Self-Planning Code Generation with Large Language Models

Human knowledge: decompose complex problems into schedule solution steps prior to implementation

Two parts:

• planning phase
  • plans out concise solution steps from the intent combined with few-shot
• implementation phase
  • generate code step by step following the solution steps

Human intent: What to do LLM Solver: How to do

self-planning

1. The prompt is specilfally designed as k examples and the k is a small number.

2. A step like “Check the number is prime” is better than “If a number is prime, it is not prime. Check if the number is divisible by any number between 2 and n-1. If the number is not dicisible by any number between 2 and n-1, it is prime”

Ealuation Methods

• RQ1: How does self-planning approach perform in code generation compared to baseline approaches?
• RQ2: How does the self-planning approach perform based on different LLMs?
• RQ3: What is the optimal design for the self-planning approach?
• RQ4: How does self-planning approach perform in multilingual code generation?
• RQ5: How does the complexity of the problem affect self-planning?