Quant Letter
The week in quantitative finance research: 90 papers from arXiv, SSRN and RePEc, screened and ranked, 17–23 Oct 2026.
This week spans AI risk in trading, neural microstructure learning, and the frontier's limits. Key tensions emerge: LLM traders fall to information attacks while benchmarks show no leading model beats equal-weight; neural filters cut arbitrage uncertainty eightfold yet ML VaR models hide regime-shift failures. Mean-field DEX liquidity and arbitrage-free IV surfaces advance pricing theory. See Information Injection Attacks, State-Space Models Learn Market Structure, and ML Value-at-Risk Validation Under Regime Shifts.
If you read five papers this week, read these.
Fabricated news reaches 62% attack success against LLM trading agents, far exceeding prompt injection, with surface-form credibility cues mattering more than content accuracy.
The study shows that machine-learning VaR models passing standard backtests breach risk limits two to nine times worse in stressed regimes, revealing false comfort in current regulatory validation.
Using 3.8 billion option contracts, the study finds implied kurtosis predicts five-day returns with 2.33 Sharpe ratio and improves volatility forecasting by 23.3 percent.
Topics drawing unusually many papers this week.
2.3% of this week's 696 new papers, 1.8x its share over the previous 4 weeks (1.3%).
2.4% of this week's 696 new papers, 1.8x its share over the previous 4 weeks (1.4%).
2.0% of this week's 696 new papers, 1.8x its share over the previous 4 weeks (1.1%).
What the quant community starred on GitHub this week, from a panel of 10,589 developers who follow financial machine learning (285 of the 859 we checked starred something). Counts only: we never publish who starred what.
Open-source market platform tracking real-time stock prices and alerts.
★ 20.8k+511 this weekTypeScript
Foundation model for the language of financial markets.
★ 43.4k+2,562 this weekPython
Research workbench for China A-shares, US and HK stocks.
★ 2,782+63 this weekTypeScript
Swarm intelligence engine for financial forecasting and predictions.
★ 81.4k+1,689 this weekPython
Unknown tool with no description provided.
★ 18.3k+17.6k this weekPythonNew repo
High-performance large-scale inference framework for DeepSeek models.
★ 20.6k+20.6k this weekPythonNew repo
Reverse engineer anything with agents from app behavior to binaries.
★ 370k+76.1k this weekTypeScript
Ephemeral workspaces with durable shared history and parallel agent support.
★ 1,978+1,529 this weekRustNew repo
Local-first search engine and MCP server for AI agent memory.
★ 13.3k+6,242 this weekRust
Unknown tool with no description provided.
★ 7,258+2,250 this weekJavaScript
Now published in SIAM J. Financial Math. · featured 1 Jun 2023, 3 days after release; 4 citations
Now published in SIAM J. Financial Math. · featured 31 Jul 2024, 8 days after release; 4 citations
Now published in SIAM J. Financial Math. · featured 21 Feb 2024, 2 days after release; 3 citations
Filter by venue, topic or fanfare.
No papers match these filters.
Quantitative finance and ML-for-finance preprints
30 papers · 28 figures
Fabricated news reaches 62% attack success against LLM trading agents, far exceeding prompt injection, with surface-form credibility cues mattering more than content accuracy.
Fanfare 4ML & AI MethodsMengxiao Wang and Nitesh SaxenaPDF
A public benchmark tests whether large language models can complete end-to-end quantitative research with real trading constraints; no frontier model beats the equal-weight market in 2024.
Fanfare 3ML & AI MethodsHoude Qian et al.PDF
Deep reinforcement learning simulates three agent archetypes in automated markets, finding herd-followers generate frequent short stress episodes while manipulators create persistent volatility.
Fanfare 2ML & AI MethodsIrene Aldridge et al.PDF
LineageEvo guides large-language-model-based alpha factor search using lineage-conditioned variation and stratified allocation, promoting diverse and predictive factor discovery.
Fanfare 2ML & AI MethodsXintong Liu et al.PDF
A Mamba sequence model trained on limit-order-book messages learns order-flow imbalance as an internal representation with R² ≈ 0.59 and generalizes across stocks and microstructure metrics.
Fanfare 3Trading, Microstructure & ExecutionValentin Mohl et al.PDF
The research models liquidity provision on decentralized exchanges as a mean-field game and derives closed-form solutions linking critical volatility to market microstructure parameters, validated on Ethereum data.
Fanfare 3Trading, Microstructure & ExecutionMikhail Belokon and Rostislav BerezovskiyPDF
Injecting metaorders into neural limit-order-book generators reveals that market impact emerges naturally from next-token prediction, enabling benchmarking of realistic simulator behavior.
Fanfare 3Trading, Microstructure & ExecutionGeorge Nigmatulin et al.PDF
High-frequency data shows skillful traders have persistent predictive edge; one-standard-deviation improvement in forecasting ability raises position P&L by 1.02–1.19 units.
Fanfare 2Trading, Microstructure & ExecutionXinlei Hao et al.PDF
Optimal execution windows maximize realized P&L by balancing alpha decay against concave impact costs; schedule shapes follow from convex optimization without depending on order size or impact coefficients.
Fanfare 3Trading, Microstructure & ExecutionDnyanesh Kulkarni and Revant NayarPDF
Wasserstein GAN post-training on limit order book foundation models cuts error accumulation and improves realism by 17%, outperforming continued pre-training with far lower compute cost.
Fanfare 3Trading, Microstructure & ExecutionAlexandre Bismuth et al.PDF
The research characterizes pure-strategy Nash equilibria for multi-market-maker games, proves equilibria exist under minimal assumptions, and validates theoretical predictions via simulation.
Fanfare 2Trading, Microstructure & ExecutionMatteo Onger and Luigi FoscariPDF
LOBGrad combines autoregressive encoders with diffusion models for limit-order-book forecasting, reproducing microstructure facts and providing calibrated predictive intervals.
Fanfare 2Trading, Microstructure & ExecutionJunyu Chen et al.PDF
OrderFlowDiT forecasts unnormalized bid-ask volume fields using unbalanced optimal transport regularization, reducing transport cost by 45.8 percent relative to mean squared error alone.
Fanfare 2Trading, Microstructure & ExecutionJung Hun Phee and Tucker BalchPDF
Decomposing a market maker into quoter and risk-manager agents trained with cooperative IPPO on Polymarket reduces held-out losses and improves quoting frequency versus baselines.
Fanfare 2Trading, Microstructure & ExecutionSkander LejmiPDF
A neural filter estimates posterior uncertainty over spread states and uses it to size positions, cutting the gap between forecasted and realized returns by two to eight times on Korean equities.
Fanfare 3Asset Pricing & FactorsMinkey ChangPDF
Older observations that conflict with recent data reduce foundation models' probability of beating zero across all tested procedures, distinguishing signal extraction from conflicting-history resistance.
Fanfare 2Asset Pricing & FactorsLevent BulutPDF
AddMart reconstructs implied volatility surfaces from sparse quotes by mapping to martingale densities, ensuring no-arbitrage by construction and enabling crash probability estimation.
Fanfare 3Derivatives & VolatilityHaochen Luo et al.PDF
The paper extends damped-sinc methods beyond Heston to general one-factor stochastic-volatility models including the 3/2 and SABR, achieving basis-point accuracy with comparable computational cost.
Fanfare 3Derivatives & VolatilityAndrey ItkinPDF
An arithmetic stochastic volatility model for crude oil provides semi-closed-form pricing for vanilla and calendar spread options, hedges daily risk effectively, and improves fit over geometric models.
Fanfare 2Derivatives & VolatilityMalthe Rauh Johansen et al.PDF
Simulation-based inference learns posterior distributions for rough Heston parameters and propagates them through neural pricers to deliver calibrated uncertainty intervals for exotic options.
Fanfare 2Derivatives & VolatilityDamiano Brigo et al.PDF
Premium-index funding cannot replicate shareholder cash flows for dividend-paying assets; fairness requires discrete event credits matching dividend atoms, not continuous basis adjustment.
Fanfare 3Crypto & DeFiDonghwa Seo et al.PDF
Testing weak-form efficiency on 8,970 cryptocurrencies over thirteen years, the study finds persistent asymmetry in market ranking but reveals that neglected assets lack liquidity rather than predictability.
Fanfare 3Crypto & DeFiAntonis Ballis and Christos IoannidisPDF
The paper establishes validity of dynamic programming equations for Value-at-Risk objectives in continuous state and action spaces, enabling optimal execution problems with VaR goals.
Fanfare 3Portfolio & AllocationXian Chen et al.PDF
Wiener-Hopf factorization yields optimal trading filters for return forecasts under transient price impact, showing fractional filters and the value of mean-reversion timing for signals.
Fanfare 2Portfolio & AllocationMatteo SmerlakPDF
The study integrates market regime detection with Nested Clustered Optimization to build multi-asset portfolios, showing improved risk-adjusted returns and reduced downside risk versus regime-agnostic approaches.
Fanfare 2Portfolio & AllocationSupriadi and Bambang SiswoyoPDF
The study applies extreme value theory to train neural hedging policies on tail risk, achieving lower losses under extreme scenarios when relevant observations are scarce.
Fanfare 2Portfolio & AllocationCarlos Pérez Octavio Mendoza and Frédéric GodinPDF
The research shows that even small amounts of directionally correct expert information materially improve portfolio allocations in both classical and machine-learning frameworks.
Fanfare 2Portfolio & AllocationOjas Kumar et al.PDF
A convex variational theory unifies worst-case risk evaluation across different ambiguity sets, yielding closed-form solutions and algorithms for portfolio selection under Wasserstein and distributional uncertainty.
Fanfare 3Risk, Credit & BankingJianjun Gao et al.PDF
![Partial-moment ambiguity with \mu=4 , \sigma=1 , \beta=3 , and admissible support [0,15] . (a) Worst-case CDFs at t=\pm](figures/arxiv-14-d2dfae68.png)
Deep-π* attributes trend inflation movements to macroeconomic indicators via neural networks, forecasting inflation more accurately than classical models and signaling the 2021 surge earlier.
Fanfare 3Econometrics & ForecastingCarlos Montes-GaldónPDF
Analysis of 13 language models on 45,964 Form 10-Q filings shows finance-specialized models produce modest advantages in return ranking but lag analyst consensus in earnings forecasts.
Fanfare 2LLMs & TextShengkun Wang et al.PDF
New working papers in finance, economics and ML
30 papers · 1 figures
The study shows that machine-learning VaR models passing standard backtests breach risk limits two to nine times worse in stressed regimes, revealing false comfort in current regulatory validation.
Fanfare 4Risk, Credit & BankingNiraj Neupane
Large purchases of mortgage-backed securities prompt exposed banks to renegotiate existing corporate loans on lender-favorable terms, reducing borrower investment by 6 percent.
Fanfare 4Risk, Credit & BankingLin Xie and Mian Liu
Higher monetary policy uncertainty reduces banks' risk-bearing capacity, widening the lending gap between well-capitalized and undercapitalized banks by 1.65 percent in riskier credit.
Fanfare 3Risk, Credit & BankingFatih Altunok et al.
As firms approach bankruptcy, lending shifts to nonbanks and fintechs despite higher losses given default, revealing worse information and more pro-cyclical credit supply during downturns.
Fanfare 3Risk, Credit & BankingManasa Gopal et al.
After ChatGPT, banks lower default risk for AI-exposed borrowers by about 5 percent but compress credit to market leaders, reordering risk in ways that predict future credit losses.
Fanfare 3Risk, Credit & BankingBruno Buchetti et al.
Asymmetric Clayton and Gumbel copulas and fat-tailed Student-t copulas substantially outperform Gaussian copulas at capturing lower-tail dependence and crisis losses.
Fanfare 2Risk, Credit & BankingHenry Mayambu
Using 3.8 billion option contracts, the study finds implied kurtosis predicts five-day returns with 2.33 Sharpe ratio and improves volatility forecasting by 23.3 percent.
Fanfare 4Asset Pricing & FactorsSimon-pierre Boucher
An intertemporal model shows that bank maturity transformation creates priced systemic risk where high liquidity-beta assets underperform by six to seventeen percent quarterly during stress.
Fanfare 3Asset Pricing & FactorsBingsen Jia and Rose Neng Lai
When U.S. industry runs hot measured by carbon emissions, expected market returns fall 4.2 percentage points the next year, forecasting returns up to five years ahead.
Fanfare 3Asset Pricing & FactorsYunlong Chen et al.
Log expected excess returns decompose exactly into joint cumulants of returns and the stochastic discount factor, explaining thirty-one anomaly factors with three structural parameters.
Fanfare 3Asset Pricing & FactorsRobert Parham et al.
Analysis of ten instruments finds U.S. equity futures are net transmitters of volatility shocks while cryptocurrencies became net receivers in 2024-2026, with total connectedness peaking in 2025.
Fanfare 2Asset Pricing & FactorsSimon-Pierre Boucher
Vision models retrieve visually similar price charts from three decades of history; similar historical patterns predict future cross-sectional returns, especially for stocks with limits to arbitrage.
Fanfare 4ML & AI MethodsHui Chen et al.
As firms adopt AI, they hold less cash, shift from revolving to term credit, pay higher spreads, and rely more on non-bank lenders while expanding investment and employment.
Fanfare 4ML & AI MethodsSuleyman Gozen
A placebo test reveals that features like central bank tone can raise credit model accuracy merely by indicating when loans were made, not by predicting defaults.
Fanfare 3ML & AI MethodsNawaz Muhammad
Five pre-registered hypotheses about machine learning for volatility, drawdowns, trend filtering, stock ranking, and news failed validation on a 2025-26 holdout despite passing development tests.
Fanfare 3ML & AI MethodsPrashant Malan
A platform couples a language model with deterministic execution to turn natural-language requests into auditable reports where every figure is from live data or reproducible code, not model hallucination.
Fanfare 3ML & AI MethodsSimon-Pierre Boucher
A model of factor crowding shows that forced exits amplify shocks up to sixfold and shift crowding costs from average returns into tail losses.
Fanfare 3Portfolio & AllocationLudwig B. Chincarini
A strategy dynamically allocating across 50 smart-beta factors based on past returns and covariances significantly outperforms static allocation even with transaction costs included.
Fanfare 2Portfolio & AllocationVaibhav Lalwani and Ashok Gunturu
Crowding combined with volatility and market stress improves the risk-adjusted performance of volatility-managed factors more than standard volatility management alone.
Fanfare 2Portfolio & AllocationXuan Feng
An agent-based model shows how rising passive market share starves the limit order book of price-sensitive capital, leading to deteriorating price discovery and emergent liquidity crises.
Fanfare 3Trading, Microstructure & ExecutionMarcos Garcés
A model shows how correlated automated trading decisions generate feedback and forced liquidations, with systemic risk rising as market depth shrinks relative to automated activity.
Fanfare 3Trading, Microstructure & ExecutionSiwon Lee
A new methodology identifies bear and bull markets by bandwidth calibration, finding median durations of 74 and 294 days, and fits regime-switching models on cumulative rather than daily returns.
Fanfare 3Trading, Microstructure & ExecutionLieven Baele and Geert Bekaert
Using observable outcomes in prediction markets to identify efficient prices, the study decomposes impact into price-moving and anticipation components, finding informed traders earn mostly through prediction.
Fanfare 3Trading, Microstructure & ExecutionOliver Padmaperuma and Andre Guettler
Resampling bootstrap of ten walk-forward validated FX strategies places their median expectancy at the 77th percentile of its own distribution, indicating modest optimization bias.
Fanfare 2Trading, Microstructure & ExecutionShepherd Mahadzva
Prediction-market prices serve as risk-adjusted state prices to measure event probability and information arrival, improving option valuation and revealing market frictions in elections and Brexit.
Fanfare 3Derivatives & VolatilityMichael Hanke et al.
Encoding intraday bars as descriptive words adds 1.7 to 3.7 points of out-of-sample R² to standard volatility models after spikes but shows no edge on daily bars.
Fanfare 3Derivatives & VolatilityJeremia Passmore
Decomposing debenture spreads into issuer, market, and contract components shows issuer risk explains 54 percent of variance, enabling out-of-sample pricing within 14 basis points.
Fanfare 3Derivatives & VolatilityGabriel Pires
The research separates pricing error from hedging loss by decomposing delta-hedge profit and loss into terminal mismatch, residuals, volatility misspecification, rebalancing, and trading costs.
Fanfare 2Derivatives & VolatilityFrancesco Demuro
State prices from binary prediction markets on the Fed decision and presidential election satisfy the law of one price for complete contingent claims, with overround declining steadily toward resolution.
Fanfare 3Macro-Finance & RatesSimon-Pierre Boucher
A transparency standard with cryptographic timestamping allows independent verification of AI forecasts before market opens, making backtests tamper-proof and comparable across models.
Fanfare 3Econometrics & ForecastingVadym Chernets
Working papers curated by RePEc's NEP field reports
30 papers · 21 figures
Using stock market data to measure AI's effect on software engineering productivity, the study estimates AI raised expected productivity by 32.6% and GDP by 3.6% to 6.5% through end-2025.
Fanfare 4ML & AI MethodsAlex Blumenfeld et al.PDF
The study proves that large language models function as voting rules and identifies millions of Condorcet cycles in ChatGPT investment recommendations, creating potential money-pump vulnerabilities.
Fanfare 3ML & AI MethodsRaša Karapandža and Yaw NyarkoPDF
Over 100,000 VC professionals show extreme concentration: 5% generate 90% of profits, with superstar status increasing access to high-quality deal flow.
Fanfare 4Asset Pricing & FactorsBlake Jackson and Ilya A. StrebulaevPDF
The shale revolution shifted how oil supply shocks affect the U.S. economy: contractionary effects weakened and became expansionary as the nation transitioned to net petroleum exporter.
Fanfare 3Asset Pricing & FactorsDiego R. Känzig et al.PDF
Inflation risk premia move inversely with demand-versus-supply narratives extracted from financial news, consistent with asset pricing theory across euro area and U.S. maturities.
Fanfare 3Asset Pricing & FactorsLuís Fonseca et al.PDF
A matched-issuer study finds euro-area stock-bond correlations have trended toward zero since 2009, with discount rates driving co-movement but financial stress effects weakening in high-inflation periods.
Fanfare 2Asset Pricing & FactorsLinas Jurkšas and Julien IdierPDF
A production economy with sticky prices explains equity and bond term structure slopes and cyclicality through countercyclical inflation and habit-driven risk premiums.
Fanfare 2Asset Pricing & FactorsPierlauro Lopez et al.PDF
A synthesis of two decades of research attributes the housing boom to improved fundamentals, credit expansion, rate declines, and shifting expectations, with defaults driving the bust.
Fanfare 2Asset Pricing & FactorsAdam GurenPDF
The research identifies AI news shocks from equity factors and shows they trigger investment-led US growth but create monetary dilemmas via imported inflation in the euro area.
Fanfare 3Econometrics & ForecastingMaria Sole PagliariPDF
Evidence from a multi-country survey shows that heterogeneous beliefs about how monetary policy works significantly alter households' consumption responses to rate changes.
Fanfare 3Econometrics & ForecastingFrancesco D’Acunto et al.PDF
Research finds that rising import exposure accounts for roughly 40 percent of U.S. Phillips curve flattening since 1976, with major implications for monetary policy trade-offs.
Fanfare 3Econometrics & ForecastingColin J. Hottman and Ricardo M. Reyes-HerolesPDF
Testing eight model features on US GDP forecasts, the study finds dynamic heterogeneity consistently improves point, density, and tail forecasts while other features help selectively.
Fanfare 2Econometrics & ForecastingFreddy García-Albán and Manuel González-AstudilloPDF
A tariff war model shows asymmetric impacts, with the imposing country bearing inflationary burden and the target experiencing larger output contraction via production network propagation.
Fanfare 2Econometrics & ForecastingPablo Aguilar et al.PDF
The paper shows that strengthening collateral pledgeability can paradoxically fuel leverage and instability, confirmed by exploiting the 2005 bankruptcy reform's expansion of repo safe harbors.
Fanfare 3Risk, Credit & BankingHusnu C. Dalgic et al.PDF
A capital structure model shows that interest coverage ratio covenants can improve firm value by reducing default costs, though they constrain shareholder discretion.
Fanfare 2Risk, Credit & BankingMichi Nishihara et al.PDF
European single supervisory mechanism adoption caused firms to reduce intangible asset shares due to tighter lending and higher collateral standards, especially for small firms.
Fanfare 2Risk, Credit & BankingMiguel Ampudia et al.PDF
Increased credit rating agency scrutiny under Dodd-Frank reduced speculative-grade firms' acquisitions, leading to more selective deals and higher announcement returns.
Fanfare 2Risk, Credit & BankingAnup Basnet et al.PDF
The paper characterizes how fiscal uncertainty shrinks the range of multiple equilibria in sovereign debt markets, eventually yielding unique interest rates as uncertainty grows.
Fanfare 2Risk, Credit & BankingOlivier J. Blanchard et al.PDF
Analysis of 4,109 loans totaling $115 billion reveals that post-crisis private debt in emerging economies outperforms both emerging-market bonds and Treasuries with public market equivalents above 1.02.
Fanfare 3Portfolio & AllocationShawn Cole et al.PDF
Estimates show a 188 GW U.S. AI campus buildout requires nearly $9 trillion through 2032, shifting toward leases, joint ventures, and structures that raise leverage and opacity.
Fanfare 3Portfolio & AllocationStijn Van NieuwerburghPDF
The study finds that early textual discussions of portfolio companies better predict ultimate performance, while interim valuations matter more later, signaling general partner quality.
Fanfare 2Portfolio & AllocationJosh Lerner et al.PDF
Analysis of PredictIt shows that limiting investor stakes to $850 per contract undermines price accuracy, with bid prices persistently implying arbitrage profits even after fees.
Fanfare 3Trading, Microstructure & ExecutionEric ZitzewitzPDF
A new parallel filter removes price microstructure noise from high-frequency trade data without bias, producing flat volatility signatures down to one-second frequencies.
Fanfare 2Trading, Microstructure & ExecutionSimon Donker van Heel and Neil ShephardPDF
Italian customs data shows firms match import and export invoicing currencies by country to reduce profit volatility from exchange-rate movements in real transactions.
Fanfare 2Trading, Microstructure & ExecutionEmanuele Brancati et al.PDF
The research examines who participates in U.S. repo markets, the trade-offs influencing repo segment structure, and the implications for monetary policy transmission.
Fanfare 2Trading, Microstructure & ExecutionGara Afonso et al.PDF
Dollar-backed payment stablecoins and tokenization likely reinforce dollar dominance despite reducing transaction frictions, with multipolarity potentially increasing fragility.
Fanfare 3Macro-Finance & RatesGordon Y. Liao et al.PDF
Completing exchange and allocation authority in the EU's carbon market ETS2 cuts present-value policy costs by EUR 1.33 trillion over 2028-2048 versus partial market creation.
Fanfare 3Macro-Finance & RatesJan Abrell and Sebastian RauschPDF
Research develops a high-frequency fragmentation index decomposing euro-area sovereign yields into common, market-driven, and fundamental components, showing Transmission Protection Instrument effectiveness.
Fanfare 2Macro-Finance & RatesAlicia AguilarPDF
A monetary asset pricing model estimates the convenience yield of liquidity at 5.1% annually, roughly ten times conventional measures, with only 24% attributable to urgency.
Fanfare 3Derivatives & VolatilityXinhao Dong and Zijian Wang Lucas HerrenbrueckPDF
The paper identifies three structural sources of measurement divergences in blockchain data and proposes approaches to align technical execution with economic meaning in DeFi ecosystems.
Fanfare 2Crypto & DeFiTimothy Aerts et al.PDF
