HSHongyu Shen

Hongyu Shen

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Hongyu Shen has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

    • 19 November 2026
    • Published by MDPI AG in Preprints.org
      • https://doi.org/10.20944/preprints202611.2094.v1
    Open Access

    Evidence-Grounded Hybrid Agentic AI for Trustworthy AI Risk Communication Assessment in SEC Form 10-K Disclosures

    by Hongyu Shen, Xiaofan Guo, Lingyun Ke, Hang Xia, Runlin Jia, Yuncong Zuo, Yihan Shi

    Public companies increasingly communicate risks related to artificial intelligence (AI), machine learning, automation, algorithmic decision-making, data privacy, security, bias, and regulatory uncertainty in annual risk-factor disclosures. However, assessing whether these AI-risk and trustworthy-AI communication claims are specific, evidence-grounded, or largely boilerplate remains difficult because Form 10-K Item 1A disclosures combine forward-looking risk language, legal disclaimers, product descriptions, governance statements, and uncertain regulatory developments. This study develops an evidence-grounded hybrid agentic AI framework for assessing AI risk communication in SEC Form 10-K disclosures. The task is framed as textual support assessment rather than an independent determination of a company's actual AI risk exposure, legal compliance, product safety, or trustworthiness. Using a corpus of 160 Form 10-K filings collected from SEC EDGAR, we identify 101 filings with AI-related Item 1A risk-factor language and construct an 80-claim manually annotated gold sample across 48 companies. We compare five approaches: a transparent rule-based baseline, a TF-IDF retrieval baseline, a single-prompt large language model (LLM) baseline, a pure multi-agent reasoning baseline, and the proposed hybrid agentic adjudication method. The hybrid method achieves the strongest performance on trustworthiness-dimension classification and evidence-support assessment, with 0.725 dimension accuracy, 0.667 dimension macro-F1, 0.750 support accuracy, and 0.734 support macro-F1. We further add an explainability and auditability analysis using claim-level audit traces and a SHAP-based structured-feature surrogate for supported versus non-supported evidence labels. The findings suggest that evidence-grounded hybrid agentic AI can support auditable analysis of trustworthy-AI risk communication, while also showing that direct LLM prompting can be poorly calibrated for evidence-support judgments.

    • 19 November 2026
    • Published by MDPI AG in Preprints.org
      • https://doi.org/10.20944/preprints202611.2060.v1
    Open Access

    Hybrid Agentic AI Improves Evidence-Based Review of Cybersecurity Incident Disclosures

    by Hongyu Shen, Xiaofan Guo, Lingyun Ke, Hang Xia, Runlin Jia, Qiuyue Liao, Yihan Shi

    Cybersecurity incident disclosures are an important source of public-company risk information, particularly after the introduction of Item 1.05 of U.S. Securities and Exchange Commission Form 8-K for material cybersecurity incidents. A persistent analytical difficulty is whether a registrant’s materiality-related statement is supported by the facts disclosed in the same filing. This study develops a hybrid agentic artificial intelligence framework for evidence-based review of cybersecurity incident disclosures. The task is framed as disclosure-based textual support assessment, not as an independent legal judgment about materiality, liability, or compliance. Using 70 SEC Form 8-K Item 1.05 filings and a 30-filing expert-adjudicated pilot sample covering 20 incident groups, we compare five methods: rule-based classification, TF-IDF retrieval, direct large language model prompting, pure multi-agent reasoning, and hybrid agentic adjudication. The hybrid method combines structured disclosure indicators with role-specific agent outputs for claim identification, evidence extraction, uncertainty assessment, judgment, and critique. It achieves the strongest pilot results, with 0.733 materiality accuracy, 0.648 materiality macro-F1, 0.767 support accuracy, and 0.683 support macro-F1. Bootstrap and paired checks indicate that the clearest gains occur for support-label classification, while the small gold sample requires cautious interpretation. The framework also records evidence sentences, agent traces, decision-source information, and exploratory surrogate feature attributions. The findings suggest that hybrid agentic AI can make cybersecurity disclosure review more consistent and inspectable, while broader validation requires larger annotated samples and cross-sector testing.

  • Open Access

    AnExplainable AI Engineering Framework for Claims-Only First-Stage Provider Audit Triage Using SHAP-Guided Hybrid Retrieval-Augmented Generation

    by Danni Huang, Litong Song, Yue Chen, Shuangjiang He, Ruiqi Wang, Hongyu Shen, Weishen Chu

    This study proposes an explainable artificial intelligence (XAI) engineering workflow for provider-level healthcare claim audit prioritization using SHAP-guided hybrid retrieval-augmented generation (RAG). The framework integrates provider-level claim aggregation, tree-based risk screening, SHAP explanation, exploratory group-level SHAP clustering, policy concept retrieval, and constrained large language model audit narrative generation. Experiments on a public Medicare provider fraud dataset use the dataset-provided PotentialFraud label as a weak audit prioritization label rather than a legal determination of fraud. The results show that reimbursement exposure, utilization duration, claim repetition, deductible patterns, and beneficiary case mix contribute to provider-level risk scores. Additional cross-validation, calibration, threshold, and scale-confounding analyses indicate that provider size and financial exposure are important confounders, while non-scale and contextual features also retain predictive information. Beyond prediction, the framework organizes local SHAP drivers into exploratory provider archetypes and maps explanation patterns to audit-relevant policy concepts. Compared with pure embedding retrieval, the SHAP-guided hybrid retriever increases policy concept diversity and explanation alignment, although these retrieval metrics do not replace independent expert audit validation. Because the public dataset does not include referral pathways, inter-facility transfers, provider–network relationships, or care-coordination records, the framework cannot determine whether utilization patterns are explained by clinically appropriate referrals, regional access constraints, or multi-level care pathways. Its current applicability is therefore limited to provider-level audit prioritization using the available claims and beneficiary variables. The proposed system is positioned as a reproducible engineering prototype for cautious, human-reviewed audit support rather than a comprehensive or automated fraud determination system.

  • Open Access

    An Explainable AI Engineering Framework for Claims-Only First-Stage Provider Audit Triage Using SHAP-Guided Hybrid Retrieval-Augmented Generation

    by Danni Huang, Litong Song, Yue Chen, Shuangjiang He, Ruiqi Wang, Hongyu Shen, Weishen Chu

    This study proposes an explainable artificial intelligence (XAI) engineering workflow for provider-level healthcare claim audit prioritization using SHAP-guided hybrid retrieval-augmented generation (RAG). The framework integrates provider-level claim aggregation, tree-based risk screening, SHAP explanation, exploratory group-level SHAP clustering, policy concept retrieval, and constrained large language model audit narrative generation. Experiments on a public Medicare provider fraud dataset use the dataset-provided PotentialFraud label as a weak audit prioritization label rather than a legal determination of fraud. The results show that reimbursement exposure, utilization duration, claim repetition, deductible patterns, and beneficiary case mix contribute to provider-level risk scores. Additional cross-validation, calibration, threshold, and scale-confounding analyses indicate that provider size and financial exposure are important confounders, while non-scale and contextual features also retain predictive information. Beyond prediction, the framework organizes local SHAP drivers into exploratory provider archetypes and maps explanation patterns to audit-relevant policy concepts. Compared with pure embedding retrieval, the SHAP-guided hybrid retriever increases policy concept diversity and explanation alignment, although these retrieval metrics do not replace independent expert audit validation. Because the public dataset does not include referral pathways, inter-facility transfers, provider-network relationships, or care-coordination records, the framework cannot determine whether utilization patterns are explained by clinically appropriate referrals, regional access constraints, or multi-level care pathways. Its current applicability is therefore limited to provider-level audit prioritization using the available claims and beneficiary variables. The proposed system is positioned as a reproducible engineering prototype for cautious, human-reviewed audit support rather than a comprehensive or automated fraud determination system.

  • Open Access

    Modeling Structural Deviation in 10-K Risk Factors: A Semantic Anomaly Detection and Explainable AI Approach

    by Fang Sun, Shuangjiang He, Ruiqi Wang, Lingyun Ke, Hongyu Shen, Qiuyue Liao

    This study presents an exploratory methodological framework for examining structural changes in regulatory risk disclosure using sentence embeddings, multivariate anomaly detection, and explainable artificial intelligence. Prior research typically relies on dictionary-based word frequencies, tone indicators, or topic proportions to quantify risk disclosure. While these measures capture disclosure intensity, they do not directly assess whether the internal semantic organization of risk narratives has shifted relative to historical patterns. We propose a structural semantic deviation framework that represents each company–year disclosure using thematic shares and embedding-based dispersion statistics and evaluates deviations from a historical baseline through unsupervised anomaly detection. Using Item 1A Risk Factors from Wells Fargo and JPMorgan Chase surrounding the 2016 regulatory shock as a focused two-firm case study, we show that traditional lexical metrics do not clearly isolate structural breaks, whereas embedding-based semantic trajectories reveal substantial narrative reconfiguration. Isolation-based modeling provides stable and discriminative anomaly scores in this setting, and SHAP decomposition highlights semantic distance, litigation emphasis, and disclosure contraction as important drivers of deviation in 2025 out-of-sample disclosures. These findings should be interpreted as methodological evidence rather than broad population-level claims. The study demonstrates how structural semantic modeling can be operationalized in regulatory disclosure analysis and provides a transparent framework that can be extended to larger panels and cross-industry settings in future research.


  • Open Access

    Modeling Structural Deviation in 10-K Risk Factors: A Semantic Anomaly Detection and Explainable AI Approach

    by Fang Sun, Shuangjiang He, Ruiqi Wang, Lingyun Ke, Hongyu Shen, Qiuyue Liao

    This study examines structural changes in regulatory risk disclosure using a semantic modeling framework that integrates sentence embeddings, multivariate anomaly detection, and explainable artificial intelligence. Prior research typically relies on dictionary-based word frequencies, tone indicators, or topic proportions to quantify risk disclosure. While these measures capture disclosure intensity, they do not directly assess whether the internal semantic organization of risk narratives has shifted relative to historical patterns. We propose a structural semantic deviation framework that represents each company-year disclosure using thematic shares and embedding-based dispersion statistics, and evaluates deviations from a historical baseline through unsupervised anomaly detection. Using Item 1A Risk Factors from Wells Fargo and JPMorgan Chase surrounding the 2016 regulatory shock, we demonstrate that traditional lexical metrics fail to isolate structural breaks, whereas embedding-based semantic trajectories reveal substantial narrative reconfiguration. Isolation-based modeling provides stable and discriminative anomaly scores, and SHAP decomposition identifies semantic distance, litigation emphasis, and disclosure contraction as key drivers of deviation in 2025 out-of-sample disclosures. The results suggest that structural semantic modeling captures risk narrative transformation beyond word accumulation, offering an interpretable and scalable framework for regulatory risk assessment.