CLASP
The Centre for Linguistic Theory and Studies in Probability

Solving complex problems with large language models

Abstract

One of the great promises that people connect with LLMs is that they can make complex problem-solving with computers accessible to lay users. Unlike optimal solvers (e.g. for planning or linear programming), LLMs accept natural-language input and require no expert training; unlike earlier task-oriented dialogue systems, they can be applied across arbitrary domains.

In my talk, I will explore the degree to which LLMs are already fulfilling this promise. I will present recent work on whether current LLMs “reason or recite” when solving NP-hard optimization problems and ongoing research on building dialogue agents that play two-player optimization games. I will conclude with some thoughts on future avenues of research.