Can LLMs Hit Moving Targets? Tracking Evolving Signals in Corporate Disclosures

2025-10-03

AI Summary

This paper addresses the “moving target” problem arising from managerial dynamic adjustments of key performance indicators (KPIs) in corporate disclosures, which undermines text-driven financial forecasting. Traditional named entity recognition (NER)-based KPI extraction suffers from high noise and contextual information loss, limiting prediction accuracy. To overcome these limitations, we propose a large language model (LLM)-driven KPI extraction framework that integrates context-aware semantic understanding with a novel indicator-level semantic similarity metric, enabling fine-grained, low-noise identification of KPI evolution. Experiments demonstrate substantial improvements in KPI tracking stability and forecasting accuracy. Across multiple financial outcome prediction tasks—including earnings surprise, revenue growth, and profitability—our method consistently outperforms NER-based baselines with robust performance gains. By explicitly modeling semantic continuity and temporal shifts in KPI definitions, the framework establishes a new paradigm for text-based dynamic signal modeling in financial analytics.

Technology Category

• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Moving targets -- managers' strategic shifting of key performance metrics when the original targets become difficult to achieve -- have been shown to predict subsequent stock underperformance. However, our work reveals that the method employed in that study exhibits two key limitations that hinder the accuracy -- noise in the extracted targets and loss of contextual information -- both of which stem primarily from the use of a named entity recognition (NER). To address these two limitations, we propose an LLM-based target extraction} method with a newly defined metric that better captures semantic context. This approach preserves semantic context beyond simple entity recognition and yields consistently higher predictive power than the original approach. Overall, our approach enhances the granularity and accuracy of financial text-based performance prediction.

Problem

• Tracking evolving performance metrics in corporate disclosures
• Addressing noise and contextual loss in target extraction
• Enhancing financial text prediction accuracy using LLMs

Innovation

• LLM-based target extraction method
• Preserves semantic context beyond NER
• Higher predictive power than original approach