Retrieval-Augmented Diffusion Modeling for Stochastic Discount Factor Portfolios

2026-09-28

AI Summary

This study addresses the portfolio optimization challenges arising from non-stationary market dynamics and multimodal noise under stochastic discount factor frameworks. To this end, we propose RADAR, a novel framework that transcends conventional isotropic Gaussian noise assumptions. Specifically, RADAR introduces a retrieval-augmented mechanism to match analogous historical states, thereby constructing context-dependent noise distributions, and leverages conditional diffusion models to denoise and learn latent market state representations. By integrating retrieval-augmented generation, conditional diffusion modeling, and empirical statistical initialization techniques, the proposed framework achieves state-of-the-art performance across key risk-adjusted metrics. Furthermore, RADAR generates economically meaningful signals for asset returns and correlations, demonstrating its practical utility in financial decision-making.

Abstract

In this work, we study portfolio optimization under the stochastic discount factor (SDF) framework by learning market state representations that capture the underlying risk structures of financial data. This is challenging due to several factors: financial markets exhibit non-stationary dynamics with shifting regimes, multimodal inputs such as price and news data often contain stochastic noise, and existing diffusion-based approaches, while effective for modeling stochastic dynamics, rely on assumptions such as isotropic Gaussian noise that fail to capture the state-dependent nature of financial uncertainty. To address these challenges, we introduce RADAR, a retrieval-augmented diffusion framework that learns market representations by conditioning on similar historical regimes. RADAR leverages retrieval to construct context-dependent noise distributions, applies conditional diffusion to denoise multimodal representations, and initializes the diffusion process using empirical statistics to reflect state-dependent uncertainty. Experiments show that RADAR achieves state-of-the-art performance on key risk-adjusted metrics while producing economically meaningful signals on asset returns and correlations.

Problem

• Portfolio Optimization
• Stochastic Discount Factor
• Non-stationary Dynamics
• Multimodal Inputs
• State-dependent Uncertainty

Innovation

• Retrieval-Augmented Diffusion
• Stochastic Discount Factor
• Portfolio Optimization
• Multimodal Denoising
• State-Dependent Uncertainty