Benchmark finds emergent market impact in generative LOB models
Original title: LOB-Impact Benchmark: Emergent Market Impact in Generative Limit Order Book Models
Five autoregressive generators showed impact exponents of 0.63 to 1.21 on five NASDAQ stocks, above the square-root law.
The LOB-Impact Benchmark tests whether neural limit-order-book generators trained only for next-token prediction produce realistic market impact. Researchers inject execution-style metaorders into five autoregressive models across five NASDAQ stocks and compare visible injections with a blind simulator-only control. Measured exponents ranged from 0.63 to 1.21, slightly above the traditional 0.5 square-root law. The authors did not observe impact decay during relaxation and release the benchmark as open source.
Unconfirmed: The authors tentatively attribute missing impact decay to short context windows.
- Phases cover free generation, metaorder injection, and post-execution relaxation.
- Impact is measured for every tested model on every stock.
- Authors could not observe impact decay in the relaxation window.
- The benchmark is released as open source.
- Organizations
- NASDAQ
From the original article
1 min read at huggingface.co
Realistic market impact is essential for generative limit order book (LOB) models to serve as world models for trading agents. If a simulator misprices this mechanism, agents learn policies that are optimal in the simulator but wrong in the real market.
By George Nigmatulin, Valentin Mohl, Sascha Frey · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.