Quant Letter: November 2023, Week 3

Weekly (23rd Edition)

Paper of Last Week

Trading with Concave Price Impact and Impact Decay - Theory and Evidence

Trading costs play a central role in designing and implementing quantitative trading strategies. For sizable funds, the crucial concern is their trades' adverse price impact.

• It is well known that impact is concave in trade sizes, in that large trades have a smaller impact than predicted by a linear model and are instead better described by a "square-root law". Price dislocations are also not static but gradually dissipate over time.

• This paper studies statistical arbitrage problems accounting for both the nonlinear and transient nature of price impact. We show that simple explicit trading rules can be derived even for general nonparametric alpha and liquidity signals, and also discuss extensions to several impact decay timescales.

Professor Johannes Muhle-Karbe wrote this small summary above for ML-Quant readers.

These results are illustrated using a proprietary dataset of Capital Fund Management metaorders, which allows us to calibrate the levels, concavity, and decay parameters of the price impact model and analyze their effects on optimal trading.

SSRN

Recently Published

Quantitative

Shapley-Based Approach to Portfolio Performance: The SPPC methodology can determine individual predictors' contributions to portfolio performance, shedding light on the sources of economic value from return predictability. (2023-11-09, shares: 3.0)

Deep hedging and delta hedging relationship: The research examines the link between deep and delta hedging, suggesting a risk-minimizing strategy that combines both with statistical arbitrage, and discusses the effects of statistical arbitrages on deep hedging. (2023-11-10, shares: 2.0)

Volatility Modeling with Neural Networks: A new neural network model is introduced for macroeconomic forecasting, designed to prevent overfitting and improve accuracy. (2023-11-09, shares: 3.0)

Salience Theory and Deep Learning in Energy Market Trading: A trading system using salience theory and deep learning is applied to Chinese new energy stocks, proving the effectiveness of these methods. (2023-11-09, shares: 2.0)

Volatility and Stock Market Sensitivity: US. macroeconomic news impacts the SP 500 more when long-term stock market volatility is high. (2023-11-14, shares: 3.0)

Financial

Data Mining's Impact on Asset Pricing: The study challenges the belief that data mining always improves price efficiency, suggesting it can actually reduce price informativeness due to complexity costs and diminishing data efficacy returns. (2023-11-10, shares: 2.0)

Generating Future Volatility Surfaces: The paper presents a new method for predicting future implied volatility surfaces using historical data, employing a conditional variational autoencoder and a long short-term memory network. (2023-11-09, shares: 17.0)

VWAP Day Trading Systems (overfit alert): The article introduces a day trading strategy based on Volume Weighted Average Price (VWAP) that can identify market imbalances, resulting in a 671% return on a $25,000 investment. (2023-11-13, shares: 1355.0)

Credit Sentiments and Bond Returns: Credit sentiments from conference calls affect bond market returns, with positive sentiments leading to better credit ratings and lower future debt costs. (2023-11-09, shares: 2.0)

Chinese Consumption Shocks and U.S. Equity Returns: China's consumption risk significantly influences U.S. equity returns, with a two-factor model explaining 40% of the variation. (2023-11-10, shares: 4.0)

Recently Updated

Quantitative

LSTM and Linear Regression for Stock Market Prediction: The article presents a study on the use of LSTM neural networks and linear regression for stock market prediction, showing superior performance over traditional models. (2023-06-09, shares: 4.0)

Mortgage Securitization and Information Frictions: The study presents a model of the U.S. housing finance system, illustrating the benefits and challenges of the securitization market. (2023-06-16, shares: 2.0)

FX Risk Management by Managers: The article shares a survey of 110 corporate risk managers on hedging foreign exchange rate risk, revealing that changes in forward and future FX rates greatly influence hedge ratios, and managers are most satisfied when FX risk doesn't affect cash flows. (2023-10-31, shares: 2.0)

The Demise of § 36(B) Litigation: The article debates the issue of mutual fund management fees, arguing that mutual funds are controlled by the investment management firms that create them and manage their portfolios, resulting in the charging of excessive management fees. (2023-05-09, shares: 2.0)

Financial

Asset Returns: Auto Debiased ML: A new machine learning method has been developed to identify risk factors in asset pricing, performing better than traditional methods by eliminating biased estimation and overfitting. (2022-09-28, shares: 2.0)

DCCA of Green and Grey Investments: The research finds that green energy ETFs offer better diversification compared to grey and conventional investment strategies. (2023-10-03, shares: 4.0)

ETFs vs Mutual Funds: Liquidity & Performance: The study suggests that ETFs may not be more liquid than mutual funds and can be subject to short-term mispricing and illiquidity. (2023-11-06, shares: 2.0)

Tracking Retail & Institutional Investors in China: The paper proposes improvements to the order size algorithm for better tracking of retail and institutional purchases in the Chinese stock market. (2023-10-20, shares: 4.0)

Chinese Capital Market Yearbook 2022: The yearbook provides a comprehensive analysis of the return and risk characteristics of stocks, government bonds, and credit bonds in the Chinese capital market. (2023-09-01, shares: 3.0)

ML for Emerging Market Bonds: Machine learning models considering nonlinearities and interactions offer better predictions of corporate bond behavior in emerging markets with high transaction costs, with key predictors tied to low-risk macro and momentum factors. (2023-10-30, shares: 2.0)

Reddit Outages & Meme Stock Trading: The predictability of retail order imbalance on future returns for meme stocks increases during Reddit outages, indicating that intense discussions can disrupt individual investors' decisions. (2023-07-11, shares: 3.0)

Simulating Spread Dynamics for VaR & CVA: A new model using a Gaussian one factor copula is suggested for simulating spread risk in banks' risk models, ensuring consistency between simulated and actual historical spreads. (2021-07-08, shares: 2.0)

Other Areas

Editing Prioritization in Survey Data with Machine Learning: Machine learning is used to identify and correct errors in household finance survey data, with Gradient Boosting Trees being the most effective method. (2023-11-14, shares: 2.0)

Machine learning for IBNR frequencies in non-life reserving: The research introduces a machine learning model for predicting the number of Incurred But Not Reported (IBNR) claims, proving its effectiveness through a study using both simulated and real data. (2023-11-14, shares: 2.0)

Market Efficiency in Blockchain Marketplaces: The research examines market efficiency in blockchain-enabled marketplaces, revealing significant inefficiencies despite complete information transparency. (2023-05-02, shares: 2.0)

ArXiv

Finance

Advanced Techniques for Algorithmic Trading: The research enhances a Deep Q-Network trading model using advanced methods, showing improved performance in automated trading and the potential of convolutional neural networks in trading systems. (2023-11-09, shares: 6)

Topic Model for Financial Textual Data: The study introduces a multi-label topic model for financial texts, achieving high accuracy and showing that stock market reactions depend on the co-occurrence of specific topics. (2023-11-10, shares: 5)

Integrating Language Models into Agent-Based Modeling: The paper introduces Smart Agent-Based Modeling (SABM), a new framework that uses Large Language Models for more realistic simulations of complex systems. (2023-11-10, shares: 5)

QLBS Model Feedback Loops: The QLBS model is expanded to include a large trader's impact on exchange rates and contingent claim prices, using reinforcement learning to find an optimal hedging strategy, reducing transaction costs and aligning with the trader's fair price. (2023-11-12, shares: 4)

Withdrawal Success Optimization: The likelihood of completing a specific investment and withdrawal schedule is maximized using adjustable portfolio weight functions, showing significant improvements when optimal weights are used instead of constant ones. (2023-11-11, shares: 4)

Portfolio Diversification for Investor Abilities: New mathematical techniques are used to determine the optimal portfolio size for investors of different abilities, suggesting that strong investors should have smaller portfolios, weak investors larger ones, and average investors a fluctuating optimal number. (2023-11-11, shares: 4)

Error Analysis of Deep PDE Solvers for Option Pricing: The practical use of Deep PDE solvers for option pricing is examined, identifying three main error sources and concluding that the Deep BSDE method performs better and is more robust against option specification changes. (2023-11-13, shares: 3)

Gaussian Process Method for American Option Pricing: Deep Kernel Learning and variational inference are used to improve high-dimensional American option pricing in the regression-based Monte Carlo method, with successful performance under geometric Brownian motion and Merton's jump diffusion models. (2023-11-13, shares: 3)

Contagion and Liquidity in Markets: The study presents a framework to understand price-mediated contagion in a system with endogenously determined market liquidity, showing the significant impact on system risk. (2023-11-10, shares: 5)

DFMM Asset: Tradeable Unit in Cross-Chain Finance: The paper investigates the Intermediating DFMM Asset in a multi-asset market, outlining its features, risk mitigation, and control levers, suggesting its potential to align the interests of various market participants. (2023-11-09, shares: 5)

Optimal Dividend Strategies for Insurers: The research examines the optimal payout of dividends from an insurance portfolio with claims from natural disasters, identifying the best dividend strategies and potential benefits for shareholders. (2023-11-09, shares: 5)

Miscellaneous

Integrating Language Models into Agent-Based Modeling: The paper introduces Smart Agent-Based Modeling (SABM), a new framework that uses Large Language Models for more realistic simulations of complex systems. (2023-11-10, shares: 5)

Generative AI: Boosting Market Prosperity and Dismissing Depression Concerns: A study finds that generative AI can lower average prices in product markets, increase order volume and revenue, and potentially benefit artists rather than causing unemployment. (2023-11-13, shares: 4)

Analysis of Hedera Hashgraph Decentralisation: The study shows that Hedera Hashgraph platform has high wealth centralization and a shrinking core, but recent indexes indicate progress towards decentralization. (2023-11-12, shares: 5)

Enhancing Pricing Models with Transformers: The paper presents new methods to improve non-life actuarial models with transformer models for tabular data, showing better results than benchmark models. (2023-11-10, shares: 5)

Historical Trending

FinGPT: Democratizing Financial Data for LLMs: The Financial Generative Pre-trained Transformer (FinGPT), a new open-source framework, automates the collection and curation of real-time financial data from various online sources, aiming to make large-scale financial data more accessible for large language models. (2023-07-19, shares: 43)

Price Interpretability in Prediction Markets: A study proposes a multivariate utility-based mechanism for prediction markets, unifying existing market-making schemes and characterizing the limiting price through systems of equations reflecting agent beliefs, risk parameters, and wealth. (2022-05-18, shares: 76)

Large-Scale Portfolio Optimization Framework: A new large-scale portfolio optimization framework, using shrinkage and regularization techniques, has been tested and proven effective using 50 years of US company return data. (2023-03-22, shares: 27)

Solution to Lillo-Mike-Farmer Model: A new Lillo-Mike-Farmer model, considering the diversity of traders' order-splitting behavior, emphasizes the importance of the ACF prefactor in data analysis. (2023-06-23, shares: 15)

Two-Way Regression for Panel Data: A new estimator for average causal effects in binary treatment with panel data has been proposed, offering better performance and robustness than the traditional two-way estimator, even with a misspecified fixed effect model. (2021-07-29, shares: 303)

Inventories and Demand Shocks in Supply Chains: Research shows that the position of industries in supply chains significantly influences the transmission of final demand shocks, with upstream industries reacting up to three times more than final goods producers. (2022-05-08, shares: 73)

ArXiv ML

Recently Published

Outlier-Robust Wasserstein DRO: The research introduces an outlier-robust framework for decision-making under data perturbations, providing optimal risk bounds and efficient computation, and validating the theory through standard regression and classification tasks. (2023-11-09, shares: 7)

Coefficient Control for SVRG: The article introduces α-SVRG, a new method for optimizing neural networks that improves training loss reduction across various architectures and datasets. (2023-11-09, shares: 16)

Diffusion Models: Cloud Removal and Urban Change Detection: Diffusion models in AI can improve Earth observation data, aiding in tasks like cloud removal, change-detection dataset creation, and urban replanning. (2023-11-10, shares: 69)

GPTV for Social Media: The study examines the abilities of Large Multimodal Models (LMMs), particularly GPT-4V, in understanding social multimedia content, noting challenges in multilingual comprehension and trend generalization. (2023-11-13, shares: 13)

Greedy PIG: Feature Attribution: The authors suggest a unified discrete optimization framework for feature attribution and selection in deep learning models, introducing an adaptive method called Greedy PIG that performs well in various tasks. (2023-11-10, shares: 11)

Offline RL: Survival: Offline reinforcement learning algorithms can still create effective policies even with incorrect reward labels due to their inherent pessimism and biases in data collection. (2023-06-05, shares: 110)

Data Contamination Quiz for LLMs: The paper introduces the Data Contamination Quiz, a method for detecting and estimating data contamination in large language models, demonstrating improved detection and accurate contamination estimation. (2023-11-10, shares: 8)

RePec

Finance

Liquidation Strategies' Effects in Multi-Asset Market: The paper reveals that certain algorithmic trading strategies can lower liquidation costs but may negatively affect market indicators in a multi-asset artificial stock market. (2023-11-15, shares: 18.0)

Global Equity Correlations and Currency Option-Implied Volatilities: The research finds that exchange rate option-implied volatilities can more accurately predict future global equity market correlations. (2023-11-15, shares: 15.0)

Portfolio Flows: Time-Variation in Push and Pull Factors: The research shows that the importance of push factors in portfolio flows during crises has increased over time, especially for EU countries, and identifies several key push and pull factors. (2023-11-15, shares: 14.0)

Preferred REITs' Portfolio Enhancement Attributes: The research indicates that REIT preferred stocks offer significant diversification benefits and enhance portfolio performance during economic expansion. (2023-11-15, shares: 21.0)

Dynamic Bond Portfolio Optimization with Stochastic Interest Rates: A new framework for multi-period dynamic bond portfolio optimization is proposed in the study, which shows it performs better than single-period optimization. (2023-11-15, shares: 26.0)

Statistical

Explainable AI for Bond Excess Returns: The SHAP technique is used to clarify bond excess return predictions made by machine learning models in a study. (2023-11-15, shares: 21.0)

Policy Uncertainty and Stock Market Volatility: A study finds that high-quality political signals can predict increased stock market volatility. (2023-11-15, shares: 16.0)

Market Momentum and Volatility Risk in China's Equity Market: Research into the Chinese equity market shows a belief-based momentum indicator can predict market volatility. (2023-11-15, shares: 16.0)

DataDriven Newsvendor Problem: High-Dimensional Method: The article explores the use of machine learning to improve demand prediction and restocking decisions in newsvendor problems, using complex historical data. (2023-11-15, shares: 27.0)

Comparative Study of Methods to Identify Sensitive Parameters: The article discusses how supervised machine learning models assign weights to input parameters to achieve the desired outcome, stressing the importance of reliable weights early in the model development process. (2023-11-15, shares: 13.0)

GitHub

Finance

Timeseries ML with Polars: The article explores the application of Polars in large-scale timeseries machine learning, particularly in parallel feature extraction and panel data forecasts. (2023-06-05, shares: 626.0)

einops: Deep Learning Operations Reinvented: The article discusses the transformation of deep learning operations across various platforms like Pytorch, Tensorflow, Jax, etc. (2018-09-22, shares: 7379.0)

elegy: High Level API for DL in JAX: The article presents a new high-level API designed specifically for deep learning in JAX. (2020-06-30, shares: 455.0)

New Grad Positions in SWE, Quant, PM: The article lists full-time job opportunities for fresh graduates in Software Engineering, Quantitative Analysis, and Project Management. (2021-06-08, shares: 8395.0)

quantnotes: Updated Quant Interview Prep Guide: The article provides an updated guide to help prepare for quantitative interviews. (2017-11-17, shares: 581.0)

Trending

Awesome Time Series: Papers, Code, and Resources: The article compiles a list of important codes, academic papers, and other key resources. (2020-03-03, shares: 764.0)

OpenAI Vision API Experiments: Resource: The article serves as a guide for those wanting to explore and improve the OpenAI Vision API. (2023-11-07, shares: 880.0)

ChatGPT Custom Instructions: Customizing Repo: The article provides a collection of tailored instructions for utilizing ChatGPT. (2023-08-15, shares: 738.0)

Context: CLI Tool API for Python Libraries: The article presents a CLI tool API for the most popular 1221 Python libraries. (2023-11-02, shares: 340.0)

LinkedIn

Trending

Python Quant GPT: Revolutionizing Quant Analysis: Python Quant GPT is a new quantitative analysis tool that provides comprehensive data analysis, advanced financial modeling, and AI-driven insights with a user-friendly interface. (2023-11-13, shares: 2.0)

Understanding the Complexity of Oil Markets: Recent oil market fluctuations are driven by broader financial markets and algorithmic behavior, not changes in supply and demand. (2023-11-14, shares: 1.0)

Talk on Large Language Models in Finance: Ioana Boier will discuss the use of Large Language Models in extracting insights from complex data and their application in quantitative finance. (2023-11-13, shares: 1.0)

Informative

ACM AI in Finance Conference '23: The ACM AI in Finance Conference ICAIF'23, showcasing the latest AI research in finance, will take place in Brooklyn, NY from November 27-29. (2023-11-13, shares: 1.0)

Data Sim Seminar with Dimitris Giannakis: Dimitris Giannakis will give a seminar on quantum information and data science at the Lawrence Livermore National Laboratory on December 15th, 2023. (2023-11-13, shares: 1.0)

FRB-CEBRA-ECONDAT Conference on Non-Traditional Data: The author discussed the use of non-traditional data in macroeconomics at the FRB-CEBRA-ECONDAT conference. (2023-11-13, shares: 1.0)

QuantMinds Summit & Workshop Day: The QuantMinds International Summit & Workshop Day will cover topics like advanced machine learning and investment modelling. (2023-11-13, shares: 1.0)

Machine Learning for Factor Models: A paper by Seisuke Sugitomo and Minami Shotaro suggests that machine learning enhances portfolio performance more than traditional methods in fundamental factor models. (2023-11-13, shares: 1.0)

Resource Alert: Quant Trading Events: Christina Qi has curated a list of quant trading events and competitions for those aspiring to enter the quantitative finance field, both students and professionals. (2023-11-13, shares: 1.0)

Podcasts

Quantitative

Volatile Times Investment Strategies: Michael Kramer discusses treasury tail events, the need for a suitable investment strategy, and the potential risks and rewards in investing. (2023-11-09, shares: 10)

ESG Progress: Fadi Zaher reveals the results of a study on long-term ESG trends, highlighting improvements and differences across regions and sectors. (2023-11-09, shares: 4)

Dark Side of Finance and Financial Revolution Hope: Lyn Alden explores the history of money systems, problems with current systems, and the potential of Bitcoin and other digital assets for a fairer economy. (2023-11-14, shares: 4)

Monetary Policy's Role in Commodity Investing: Leigh Goehring talks about the link between monetary policy and commodities investing, and the idea of gold-backed digital currencies on a podcast. (2023-11-09, shares: 3)

Developments in EM Fixed Income Market: Jonny Goulden and Saad Siddiqui discuss the impact of recent market developments on the EM fixed income asset class in a 2023 podcast. (2023-11-09, shares: 3)

Related

Fed's 2024 Balance Sheet Outlook: The podcast discusses the future of the Federal Reserve's balance sheet, considering the monetary policy and expected changes in reserves by 2024. (2023-11-09, shares: 3)

Linda Gibson: Quant Investing Pioneer: Barry Ritholtz interviews Linda Gibson, CEO of PGIM Quantitative Solutions, about her career and insights into quantitative investing on Bloomberg Radio. (2023-11-10, shares: 3)

Enhancing Investment Outcome with Options: Giuseppe Sette, President of Toggle AI, shares insights on using options to enhance investment outcomes in an interview with IBKR's Jeff Praissman. (2023-11-13, shares: 3)

Presidential Campaign Fundraising: Fundraiser in Chief: Brendan Doherty's new book examines the relationship between campaigning and governing, with a focus on presidential fundraising in the past 60 years. (2023-11-13, shares: 2)

Global Value Stocks Opportunity: In his podcast, Meb discusses the potential in global value stocks using a shareholder yield approach. (2023-11-11, shares: 1)

News

Quantitative

JPMorgan AI MD Joins Trading Rush: A commodities trading firm is integrating quantitative analysis and machine learning into its business processes. (2023-11-09, shares: 9

Hedge Fund Talent War Tied to High Payouts: Industry leaders suggest that large payouts by some multistrategy majors are triggering a talent war in the global hedge fund industry. (2023-11-09, shares: 2)

Hunterbrook Hedge Fund Seeks Financial Journalists: VC-backed hedge fund firm, Hunterbrook, is recruiting financial journalists to help its research team identify potential investments. (2023-11-13, shares: 2)

Twitter

Quantitative

Digital Asset Volatility in Crypto Winter: The study uses LSTM and RFSV techniques to analyze the volatility of digital assets during a period of significant decline in cryptocurrency values. (2023-11-14, shares: 2)

Econometrics: Analyzing Economic Data Statistically: Econometrics, a field that combines economics, statistics, and math, is offering a free PDF download for data analysis. (2023-11-11, shares: 2)

Currency Anomalies' Performance Decline Post Publication: Research indicates that the performance of equity and currency anomalies decreases after their publication, implying that market players use these publications to correct mispricing. (2023-11-14, shares: 1)

Towhee: LLM-based data transformation: Towhee is a pipeline orchestration tool that uses Large Language Models to convert raw multimodal data into specific formats. (2023-11-10, shares: 1)

GraphCast: DeepMind's opensource weather model: Google Deep Mind has made its advanced weather forecasting model, GraphCast, open-source. (2023-11-14, shares: 0)

Funds leverage algorithms for earnings call analysis: Investment funds are utilizing algorithms to analyze earnings call transcripts, leveraging audio for more comprehensive information than text. (2023-11-14, shares: 0)

Miscellaneous

Fractal Geometry and Market Microstructure: Article 2: The study delves into the basics of fractal geometry and its use in understanding market structures. (2023-11-11, shares: 0)

Optimal kmeans clusters: Article 3: The piece explores the best use of k-means clusters through the k-scorer algorithm. (2023-11-10, shares: 0)

FT DataViz on bondmarket malaise depth: Article 4: The Financial Times offers a detailed examination of the current problems in the bond market. (2023-11-10, shares: 0)

Videos

Cryptos: Are They Different? Evidence from Retail Trading: The ABFR forum, a group of scholars studying AI and big data's impact on society, plans monthly discussions on economics and finance, with the next one on October 26, 2023. (2023-11-09, shares: 0.0)

Workshop on AI in Finance: Raw & Unedited Recording: A raw screen recording from the Workshop on AI in Finance at Texas State University San Marcos on October 27, 2023, is accessible on GitHub. (2023-11-10, shares: 4.0)

Reddit

Quantitative

Underrated Quant Channel: The author posts a video link without providing any context or description of its content. (2023-11-12, shares: 142.0)

Spread Failures: The author shares their failed attempts at making profitable strategies on calendar spreads using NIFTY options. (2023-11-10, shares: 22.0)

C as a Low-Level Language: The author, an actuary, is interested in transitioning to a role in pricing and risk monitoring models, highlighting the need for C programming skills and understanding of derivative markets. (2023-11-09, shares: 19.0)

Pairs Trading Practicality: The author discusses the practicality and challenges of implementing pairs trading strategies, including issues like trade discoverability. (2023-11-09, shares: 15.0)

Linux/Unix in Quant Firms: The author seeks understanding on why quant firms require knowledge of Linux/Unix systems from applicants, asking for industry professionals' insights. (2023-11-09, shares: 13.0)

Rising

Overestimation of Quant Knowledge by Nonquants: The article provides insights into the quant industry, discussing the importance of understanding its realities before entering, including strategies, programming languages, salaries, and daily tasks. (2023-11-09, shares: 239.0)

Data Processing Skills Gap Among Discretionary Traders: The author shares their experience in the systematic side of the quant industry, highlighting the lack of data processing skills among many quant traders from the discretionary side. (2023-11-12, shares: 32.0)

Risk Quants' Salary Progression: The article explores the potential annual salary increase for a beginner in the risk quant industry. (2023-11-09, shares: 25.0)

Performance and Bonus Systems Evaluation at TMG: The author seeks updated information about a Chicago-based company where they are interviewing for an algo trader role, focusing on the company's performance, bonus systems, and market standing. (2023-11-09, shares: 13.0)

Risk Evaluation in New Quant Firms: The article discusses the risks and considerations of joining a new quant firm, comparing it to larger, established firms, and provides advice on evaluating an offer. (2023-11-10, shares: 40.0)

Paper with Code

Language Models as Teachers: The article assesses a technique's effectiveness on different public models and complex tasks. (2023-11-15, shares: 237.0)

DeciAI PPLiteSeg: Real-Time Segmentation Model: The article highlights the increasing need for semantic segmentation techniques in practical applications. (2023-11-11, shares: 3697.0)

Posttraining with Data Curriculum: The article presents a data curriculum learning scheme that enhances alignment in contrastive posttraining by progressing from simpler to more complex pairs. (2023-11-13, shares: 2755.0)

Accelerating Model Pretraining with Pruning: The study explores the use of structured pruning to develop smaller LLMs from larger pre-existing models. (2023-11-09, shares: 194.0)

Serving Concurrent LoRA Adapters: The article presents SLoRA, a system engineered for the scalable usage of multiple LoRA adapters. (2023-11-09, shares: 190.0)