Language Models Guidance with Multi-Aspect-Cueing: A Case Study for Competitor Analysis
Amir Hadifar, Christopher Ochs, Arjan Van Ewijk
TL;DR
This paper tackles the gap that LLMs face in dynamic competitive landscapes by proposing Multi-Aspect Cueing (MAC), a prompt-level framework that injects business-related aspects into reasoning prompts. MAC decomposes the analysis into aspects, input values, and demonstrations, and uses Shapley Value Sampling to attribute outputs to each aspect; experiments show improved relevance scoring and memorization in telecom and other domains. Across three datasets (telecom InHouse, Recipe-MPR, FiQA SA), MAC outperforms baselines, with notable improvements on the InHouse dataset, where gains reach $>30\%$ at 5-shot and $>40\%$ at 10-shot. The results suggest that integrating domain-specific aspects into prompts enhances LLMs' ability to trade off criteria and recall relevant signals in shifting competitive landscapes, with practical benefits for competitor analysis workflows.
Abstract
Competitor analysis is essential in modern business due to the influence of industry rivals on strategic planning. It involves assessing multiple aspects and balancing trade-offs to make informed decisions. Recent Large Language Models (LLMs) have demonstrated impressive capabilities to reason about such trade-offs but grapple with inherent limitations such as a lack of knowledge about contemporary or future realities and an incomplete understanding of a market's competitive landscape. In this paper, we address this gap by incorporating business aspects into LLMs to enhance their understanding of a competitive market. Through quantitative and qualitative experiments, we illustrate how integrating such aspects consistently improves model performance, thereby enhancing analytical efficacy in competitor analysis.