---
title: Structured SDEs for Realized GARCH: Continuous-Time Volatility Modelling and Component-Level Analysis
url: https://www.ml-quant.com/papers/arxiv/2610.17732/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-10-16
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2610.17732
source_url: https://arxiv.org/abs/2610.17732
featured: 2026-10-16
citations: unknown
topic: Derivatives & Volatility
---


# Structured SDEs for Realized GARCH: Continuous-Time Volatility Modelling and Component-Level Analysis

A continuous-time volatility model incorporating realized measures forecasts one-day-ahead integrated variance, showing strong out-of-sample performance relative to GARCH and related benchmarks across 17 equity indices.

- Source: https://arxiv.org/abs/2610.17732
- Identifier: arXiv:2610.17732
- Released: 2026-10-14
- First featured: Quant Letter No. 135 (2026-10-16): https://www.ml-quant.com/issues/2026-10-16/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Derivatives & Volatility
- Authors: Hongyu Yao, Zhengyang Chi, Junbin Gao

## Abstract (arXiv, CC0)

We propose a structured continuous-time stochastic volatility model for one-day-ahead integrated-variance forecasting. The model introduces a jointly evolving realized-measure state that feeds directly into latent variance dynamics, while correlated Brownian innovations capture return-linked dependence. Parameters are estimated through simulation-based predictive learning, and Monte Carlo simulation forecasts integrated variance. Across 17 equity indices, the proposed model shows strong out-of-sample performance relative to GARCH, GARCH-Itô, Realized GARCH, HAR, and HEAVY benchmarks under QLIKE, MSE, and MAE. Ablation results show that realized-measure feedback provides substantial incremental forecasting value within the proposed framework, while return-linked Brownian dependence contributes more modest complementary gains. The results support incorporating measurement-driven information directly into continuous-time latent volatility dynamics to improve integrated-variance forecasting.

## Related

- [Pricing Multi-strike Quanto Call Options on Multiple Assets with Stochastic Volatility, Correlation, and Exchange Rates](https://www.ml-quant.com/papers/arxiv/2411.16617/): The study explores quanto options involving multiple assets in different currencies, concluding that a mix of GARCH-Jump SV, Weibull SC, and Ornstein Uhlenbeck (OU) SER is most effective for Monte Carlo simulation pricing.
- [Numerical analysis of a particle system for the calibrated Heston-type local stochastic volatility model](https://www.ml-quant.com/papers/arxiv/2504.14343/): The article presents a Monte Carlo method for simulating the Heston-type local stochastic volatility model, addressing drift and diffusion coefficient challenges and proving a strong chaos propagation under certain conditions.
- [Unified GARCH-Recurrent Neural Network in Financial Volatility Forecasting](https://www.ml-quant.com/papers/arxiv/2504.09380/): The paper proposes a new GARCH-GRU model for financial volatility forecasting, showing better computational efficiency and forecasting accuracy than other models.
- [Deep Reinforcement Learning Algorithms for Option Hedging](https://www.ml-quant.com/papers/arxiv/2504.05521/): A comparison of eight Deep Reinforcement Learning algorithms for dynamic hedging found that Monte Carlo Policy Gradient and Proximal Policy Optimization performed best, with the former outperforming the Black-Scholes delta hedge baseline.
- [Is the difference between deep hedging and delta hedging a statistical arbitrage?](https://www.ml-quant.com/papers/arxiv/2407.14736/): The research compares deep hedging and delta hedging in a GARCH-based market model, suggesting that the difference between the two can be a statistical arbitrage if the risk measure doesn't adequately consider negative outcomes.
- [A path-dependent PDE solver based on signature kernels](https://www.ml-quant.com/papers/arxiv/2403.11738/): The article introduces a new, verifiably effective kernel-based solver for path-dependent partial differential equations (PPDEs). This provides a practical alternative to Monte Carlo methods, especially for option pricing under rough volatility.