log in  |  register  |  feedback?  |  help  |  web accessibility
PhD Defense: Gradient-Based Prompt Optimization for LLMs: Progress and Challenges
Sicheng Zhu
IRB-4137 or https://umd.zoom.us/j/7541393444
Thursday, April 3, 2025, 4:00-6:00 pm
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract

While gradients are fundamental to optimization in deep learning, their utility in prompt optimization for large language models (LLMs) remains limited. This gap stems from challenges unique to language, such as its discrete nature and the fact that LLM objectives are only defined over discrete inputs. As a result, gradient information is often underutilized, limiting the effectiveness of tasks like prompt engineering and red-teaming.

In this talk, I will present my work on gradient-based prompt optimization, focusing on two components: optimization methods and objectives. First, I will introduce a controllable generation method that generates fluent prompts while eliciting desired model behaviors, such as jailbreaks or false refusals. Next, I will show that existing optimization objectives offer limited control over model behavior. To address this, we propose a new objective that improves behavior control and simplifies the optimization process. Finally, I will explore the potential and challenges of grad-based optimization without search and share early results of a prototype method. This talk aims to provide an overview of recent progress and remaining challenges in gradient-based prompt optimization for LLMs.

Bio

Sicheng Zhu is a PhD candidate in the Computer Science Department at UMD, advised by Prof. Furong Huang. His works on trustworthy machine learning, including robustness and generalization, with a recent focus on safety alignment of LLMs. He has interned at Bosch, Adobe, and Meta.

This talk is organized by Migo Gui