This paper explores the effectiveness of two methods, *Prompt Chaining* and *Stepwise Prompt*, in text summarization to improve the quality of initial summaries through iterative refinement. Prompt Chaining involves a series of three discrete prompts for drafting, critiquing, and refining, while Stepwise Prompt integrates these steps into a single prompt. The study uses the InstruSum dataset, which evaluates LLMs' ability to summarize articles based on specific requirements. Experimental results show that Prompt Chaining consistently outperforms Stepwise Prompt in generating higher-quality summaries. Additionally, the findings suggest that Stepwise Prompt might simulate a refinement process by intentionally producing errors to be corrected. The conclusions have broader implications for various NLP tasks, highlighting the potential of these methods in advancing LLMs.This paper explores the effectiveness of two methods, *Prompt Chaining* and *Stepwise Prompt*, in text summarization to improve the quality of initial summaries through iterative refinement. Prompt Chaining involves a series of three discrete prompts for drafting, critiquing, and refining, while Stepwise Prompt integrates these steps into a single prompt. The study uses the InstruSum dataset, which evaluates LLMs' ability to summarize articles based on specific requirements. Experimental results show that Prompt Chaining consistently outperforms Stepwise Prompt in generating higher-quality summaries. Additionally, the findings suggest that Stepwise Prompt might simulate a refinement process by intentionally producing errors to be corrected. The conclusions have broader implications for various NLP tasks, highlighting the potential of these methods in advancing LLMs.