Towards Financial World Modeling

2026-10-06

AI Summary

This study addresses the limitation that existing evaluations of financial representation learning are confined to single tasks and narrow datasets, thereby hindering their utility for world model-based decision-making. To overcome this, we construct Market-1T, a trillion-scale high-frequency market dataset, alongside a rigorous cross-cycle scaling evaluation protocol. Through extensive comparative experiments, latent space probing analyses, and diverse encoder training strategies—including architectures such as DINO-WM—we systematically reveal underlying structural discrepancies among encoders that exhibit comparable predictive performance. Ultimately, this work establishes a foundational paradigm for training and evaluating financial world models, providing a robust representation framework for inferring market states under previously unseen tasks.

Abstract

Building a world model requires a state representation useful for planning and decision-making---potentially over tasks unknown at training time. In the context of financial markets, planning and decision-making may require a model to reason about market-wide conditions, asset-specific expected returns, liquidity, volatility, and cross-asset relationships. Yet financial representation learning has largely been evaluated on individual predictive tasks, oftentimes on a single time period using comparatively narrow datasets. We address this through three primary contributions. First, we introduce Market-1T, a dataset containing nearly one trillion observations across U.S. equities from 2008 to 2025 at 1 Hz resolution. Second, we develop and implement a rigorous evaluation protocol. Third, we conduct a systematic large-scale study of financial representation learning, comparing 18 encoder-training strategies across nearly two decades of market regimes. We evaluate learned representations both by their predictive utility on common finance tasks and through probes of latent structure. We find that encoders with similar predictive performance can organize market state very differently. Collectively, we establish a foundation for training and evaluating financial market representations in support of world models such as DINO-WM, V-JEPA 2, and LeWM.

Problem

• Financial World Model
• Representation Learning
• State Representation
• Market Modeling
• Evaluation Protocol

Innovation

• Financial World Model
• Representation Learning
• Market-1T Dataset
• Evaluation Protocol
• Latent Structure Probing