MENTOR: a multi-agent framework for event and narrative trend prediction with optimized reasoning

Abstract

Narrative economics suggests that financial markets are strongly influenced by evolving narratives, creating opportunities for forecasting emerging events and their economic impacts. However, existing large language model (LLM)-based approaches are inadequate in terms of systematic task decomposition and alignment with financial applications. We propose MENTOR, a multi-agent framework for event and narrative trend prediction that integrates teacher–student iterative reasoning with progressive subtasks: detecting and ranking trending events, forecasting future events from current narratives, and predicting industry index performance influenced by these events. Experiments on our self-constructed Chinese key opinion leader (KOL) articles dataset and English financial news dataset show that MENTOR consistently outperforms recent baselines such as the stakeholder-enhanced future event prediction (StkFEP) and summarize–explain–predict (SEP) frameworks in both event prediction and industry ranking tasks. In addition, the backtest results at the portfolio level show that improved event and industry forecasts can bring about a practical improvement in investment performance. These results demonstrate that incorporating structured reasoning and multi-agent feedback enables more reliable event forecasting and strengthens the connection between narrative dynamics and financial market outcomes.

1. Introduction

Narrative economics emphasizes that financial markets are driven not solely by fundamentals but also by the narratives that circulate among investors and the public (Shiller, 2019). Narratives shape expectations, guide sentiment, and can amplify or mitigate economic fluctuations. In practice, analysts often anticipate emerging narratives and then map them to likely sectoral or market responses. This decomposition, predicting events first and then inferring financial impacts, offers a structured way to avoid modeling the randomness inherent in price movements. However, automating this reasoning process remains an open challenge.

Although narratives can spread in complex and sometimes unpredictable ways, Robert Shiller's theory of narrative economics provides a foundation for structured forecasting (Shiller, 2019). Specifically, Shiller argues that (1) new narratives typically emerge from existing ones through association or amplification, (2) economically relevant narratives tend to recur across different historical episodes (e.g., inflation fears and technological disruption), and (3) despite surface-level diversity, narratives can be meaningfully clustered by theme and market impact. These properties imply a degree of bounded predictability: although we cannot foresee every narrative shock, we can identify high-potential trajectories rooted in current discourse. As illustrated in Fig. 1, our framework operationalizes this insight by analyzing the narrative constellation of today to forecast future events and their financial implications.