Paper: arXiv 2506.23767

Authors: Xue Wen Tan, Stanley Kok

Abstract

Every publicly traded U.S. company files an annual 10-K report containing critical insights into financial health and risk. We propose Tiny eXplainable Risk Assessor (TinyXRA), a lightweight and explainable transformer-based model that automatically assesses company risk from these reports. Unlike prior work that relies solely on the standard deviation of excess returns (adjusted for the Fama-French model), which indiscriminately penalizes both upside and downside risk, TinyXRA incorporates skewness, kurtosis, and the Sortino ratio for more comprehensive risk assessment. We leverage TinyBERT as our encoder to efficiently process lengthy financial documents, coupled with a novel dynamic, attention-based word cloud mechanism that provides intuitive risk visualization while filtering irrelevant terms. This lightweight design ensures scalable deployment across diverse computing environments with real-time processing capabilities for thousands of financial documents which is essential for production systems with constrained computational resources. We employ triplet loss for risk quartile classification, improving over pairwise loss approaches in existing literature by capturing both the direction and magnitude of risk differences. Our TinyXRA achieves state-of-the-art predictive accuracy across seven test years on a dataset spanning 2013-2024, while providing transparent and interpretable risk assessments. We conduct comprehensive ablation studies to evaluate our contributions and assess model explanations both quantitatively by systematically removing highly attended words and sentences, and qualitatively by examining explanation coherence. The paper concludes with findings, practical implications, limitations, and future research directions. Our code is available at https://github.com/Chen-XueWen/TinyXRA.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper introduces a highly practical, lightweight NLP model (TinyXRA) for evaluating 10-K report risks using downside risk metrics like the Sortino ratio alongside higher moments. It features exceptional empirical validation across a multi-year dataset (2013-2024), quantitative and qualitative XAI evaluations, and fully open-sourced code.

Research Flowchart

  flowchart TD
    A[Research Goal: Explainable, Lightweight Risk Assessment for 10-K Filings] --> B[Data Input: 10-K Reports 2013-2024 & Multi-Metric Risk Labels]
    B --> C[TinyBERT Encoder: Process Lengthy Financial Documents]
    C --> D[Model Training: Triplet Loss Optimization for Risk Quartiles]
    D --> E[XAI Mechanism: Dynamic Attention-Based Word Clouds]
    E --> F[Outcomes: SOTA Predictive Accuracy & Real-Time Production Scalability]