Yuewen Group, recognized as a global leader in online literature and intellectual property operations, has managed to attract approximately 260 million users in over 200 countries through its international platform WebNovel. The company is committed to spreading Chinese web literature globally, adapting quality novels into movies and animations with the aim of expanding China’s cultural influence.
Recently, the company has announced its integration of Prompt Optimization on Amazon Bedrock, a tool that allows optimizing prompts for various uses with a single API call or a simple click in the Amazon Bedrock console. This innovation represents a significant advancement in the performance of large language models (LLMs) for intelligent text processing tasks within the company.
From the beginning, Yuewen Group faced long development cycles and slow updates with its own natural language processing (NLP) models. To overcome these challenges, it adopted Anthropic’s Claude 3.5 Sonnet model on Amazon Bedrock, which offers enhanced capabilities in language understanding and generation, allowing for handling multiple tasks concurrently and with a better understanding of context. However, the company initially struggled to fully leverage the potential of LLMs due to their lack of prompt engineering experience, highlighting the need for effective optimization.
One of the biggest challenges in prompt optimization is the difficulty of evaluation, as the quality of a prompt and its effectiveness in generating desired responses depend on multiple factors, including the language model architecture and training data. Furthermore, context variability implies that an effective prompt in one situation may not work the same in another, requiring significant adjustments. With the increasing use of LLMs, the number of prompts needed also grows, making manual optimization increasingly laborious.
To address these challenges, the automatic prompt optimization technology has started to gain attention. Prompt optimization on Amazon Bedrock provides efficiency by automatically creating high-quality prompts tailored to different LLMs, significantly reducing the time and effort required in manual prompt engineering. Thanks to this, Yuewen Group has experienced significant improvements in the accuracy of intelligent text analytics tasks. For example, in attributing character dialogues, optimized prompts reached a 90% accuracy, surpassing traditional NLP models by 10%.
The implementation of this new technology has revolutionized the prompt engineering process, allowing the company to complete tasks in a more agile and efficient manner. A set of best practices has been compiled to maximize user experience, including the use of clear and concise prompts and avoiding overly long examples.
As artificial intelligence continues to evolve, tools like prompt optimization will become essential for companies to fully capitalize on the benefits of large language models in their operations. Yuewen Group’s experience is a clear example of how these innovations can transform applications in various industries, resulting in considerable time savings and performance improvements.
Referrer: MiMub in Spanish