Topic: Enhance Reasoning of Large Language Models through Prompt Engineering
Time: 1/24/2025, Friday, 12:00PM
Speaker: Yu Yang (Ph.D., PE, PMP)
Slides: download

The seminar focuses on enhancing the reasoning capabilities of large language models (LLMs) through advanced prompt engineering. Dr. Yang will begin with a brief overview of LLM intelligence before exploring key prompting techniques such as Socratic Questioning, Few-shot Learning, Chain-of-Thought, Self-Consistency, and Universal Self-Consistency.

He will also introduce his Unified Prompt Framework, inspired by Dr. Denny Zhou (Google DeepMind), which synthesizes these methods under the hypothesis that LLMs function as Pattern-Matching Bayesian Machines. The seminar will conclude with practical recommendations for applying these techniques to improve LLM reasoning in real-world use cases.
 

本次研讨会聚焦于通过高级提示工程技术提升大语言模型（LLMs）的推理能力。杨博士将首先简要概述 LLM 的智能特性，然后深入探讨关键的提示技术，包括苏格拉底式提问、Few-shot Learning（少样本学习）、Chain-of-Thought（思维链）、Self-Consistency（自一致性）以及Universal Self-Consistency（通用自一致性）。
他还将介绍其统一提示框架（Unified Prompt Framework），该框架受到 Google DeepMind 的Denny Zhou博士的启发，并基于 LLM 是模式匹配贝叶斯机器的假设。研讨会最后将提供实用建议，帮助参会者将这些技术应用于实际场景中，进一步提升 LLM 的推理能力。
 

Speaker

Yu Yang is a Senior Data Scientist at Xylem with expertise in web application development, machine learning, large language models (LLMs), and AI-generated content (AIGC). He has an impressive academic record, with 34 peer-reviewed publications, an H-index of 24, and over 3,400 citations. Dr. Yang is also the author of one of the first bilingual AIGC books, Poems from Tang Dynasty Meet AI (《唐诗遇见AI》), which is available on Amazon. In addition to his professional and academic pursuits, he actively shares his technical insights and perspectives on platforms like Medium and X.