LOB-Impact Benchmark: Emergent Market Impact in Generative Limit Order Book Models

Key points

• Here, we show that market impact emerges naturally in neural LOB generators trained solely for next-token prediction.1
• We inject execution-style metaorders into the generated message flow of five autoregressive models across five NASDAQ stocks and measure the response using estimators from the empirical impact literature, validated on baselines whose physics is known in advance.1
• Interestingly, we observe emergent impact in generative models of market data, with exponents in the range of 0.63 to 1.21, slightly above the traditional 0.5 of the square-root law of impact.1
• We release the benchmark as open source to validate future generative models for market data.1

Sources (1)

• [1]LOB-Impact Benchmark: Emergent Market Impact in Generative Limit Order Book Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 06:01 PM

  Realistic market impact is essential for generative limit order book (LOB) models to serve as world models for trading agents.

  Here, we show that market impact emerges naturally in neural LOB generators trained solely for next-token prediction.

Extractive summary: sentences quoted from the sources.