LabelFusion-TS: Fusing Large Language Models, Transformer Encoders, and Financial Time Series for Monetary-Policy Stance Classification

2026-08-12

AI Summary

This work addresses the limitations of existing financial text classification methods, which often neglect market context and struggle to accurately discern hawkish, dovish, or neutral stances in Federal Reserve communications. To overcome this, the authors propose LabelFusion-TS, a novel system that, for the first time, incorporates financial market time series as an auxiliary modality. The approach fuses a fine-tuned RoBERTa model, prompt-driven large language models, and a time series Transformer, employing a two-stage training strategy to mitigate the scarcity of labeled data. Evaluated on a test set spanning 2015–2022, the model achieves a weighted F1 score of 70.2% using only 240 manually annotated samples—significantly outperforming zero-shot large language models (64.1%)—thereby demonstrating the efficacy of multimodal fusion and few-shot learning in this domain.

Abstract

Financial text is produced and interpreted within a market environment, yet financial text classifiers almost always receive text alone. We study whether financial time series are useful as an additional input on the task of classifying sentences from Federal Reserve communication as hawkish, dovish, or neutral. Our system, \lfts{}, extends the \lf{} architecture with this modality: a small voting network combines three independently trained components, a fine-tuned RoBERTa encoder, a prompted large language model (LLM), and a fused ensemble of time-series transformers over the market series of the months preceding publication. Because only about a thousand annotated sentences are available for training, the RoBERTa encoder is first pre-trained on sentences annotated automatically by the LLM and only then fine-tuned on the human labels. Trained on Federal Open Market Committee (FOMC) communication up to 2015 and evaluated on 2015--2022, the fused system achieves 70.2\% weighted F1 -- against 64.1\% for the zero-shot LLM -- and overtakes it with as few as 240 human-labelled sentences. We take this as initial evidence for market time series as an input modality in financial text classification.

Problem

• financial text classification
• monetary-policy stance
• time series
• multimodal fusion
• FOMC communication

Innovation

• multimodal fusion
• financial time series
• large language models
• monetary-policy stance classification
• few-shot learning