In brief4Cast Lab: An Open-Source App for Machine Learning and Statistical Analysis
Summary4Cast Lab is a free, open-source application for machine learning, econometrics and statistical analysis that runs on the user's own computer without installation or programming. Its Forecasting mode trains eight learning algorithms under leakage-safe validation for time-ordered data and compares selected models with alternatives and naive benchmarks via formal tests. Its Estimation mode covers applied econometrics from robust OLS to ARIMA, VAR/VECM and GARCH. The paper describes version 4.6.2 and reports validation of 1,234 numerical results against reference libraries and NIST certified values.
In briefGROMM: A risk-neutral option market model that fits European option prices without arbitrage.
SummaryThe paper presents a statically and dynamically arbitrage-free model of European option markets, built on the SANOS framework to smoothly interpolate option prices, and proves its dynamics have a unique strong solution. It serves as a stochastic local volatility representation of an option price surface rather than a pricing model. Tested on S&P Index option data from January 2022 to August 2025, the model's initial state fits within bid/ask for over 95% of relevant prices. The authors also propose a range of BHW tests for checking the absence of arbitrage.
In briefReinforcement learning with Hawkes-structured networks for market-making strategies
SummaryThis paper studies market making when order arrivals are clustered and self-exciting, modeled with non-Markovian Hawkes processes. The authors propose a model-free reinforcement learning framework and a Hawkes-Structured Network that encodes order flow history with a learnable mixture-of-exponentials filter. Experiments with several kernel types show the network earns higher trading rewards than general-purpose neural architectures and implicitly recovers order-flow intensities.
In briefNeural Estimation of Irreversibility in Market Order Book Paths
SummaryThe paper defines a time irreversibility rate for coarse-grained limit order book paths as the forward-to-reverse log-density ratio per event. It estimates this rate from three weeks of Nasdaq order data using order-invariant and order-sensitive neural networks. Real paths show more irreversibility than a surrogate preserving the one-step transition table, while a calibrated queue-reactive simulator matches the overall rate but shows almost no ordering-based increment. The authors propose the ordering and scale profile of path irreversibility as a benchmark for order book simulators.
In briefA theory model of household spending that compares with neighbors and resists cutting back, showing how group consumption shapes saving and investment decisions
SummaryThis paper studies how households choose consumption and investment when their well-being depends on spending relative to the population's geometric average, with spending cuts hurting more than increases help. The authors show that a mean field equilibrium exists under a weak interaction condition, and in a complete market with common risk it is an exact Nash equilibrium of the finite game. In a baseline example, ignoring inherited spending standards costs up to 17% of wealth, while small errors about the reference group cost less than 0.5%.
In briefExecuted and Recoverable Credit: How Action Order Shapes LLM Borrowing in a Simulated Lending Market
SummaryThis paper tests whether the order in which shared actions occur changes how much credit is actually executed in a simulated collateralized lending market with three small open-weight language models. An exact randomization test rejects the claim that common-market net debt is unaffected by complete-order assignment (two-sided p = 2/8,100). All 24 markets increased debt, and no market experienced liquidation or realized bad debt. The authors also report that reference-priced collateral narrows the observed order gap under fixed requests.
