Research digest
Every issue of our research digest, as it was sent. The daily update leads with the three papers our screen found most interesting; the weekly digest opens with the week in numbers — the research funnel, the backtests and how the AI Quant portfolio did.
The paper builds leakage-aware classifiers to predict next-day firm-level stress and crash risk using technical price, volatility, and volume features. It does not present a tradable portfolio backtest, but the predictions can inform equity exposure reduction, position sizing, and risk-aware trading decisions.
The paper studies long-only portfolio allocation when cross-asset dependence is uncertain, using VaR, Expected Shortfall, Range-VaR, and volatility objectives. It is directly relevant to robust portfolio construction and proposes blending risk under a reference model with risk under worst-case dependence.
The paper develops a reinforcement-learning method for daily dynamic hedging of a high-dimensional basket option under time-consistent conditional CVaR objectives. Its numerical results are based on a simulated DCC-GARCH environment rather than a measured real-market trading backtest, but the hedging method can be tested on historical US-equity baskets with synthetic option liabilities.
The paper presents a practical framework for evaluating LLM-discovered trading strategies without look-ahead leakage and while accounting for the full number of searched candidates. It reports real-data backtests on US stocks and ETFs, finding that passive benchmarks pass its tests while all evaluated LLM-discovered strategies fail after honest out-of-sample and multiple-testing-aware evaluation.