arXiv subjects

Xinyue Fang

Publications and source records attributed to Xinyue Fang.

3 recordsLinked to original sources

arXiv2026

Volatility Forecasting and Return Prediction under Market Regimes: Evidence from High-Frequency Chinese Equity Data

This study investigates whether regime-dependent volatility forecasting and machine-learning-based return prediction can be jointly integrated to improve both statistical forecasting performance and economic strategy outcomes in equity markets. Using high-frequency CSI 300 Index data from 2005 to 2023, a sequential twostage framework is developed. In the first stage, realized volatility is modeled using regime-augmented HARQ specifications combined with Markov-switching GJR-GARCH filtering to capture long-memory dynamics, asymmetry, and structural market regimes. In the second stage, volatility forecasts, regime indicators, and return-related predictors are incorporated into an XGBoost return-prediction model estimated through a strictly walk-forward out-of-sample procedure. The empirical results demonstrate that regime-aware volatility forecasting consistently outperforms baseline HARQ models across forecast evaluation metrics and is generally supported by formal forecast comparison tests. In contrast, return predictability remains weak, state-dependent, and concentrated primarily in low-volatility regimes. Although naive predictive trading strategies generally fail after accounting for realistic transaction costs, carefully designed implementations incorporating volatility scaling, low-volatility gating, threshold calibration, and turnover controls can improve defensive economic performance. The findings suggest that the practical value of predictive systems in financial markets may depend less on generating strong unconditional return forecasts and more on transforming weak state-dependent signals into economically robust portfolio allocation rules. Overall, the study contributes by integrating econometric volatility modeling, regime classification, machine-learning return prediction, and implementation realism within a unified framework.

q-fin.TR↗

arXiv2026

Knowledge Injection Exists in MoE? Exploring Expert-Aware Contrast Decoding in MoE for Mitigating LLMs'Hallucinations

Existing LLM hallucination mitigation methods, including prompt engineering and model optimization, either hardly alter models'internal knowledge or have poor cross-domain generalization. Contrastive decoding mitigates hallucinations by using layer-wise differences in LLMs. However, prior studies only explore transformer-based models (e.g., GPT), ignoring other effective frameworks like mixture-of-experts (MoE) models. Since MoE alters the traditional transformer architecture, we conduct empirical studies to investigate whether similar layer-wise differences exist in MoEs. Our results show that they do not exist in MoE with shared experts; nevertheless, across different MoEs, higher layers exhibit distinct expert activation patterns between factual and non-factual outputs. Building on these, we propose EAACD, an expert-aware adaptive contrast decoding that uses expert differences in MoE's higher layers to mitigate hallucinations on QA tasks. EAACD splits high-layer experts into a higher-reliability group and several lower-reliability groups based on their confidence and consistency. It contrasts the higher-reliability group's prediction with each lower-reliability group's prediction to calibrate the model's original predictions. To strengthen this contrast, EAACD amplifies hallucinations from lower-reliability experts via attention and masking to provide stronger negative references. EAACD outperforms all baselines on four datasets.

cs.CL↗

arXiv2024

Zero-resource Hallucination Detection for Text Generation via Graph-based Contextual Knowledge Triples Modeling

LLMs obtain remarkable performance but suffer from hallucinations. Most research on detecting hallucination focuses on the questions with short and concrete correct answers that are easy to check the faithfulness. Hallucination detections for text generation with open-ended answers are more challenging. Some researchers use external knowledge to detect hallucinations in generated texts, but external resources for specific scenarios are hard to access. Recent studies on detecting hallucinations in long text without external resources conduct consistency comparison among multiple sampled outputs. To handle long texts, researchers split long texts into multiple facts and individually compare the consistency of each pairs of facts. However, these methods (1) hardly achieve alignment among multiple facts; (2) overlook dependencies between multiple contextual facts. In this paper, we propose a graph-based context-aware (GCA) hallucination detection for text generations, which aligns knowledge facts and considers the dependencies between contextual knowledge triples in consistency comparison. Particularly, to align multiple facts, we conduct a triple-oriented response segmentation to extract multiple knowledge triples. To model dependencies among contextual knowledge triple (facts), we construct contextual triple into a graph and enhance triples' interactions via message passing and aggregating via RGCN. To avoid the omission of knowledge triples in long text, we conduct a LLM-based reverse verification via reconstructing the knowledge triples. Experiments show that our model enhances hallucination detection and excels all baselines.

cs.CL↗