Google

Text completion, also referred to as continual writing, is a class of text generation tasks where the primary goal is to extend or continue a given piece of input text in a coherent manner.

Text Completion (Continual Writing)

A straightforward example of text completion is the task of continuing a story from a set of initial sentences. The model is given the beginning of a narrative and is expected to generate the subsequent text.

Example of Text Completion: Story Continuation

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

Chain-of-Thought (CoT) prompting is a technique used to guide a large language model to detail its reasoning process before providing a final answer. This is often achieved by including a simple instruction like 'Let’s think step by step' in the prompt. By verbalizing its thought process, the model can break down complex problems into smaller, manageable parts, which often improves the accuracy and reliability of its conclusions.

Chain-of-Thought Prompting

A software developer is building two new AI-powered features for a word processing application. Analyze the descriptions below and determine which feature is designed for the task of extending or continuing a given piece of input text. Justify your choice by explaining how the feature's function aligns with this goal.

AI Writing Assistant Feature Analysis

Consider two outputs generated by a language model based on the input 'The sun began to set, casting long shadows across the valley.'

Output A: '...and painting the sky in hues of orange and purple. A cool breeze rustled the leaves in the trees, signaling the end of another day.'

Output B: 'As evening approached, the valley was covered in long shadows from the setting sun.'

Which output (A or B) is an example of extending the input text? Explain your reasoning by describing the primary goal demonstrated in the chosen output.

Distinguishing Text Generation Tasks

A prompt for a conversation completion task can be structured by first setting the context and then providing a clear instruction. For example, a model can be told, 'You are provided with a conversation between a tourist and a taxi driver,' to establish the scenario, followed by the direct command, 'Please complete the conversation.'

Example Prompt for Conversation Completion

A language model is provided with the following text: 'The old bookstore was a labyrinth of towering shelves and narrow aisles. The air smelled of aging paper and leather bindings. In the quietest corner, a single book lay open on a dusty table.' Which of the following model outputs best demonstrates the task of extending the given text in a coherent manner?

Learn Before

Related