Research

Paper

AI LLM February 24, 2026

FinAnchor: Aligned Multi-Model Representations for Financial Prediction

Authors

Zirui He, Huopu Zhang, Yanguang Liu, Sirui Wu, Mengnan Du

Abstract

Financial prediction from long documents involves significant challenges, as actionable signals are often sparse and obscured by noise, and the optimal LLM for generating embeddings varies across tasks and time periods. In this paper, we propose FinAnchor(Financial Anchored Representations), a lightweight framework that integrates embeddings from multiple LLMs without fine-tuning the underlying models. FinAnchor addresses the incompatibility of feature spaces by selecting an anchor embedding space and learning linear mappings to align representations from other models into this anchor. These aligned features are then aggregated to form a unified representation for downstream prediction. Across multiple financial NLP tasks, FinAnchor consistently outperforms strong single-model baselines and standard ensemble methods, demonstrating the effectiveness of anchoring heterogeneous representations for robust financial prediction.

Metadata

arXiv ID: 2602.20859
Provider: ARXIV
Primary Category: cs.CL
Published: 2026-02-24
Fetched: 2026-02-25 06:05

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.20859v1</id>\n    <title>FinAnchor: Aligned Multi-Model Representations for Financial Prediction</title>\n    <updated>2026-02-24T13:02:09Z</updated>\n    <link href='https://arxiv.org/abs/2602.20859v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.20859v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Financial prediction from long documents involves significant challenges, as actionable signals are often sparse and obscured by noise, and the optimal LLM for generating embeddings varies across tasks and time periods. In this paper, we propose FinAnchor(Financial Anchored Representations), a lightweight framework that integrates embeddings from multiple LLMs without fine-tuning the underlying models. FinAnchor addresses the incompatibility of feature spaces by selecting an anchor embedding space and learning linear mappings to align representations from other models into this anchor. These aligned features are then aggregated to form a unified representation for downstream prediction. Across multiple financial NLP tasks, FinAnchor consistently outperforms strong single-model baselines and standard ensemble methods, demonstrating the effectiveness of anchoring heterogeneous representations for robust financial prediction.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-02-24T13:02:09Z</published>\n    <arxiv:comment>11 pages, 4 figures, 5 tables</arxiv:comment>\n    <arxiv:primary_category term='cs.CL'/>\n    <author>\n      <name>Zirui He</name>\n    </author>\n    <author>\n      <name>Huopu Zhang</name>\n    </author>\n    <author>\n      <name>Yanguang Liu</name>\n    </author>\n    <author>\n      <name>Sirui Wu</name>\n    </author>\n    <author>\n      <name>Mengnan Du</name>\n    </author>\n  </entry>"
}