Do LLMs Understand Limit Order Book Dynamics?

📰 ArXiv cs.AI

LLMs can generate valid limit order book event sequences but fail to understand underlying dynamics, leading to biased estimates and spurious predictability

advanced Published 26 Aug 2026

Action Steps

Train an LLM on synthetic limit order book data to generate valid event sequences
Evaluate the LLM's implicit world model using novel tests to identify potential biases
Assess the LLM's predictability in forecasting future limit order book events
Compare the LLM's performance with traditional models to determine its effectiveness
Refine the LLM's training data and architecture to improve its understanding of limit order book dynamics

Who Needs to Know This

Quantitative traders and researchers working with limit order book data can benefit from understanding the limitations of LLMs in this context, as it affects the accuracy of their predictions and trading decisions

Key Insight

💡 LLMs can be biased and predictably incorrect when forecasting limit order book events due to their limited understanding of the underlying dynamics

Key Takeaways

LLMs can generate valid limit order book event sequences but fail to understand underlying dynamics, leading to biased estimates and spurious predictability

Full Article

Title: Do LLMs Understand Limit Order Book Dynamics?

Abstract:
arXiv:2608.23706v1 Announce Type: new Abstract: A large language model (LLM) trained on synthetic limit order book (LOB) data achieves near perfect scores in generating valid sequences of LOB events. However, the LLM's implicit world model fails to learn the state of the LOB. This deficiency leads to biased estimates and spurious predictability in using the LLM to forecast future LOB events. Our analysis uses novel tests of an LLM's world model, extending prior work from deterministic settings t

Read full paper → ☆ Save to playlist ← Back to Reads