AIQF Weekly digest — 28 new papers

The week in numbers

Max drawdown -2.4%, -7.0% at market β · 10 strategies live · as of 2026-09-11

At market β restates the record at the market's beta — the same strategies held at 3.70× the exposure: the return is multiplied by that factor, the drawdown re-measured on the scaled line, and Sharpe is unchanged by it. Arithmetic on the record, not a second track record — no leverage was used and no financing cost is modelled.

Quant Paper Radar · 5 of 28

1. 1

   Simple Dynamic Stock/Bond/Gold Portfolios

   Nikhil Devanathan, Alexandros E. Tzikas, Stephen P. Boyd · ARXIV

   we backtested it — Sharpe 0.43 · quant relevance 1.00

   The paper evaluates monthly, long-only dynamic allocation among SPY, AGG, GLD, and cash, using volatility targeting and constrained Markowitz optimization. It reports improved Sharpe ratios, drawdowns, and risk consistency relative to fixed-weight stock/bond and stock/bond/gold benchmarks. The platform can implement the ETF and FRED-based methodology, although its price history supports replication from approximately 2010 rather than the paper's full 2006--2026 evaluation window.

   original paper

2. 2

   Entropic Value-at-Risk parity for tempered stable returns

   Jaehyung Choi · ARXIV

   we backtested it — Sharpe 0.37 · quant relevance 0.98

   The paper constructs long-only inverse-risk-parity and equal-risk-contribution portfolios using Entropic Value-at-Risk under tempered-stable return models. Across ETF and portfolio universes, it reports positive Sharpe-ratio differences for EVaR-ERC portfolios relative to equal-weight and Gaussian-EVaR benchmarks.

   original paper

3. 3

   A hybrid deep learning framework for volatility prediction in financial markets

   Sanjay Agal · OPENALEX

   we backtested it — Sharpe 0.64 · quant relevance 0.96

   The paper forecasts market volatility for risk management, VaR, options applications, and an explicitly reported trading strategy. A closely related volatility-targeting implementation using daily and one-minute US equity and ETF data is backtestable, but the full multi-asset sample cannot be reproduced because the required FX data and long history are unavailable.

   original paper

4. 4

   Nyström Attention Matches Full Attention for Cross-Sectional Stock Prediction

   Kunhan Guo · ARXIV

   we backtested it — Sharpe -0.58 · quant relevance 0.96

   The paper tests a low-rank Nyström approximation to cross-sectional Transformer attention for predicting and ranking future stock returns. On Chinese A-share universes, it reports that the method preserves the Rank IC of full attention while reducing attention complexity from O(N²) to O(mN), though gains from the cross-stock module did not reliably extend to the largest universe.

   original paper

5. 5

   AlphaRJM: Reward-Jump Memory for Stochastic Return-Guided Alpha Discovery

   Sayan Dhan, Selvaraju Natarajan · ARXIV

   quant relevance 0.96

   AlphaRJM uses reinforcement learning to discover interpretable formulaic equity alpha signals, with rewards based on incremental improvements in ensemble information coefficient. The paper reports gains and stability across equity universes, forecast horizons, and random seeds.

   original paper

28 new papers written up in this period. Open the Quant Paper Radar →