DeLLMa: Decision Making Under Uncertainty with Large Language Models

  • 2024-10-11 18:43:48
  • Ollie Liu, Deqing Fu, Dani Yogatama, Willie Neiswanger
  • 0

Abstract

The potential of large language models (LLMs) as decision support tools isincreasingly being explored in fields such as business, engineering, andmedicine, which often face challenging tasks of decision-making underuncertainty. In this paper, we show that directly prompting LLMs on these typesof decision-making problems can yield poor results, especially as the problemcomplexity increases. To aid in these tasks, we propose DeLLMa (Decision-makingLarge Language Model assistant), a framework designed to enhancedecision-making accuracy in uncertain environments. DeLLMa involves amulti-step reasoning procedure that integrates recent best practices in scalinginference-time reasoning, drawing upon principles from decision theory andutility theory, to provide an accurate and human-auditable decision-makingprocess. We validate our procedure on multiple realistic decision-makingenvironments, demonstrating that DeLLMa can consistently enhance thedecision-making performance of leading language models, and achieve up to a 40%increase in accuracy over competing methods. Additionally, we show howperformance improves when scaling compute at test time, and carry out humanevaluations to benchmark components of DeLLMa.

 

Quick Read (beta)

loading the full paper ...