Google

Few-shot Chain-of-Thought (CoT) prompting is a technique that involves providing a language model with one or more demonstrations of CoT reasoning within the prompt. Each demonstration consists of a problem paired with the detailed, step-by-step thought process used to arrive at the solution, guiding the model on how to reason through similar problems.

Few-Shot Chain-of-Thought (COT) Prompting

The primary distinction between few-shot and zero-shot Chain-of-Thought (CoT) prompting lies in the use of examples. Few-shot CoT requires providing one or more demonstrations of step-by-step reasoning within the prompt. In contrast, zero-shot CoT provides no such examples and instead relies on incorporating a direct instruction to trigger the model's reasoning process.

Comparison of Few-Shot and Zero-Shot CoT Prompting

A significant practical challenge in using few-shot Chain-of-Thought (CoT) prompting is the difficulty of creating the necessary demonstrations. Sourcing or manually crafting detailed, high-quality, multi-step reasoning examples can be a labor-intensive and complex task, whether attempted automatically or by hand.

Difficulty of Creating Few-Shot CoT Demonstrations

You are building a prompt to help a language model solve simple logic puzzles. For the problem below, create a single, complete example demonstration (including the problem, a step-by-step thought process, and the final answer) that could be used as one of the examples in a few-shot chain-of-thought prompt.

**Problem:** There are 5 boxes. The red box is to the left of the blue box. The green box is to the right of the yellow box. The blue box is between the yellow and purple boxes. The yellow box is not at either end. What is the order of the boxes from left to right?

Constructing a CoT Demonstration

A data scientist is struggling to get a language model to consistently extract structured information from unstructured product reviews. After several failed attempts with simple instructions, they design the prompt shown in the case study. Analyze this prompt: identify the core technique being demonstrated and explain how the inclusion of multiple, detailed examples is intended to improve the model's performance on the final, unseen review.

Analyzing a Prompt for Structured Data Extraction

A developer wants to improve a language model's ability to solve multi-step logic puzzles. They decide to construct a prompt that includes examples to guide the model. Which of the following prompt structures best implements the technique of providing multiple, detailed, step-by-step reasoning demonstrations before presenting the final problem to be solved?

Learn Before

Related