Paper: arXiv 2509.05107

Authors: Alfred Backhouse, Kang Li, Jakob Foerster, Anisoara Calinescu, Stefan Zohren

Abstract

Simulating limit order books (LOBs) has important applications across forecasting and backtesting for financial market data. However, deep generative models struggle in this context due to the high noise and complexity of the data. Previous work uses autoregressive models, although these experience error accumulation over longer-time sequences. We introduce a novel approach, converting LOB data into a structured image format, and applying diffusion models with inpainting to generate future LOB states. This method leverages spatio-temporal inductive biases in the order book and enables parallel generation of long sequences overcoming issues with error accumulation. We also publicly contribute to LOB-Bench, the industry benchmark for LOB generative models, to allow fair comparison between models using Level-2 and Level-3 order book data (with or without message level data respectively). We show that our model achieves state-of-the-art performance on LOB-Bench, despite using lower fidelity data as input. We also show that our method prioritises coherent global structures over local, high-fidelity details, providing significant improvements over existing methods on certain metrics. Overall, our method lays a strong foundation for future research into generative diffusion approaches to LOB modelling.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 9.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts from diffusion models and stochastic processes, and demonstrates high empirical rigor through rigorous benchmarking on LOB-Bench with Level-2 and Level-3 data, showcasing state-of-the-art performance.

Research Flowchart

  flowchart TD
  A["Research Goal: Generate & Forecast LOB Data<br>using Diffusion Models"] --> B["Data Preparation: LOB-Bench<br>Level-2 & Level-3 Data"]
  B --> C["Methodology: Transform LOB Data<br>into Structured Image Format"]
  C --> D["Spatio-temporal Diffusion Model<br>with Inpainting"]
  D --> E["Parallel Generation of<br>Long Sequences"]
  E --> F{"Evaluation on<br>LOB-Bench Metrics"}
  F --> G["Outcomes: SOTA Performance<br>Coherent Global Structures"]
  G --> H["Future Foundation for<br>Generative LOB Modelling"]

Cited References (62)

  1. Language models are few-shot learners (2020)
  2. Limit order books (2013)
  3. A Continuous-Time Measurement of the Buy-Sell Pressure in a Limit Order Book Market (2004) — link
  4. STRATEGIC INTERACTIONS AND NEGATIVE OIL PRICES (2021) — link
  5. Empirical properties of asset returns: stylized facts and statistical issues (2001) — link
  6. Hippo: Recurrent memory with optimal polynomial projections (2020)
  7. Efficiently Modeling Long Sequences with Structured State Spaces (2022)
  8. Simplified State Space Layers for Sequence Modeling (2023) — link
and 54 more
  1. Attention is all you need (2017)
  2. Analysis and Control of Nonlinear Process Systems (2004)
  3. Machine Learning — Lecture 17: Recurrent Neural Networks (2025) — link
  4. A Study Of N-Gram And Decision Tree Letter Language Modeling Methods (1998) — link
  5. Diffusion-lm improves controllable text generation (2022)
  6. CS294-158 Deep Unsupervised Learning, Spring 2024 (2024)
  7. Improving Image Generation with Better Captions (2023)
  8. Sora: Creating Video from Text (2024)
  9. Sora: A review on background, technology, limitations, and opportunities of large vision models (2024) — link
  10. Denoising diffusion probabilistic models (2020)
  11. Sdxl: Improving latent diffusion models for high-resolution image synthesis (2023) — link
  12. Diffedit: Diffusion-based semantic image editing with mask guidance (2022) — link
  13. Auto-Encoding Variational Bayes (2014)
  14. Failure modes of variational autoencoders and their effects on downstream tasks (2020) — link
  15. Generative adversarial nets (2014)
  16. Improved techniques for training gans (2016)
  17. Towards realistic market simulations: a generative adversarial networks approach (2021)
  18. Conditional generators for limit order book environments: Explainability, challenges, and robustness (2023)
  19. Generating synthetic data in finance: opportunities, challenges and pitfalls (2020)
  20. Generating Synthetic Data in Finance: Opportunities, Challenges and Pitfalls (2020) — link
  21. Unlocking the power of lstm for long term time series forecasting (2025)
  22. Enhancing time series momentum strategies using deep neural networks (2019) — link
  23. MarS: a Financial Market Simulation Engine Powered by Generative Foundation Model (2025)
  24. JAX-LOB: A GPU-Accelerated limit order book simulator to unlock large scale reinforcement learning for trading (2023)
  25. Generative ai for end-to-end limit order book modelling: A token-level autoregressive generative model of message flow using a deep state space network (2023)
  26. LOB-Bench: Benchmarking Generative AI for Finance – an Application to Limit Order Book Data (2025) — link
  27. TRADES: Generating Realistic Market Simulations with Diffusion Models (2025) — link
  28. Conditional Image Synthesis with Diffusion Models: A Survey (2024) — link
  29. High-resolution image synthesis with latent diffusion models (2022)
  30. Photorealistic text-to-image diffusion models with deep language understanding (2022)
  31. Image super-resolution via iterative refinement (2021) — link
  32. Diffusion models beat gans on image synthesis (2021)
  33. Learning transferable visual models from natural language supervision (2021)
  34. Repaint: Inpainting using denoising diffusion probabilistic models (2022)
  35. Sdedit: Guided image synthesis and editing with stochastic differential equations (2021) — link
  36. Deeplob: Deep convolutional neural networks for limit order books (2019)
  37. BDLOB: Bayesian Deep Convolutional Neural Networks for Limit Order Books (2018) — link
  38. The information content of an open limit‐order book (2008) — link
  39. Axial-lob: High-frequency trading with axial attention (2022)
  40. Deep inception networks: A general end-to-end framework for multi-asset quantitative strategies — → 2307.05522
  41. Multi-Horizon Forecasting for Limit Order Books: Novel Deep Learning Approaches and Hardware Acceleration using Intelligent Processing Units (2021) — link
  42. Gradient-based learning applied to document recognition (1998) — link
  43. Overfeat: Integrated recognition, localization and detection using convolutional networks (2014)
  44. Deep inside convolutional networks: Visualising image classification models and saliency maps (2013) — link
  45. Instance normalization: The missing ingredient for fast stylization (2016) — link
  46. Explaining ai in finance: Past, present, prospects — → 2306.02773
  47. Diffusers: State-of-the-art diffusion models (2022)
  48. The price impact of order book events: market orders, limit orders and cancellations (2012)
  49. ClusterLOB: Enhancing Trading Strategies by Clustering Orders in Limit Order Books — → 2504.20349
  50. LOBSTER: Limit Order Book Reconstruction System (2011) — link
  51. Learning to simulate realistic limit order book markets from data as a world agent (2022)
  52. Rwkv: Reinventing rnns for the transformer era (2023) — link
  53. Eagle and finch: Rwkv with matrix-valued states and dynamic recurrence (2024) — link
  54. A stochastic model for order book dynamics (2010)

3 of 62 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.