Weekly Quantitative Paper Digest (Feb 14‑Feb 20, 2026)

Factor Engine: A Python Library for Systematic Financial Factor Computation and Analysis

Factor Engine is an open‑source, high‑performance Python library that provides a modular and extensible API for defining custom financial factors. The API uses Python decorators so that users can declare factor functions concisely and integrate them with data‑science tools such as pandas and NumPy. To evaluate the library, the authors compute pricing factors with Factor Engine and compare the results to a reference implementation written in Stata. Back‑test outcomes (e.g., factor returns, t‑statistics, and information ratios) are highly similar between the two implementations, and runtime performance is comparable. The authors also embed the computed factors into a machine‑learning pipeline—training a regression model to predict future returns—demonstrating that the library can be used directly in quantitative‑finance research workflows.

FactorMiner: A Self‑Evolving Agent with Skills and Experience Memory for Financial Alpha Discovery

FactorMiner is a lightweight, flexible framework that treats alpha‑factor discovery as an iterative search guided by accumulated knowledge. The system consists of:

Modular skill architecture : each skill encapsulates a systematic financial evaluation tool (e.g., factor generation, back‑testing, performance scoring) as an executable component.

Experience memory : a structured repository that records the outcomes of past mining experiments, distilling successful patterns and failure constraints.

FactorMiner operationalizes the Ralph loop—Retrieve, Generate, Evaluate, Distill—by retrieving priors from memory, generating new factor candidates, evaluating them on historical data, and distilling the results back into memory to bias subsequent searches. Experiments on multiple asset classes and markets show that the agent builds a diverse library of high‑quality factors while keeping redundancy low as the library expands. The authors report competitive predictive performance (e.g., out‑of‑sample Sharpe ratios) and argue that the approach scales under a “relevance red‑sea” constraint, where the search space is saturated with redundant signals.

Market Efficiency and the Heterogeneous Impact of Financial Liberalization: Evidence from the Shanghai‑Hong Kong Stock Connect

The study investigates how the Shanghai‑Hong Kong Stock Connect (SHHK) affected A‑H share price premiums and whether the effect depends on market efficiency. Using monthly data from January 2011 to May 2019 for 67 dual‑listed firms, the authors estimate a two‑step system GMM dynamic panel model to address premium persistence and potential endogeneity. Market efficiency is proxied by the daily high‑low price range, interpreted as a measure of trading friction.

Key findings:

Implementation of SHHK is associated with an average 18.4 % increase in the A‑H premium.

The premium effect is heterogeneous: firms operating in markets with higher trading friction (lower efficiency) experience a larger marginal increase, whereas firms in more efficient markets see a weaker effect.

No statistically significant price reaction is observed during the policy announcement period.

Robustness checks—including placebo tests and alternative efficiency proxies—confirm the efficiency‑dependent impact. The results highlight that pre‑existing information frictions shape the outcomes of financial liberalization policies.