model's output below, then identify which specific requirement the output fails to meet and explain your reasoning.

Evaluating Generated Text Against User Requirements

This formula, expressed as $$Pr(y|x, y, ..., x)$$, calculates the probability of a particular element $$y$$ occurring next in a sequence, given the preceding context. This context includes the initial input $$x$$ and any elements that have already been generated.

Conditional Probability Formula for Sequence Generation

A user provides the following vague instruction to a text generation model: 'Write about electric cars.' The model's output is too general to be useful. Revise this instruction by adding three distinct and specific requirements to guide the model toward generating a more focused and helpful response. Your revised instruction should clearly specify constraints related to the desired output's format, tone, and content.

Formulating Effective Text Generation Instructions

A user provides the following instructions to a text generation model: 'Write a three-sentence summary of a company's quarterly earnings report. The tone must be formal and objective, and it must include the exact phrase "year-over-year growth".' Which of the following outputs best fulfills all the user's requirements?

Learn Before

Related