Quant LetterNo. 42
74 items across 8 sections, as sent to readers on 27 March 2024. Paper titles open their ML-Quant page; ↗ goes to the source.
Quantitative-finance and ML-for-finance preprints from arXiv.
13 items
The rHW model has been introduced to enhance Valuation Adjustments calculations by capturing market-implied skew and smile, significantly impacting interest rate derivatives' exposures and xVAs.
6 shares1 citation todaySource ↗
A study has created a standard for assessing large language models in Japanese and finance, with GPT-4 showing excellent performance.
7 shares23 citations todaySource ↗
A novel credit rating migration model has been developed, incorporating economic state changes and utilizing Markov theory for various rating philosophies analysis.
6 shares3 citations todaySource ↗
The study uses a generative adversarial network (GAN) to solve utility optimization problems in market settings, performing as well as optimal strategies and better in settings without a known optimal strategy.
5 shares2 citations todaySource ↗
The research applies Merton's optimal portfolio problem to a stock market with potential bankruptcy, creating a new version of Merton's ratio using Bellman's principle and validating it with a verification theorem.
3 sharesSource ↗
The paper introduces new mathematical methods to detect temporal shifts in US equities, investigates nonlinear shifts in market sector structure, studies network structure in new market sectors, and conducts sampling experiments across various sample spaces.
2 shares2 citations todaySource ↗
A study on AI's impact on freelancers revealed that exposure to language modeling technologies led to more job applications and increased specialization post the introduction of ChatGPT.
2 shares1 citation todaySource ↗
A football match data study showed that coworker performance significantly influences individual performance evaluations and career advancements, with positive performances creating beneficial effects.
2 sharesSource ↗
An experiment comparing ranking algorithms for online marketplace listings found that linear programming-based algorithms improved all key metrics, including revenue and purchase rates.
2 shares1 citation todaySource ↗
Football teams' blockchain fan tokens increase in value before the World Cup but drop during matches, especially if the team loses.
15 shares15 citations todaySource ↗
A new model using fractional stochastic volatility helps manage cryptocurrency exchange rate risk, highlighting the need for price and volatility jumps in the crypto market.
10 shares5 citations todaySource ↗
The growth of blockchain technology and crypto assets presents regulatory and tax challenges, prompting discussions on crypto principles, tax issues, and policy responses.
7 shares4 citations todaySource ↗
A proposed DAO governance model uses a sequential auction mechanism to address control issues, offering a robust solution to empty voting and varied regulations.
2 shares1 citation todaySource ↗
Working papers in finance and economics from SSRN.
25 items
Derman et. al.'s concept is improved to create a static minimum variance hedge for volatility swaps using variance swaps, with numerical examples showing the hedge's effectiveness.
85 sharesSource ↗
The paper shows that the bias in the two-stage least squares estimator can be split into an observable and unobservable bias, without needing to specify the first stage's functional form or validate the instrumental variable.
3 sharesSource ↗
Machine learning reveals an N-shaped pattern in corporate credit rating standards from 1986 to 2016, with the Gradient Boosting Machine model predicting actual credit ratings better than traditional regression models.
2 sharesSource ↗
The paper investigates house price volatility and its causes in 70 Chinese cities from 2005 to 2019, finding significant geographical differences in volatility patterns and the influence of education and healthcare amenities.
3 sharesSource ↗
The research investigates the use of deep reinforcement learning in creating adaptable and compliant gaming AI, overcoming limitations of traditional methods.
3 sharesSource ↗
The research indicates that higher equity market returns occur when the VIX exceeds a certain level and are lower following high market sentiment, with consistent results across different return periods and evaluations.
2 sharesSource ↗
The research highlights the detrimental effect of technology-related political risks on firm innovation, with unrelated diversification intensifying these effects, based on data from 598 firms from 2012 to 2020.
3 sharesSource ↗
The article discusses the complexity of data commodification, suggesting it's a spectrum rather than a binary concept, and examines the impact of new EU data policies.
9 shares5 citations todaySource ↗
The paper provides a proof of Beals conjecture in number theory using machine learning, highlighting its potential in discovering mathematical proofs.
2 sharesSource ↗
The study uses machine learning to identify leading stocks, demonstrating that stocks with negative leaders can predict future market returns.
4 sharesSource ↗
The article introduces a new estimation algorithm for high-dimensional panel datasets, including the asymptotic distribution theory and a test for the number of factors in a tensor factor model.
2 sharesSource ↗
The article explores the paradox of larger deep neural networks performing better than smaller ones, despite theories suggesting one layer should suffice.
2 sharesSource ↗
The article evaluates the efficiency of four deep learning models in predicting high-frequency returns in equities, emphasizing the role of network structure, input choice, and time inclusion.
125 shares3 citations todaySource ↗
The article examines the theory that the superior performance of volatility targeting strategies over basic buy-and-hold positions is due to trend following, and explores the link between volatility targeting and trend following.
3 sharesSource ↗
The paper addresses the mean variance hedging issue assuming the underlying trading strategy doesn't have to be self-financing, introducing a non-self-financing trading strategy with an extra jump noise source.
3 sharesSource ↗
ChatGPT can create profitable day trading strategies by analyzing Twitter posts and suggesting stocks to buy or sell, according to a study.
2 sharesSource ↗
Adjusting a portfolio's stock percentage based on stock earnings yield and bond current yield can help counter weak medium-term returns, suggests a paper.
2 sharesSource ↗
Cloned portfolios from SEC EDGAR Form 13F filings can match original funds' performance and outperform the S&P 500 index by 24.25% annually on a risk-adjusted basis, according to research.
3 sharesSource ↗
Credit VIX Effects: The study indicates that uncertainty in corporate credit risk, particularly for US investment-grade firms, significantly influences volatility in major asset classes and markets.
2 sharesSource ↗
The study suggests that the amount of information disclosed in an economy with large trading data depends on the degree of information overlap across data sources, with derivative securities trades providing significant information about macroeconomic shocks.
3 sharesSource ↗
The study shows that standard machine learning models and publicly available transaction data can effectively predict Bitcoin price movements, surpassing existing prediction models.
2 sharesSource ↗
The study analyzes irregularities in daily returns of UK companies on the London Stock Exchange from 1990-2022 using volatility models.
3 sharesSource ↗
The article proposes a model of litigation finance, suggesting funders prefer high-risk cases, potentially increasing lawsuits and challenging the legal system.
2 sharesSource ↗
The research shows that insider investors often deviate from suggested trading strategies, choosing to delay information acquisition costs when trading returns are unaffected.
2 sharesSource ↗
The study finds that investors gather more public information about firms they are socially closer to, leading to predicted short-term earnings, stock returns, and increased volatility.
2 sharesSource ↗
The general machine-learning papers the letter carried in 2023-25.
6 items
The book offers an in-depth look at differentiable programming, a new method that allows complex computer programs to be differentiated, enabling optimization of program parameters and introducing probability distributions.
207 shares82 citations todaySource ↗
Large Language Models such as GPT-3.5, GPT-4, and Llama2 struggle to explore in reinforcement learning environments without significant interventions, indicating the need for algorithmic interventions in complex decision-making scenarios.
196 shares92 citations todaySource ↗
Embedding Generation: CoLLEGe, a new meta-learning framework, is presented which enhances the ability of language models to learn new concepts quickly using a few example sentences or definitions.
21 shares4 citations todaySource ↗
The research shows that large language models can be updated efficiently with new data, saving significant computational resources and matching the performance of re-training from scratch.
1,205 shares132 citations todaySource ↗
The paper presents TableLlama, an open-source large language model fine-tuned for table-based tasks, and introduces a new dataset, TableInstruct, which improves the performance and generalizability of these models.
84 shares233 citations todaySource ↗
The paper evaluates different explainable artificial intelligence methods in the context of climate science, comparing their suitability for specific research problems.
53 shares90 citations todaySource ↗
Papers that shipped their code, from the Papers with Code feed (2023-25).
2 items
LLM Hypothesis Testing: Assessing large language models (LLMs) is a complex process that involves analyzing multiple responses.
1,718 shares
Current foundation models struggle with time series data due to unique challenges, necessitating task-specific models.
197 shares
Repositories the letter featured.
7 items
LOBFrame is a new open-source code base that provides a novel way to process large-scale Limit Order Book data.
20 shares
The HARNet model, used for forecasting realized volatility, has been incorporated into TensorFlow.
21 shares
GPT 100 offers private interaction with documents, ensuring data security and preventing leaks.
50,009 shares
Python 3 source code is utilized for conducting directional change analysis of financial time series data.
14 shares
The piece delves into GRANDE, a decision tree method that employs gradient-based learning.
43 shares
The article presents Devika, an AI software engineer designed to understand human instructions, conduct research, and write code, rivaling Cognition AI's Devin.
2,226 shares
The piece examines the application of large language models in reverse engineering to decompile binary code.
1,904 shares
Industry news: funds, hiring, markets and regulation.
4 items
Workers of Elon Musk's adopted child generally don't seek jobs in the financial sector.
2 shares
New Era: A report based on data emphasizes the fast expansion and difficulties in the family office sector, with a focus on operational problems and knowledge from top companies.
2 shares
Awkwardness: A prop trading employee has reported a negative workplace experience.
2 shares
Two ex-Bridgewater Associates executives, dismissed in 2023, are suing the firm for alleged favouritism, age, and sex discrimination, seeking a substantial payout.
1 shares
Posts from quant and economics blogs and newsletters.
8 items
The article talks about the use of PositionBook data to modify existing volume-based indicators.
7 shares
The author discusses how PositionBook data can be used to alter existing volume indicators in forex trading.
7 shares
The post highlights the use of PositionBook data to change existing volume-related indicators in forex trading.
7 shares
The article delves into the application of PositionBook data to adjust volume-based indicators.
7 shares
The author studies how PositionBook data can be used to modify existing volume indicators in forex trading.
7 shares
The article explores the modification of volume-incorporating indicators using PositionBook data.
7 shares
Favorite LLMs Papers: The article reiterates the idea of altering volume-based indicators with PositionBook data.
7 shares
Two Sigma researchers highlight key machine learning papers and presentations from NeurIPS 2023.
4 shares
Posts from quant researchers on X.
9 items
The article discusses the application of machine learning and large language models in the fields of economics and finance.
5 shares
The author provides a weekly summary of recent research, mainly on machine learning and large language models, and also on ESG and macro.
4 shares
The article gives a detailed review of the use of synthetic data in finance, along with a comprehensive list of references for further study.
2 shares
The article presents LLM Agents for Data Science, a tool that generates detailed outputs from a single-line requirement, including user stories and competitive analysis.
2 shares
Python Synthetic Data Library: The article introduces DataDreamer, a new open-source Python library for generating synthetic data and managing training workflows.
1 shares
The article discusses the Moira Universal Time Series model, pretrained on HuggingFace LOTSA data, capable of handling multiple variables.
1 shares
The article offers detailed lecture notes on the topics of ridge regressions and lasso elastic net.
0 shares
The article explores various aspects of MLOps, presenting them in a diverse, kaleidoscopic perspective.
0 shares
The paper explores a method for predicting a stock's performance against the market, noting simple models can be as effective as complex ones.
0 shares