Paper: arXiv 2407.18583
Authors: Stéphane Crépey, Botao Li, Hoang Nguyen, Bouazza Saadeddine
Abstract
We present a unified framework for computing CVA sensitivities, hedging the CVA, and assessing CVA risk, using probabilistic machine learning meant as refined regression tools on simulated data, validatable by low-cost companion Monte Carlo procedures. Various notions of sensitivities are introduced and benchmarked numerically. We identify the sensitivities representing the best practical trade-offs in downstream tasks including CVA hedging and risk assessment.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 7.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced probabilistic machine learning, SVD, and neural networks with complex derivations (Malliavin calculus referenced), indicating high mathematical density. It is highly empirical with a provided GitHub code link, explicit backtesting on financial CVA, and implementation on specific hardware (GPU/CPU) with statistical rigor.
Research Flowchart
flowchart TD A["Research Goal<br>Unified Framework for CVA<br>Sensitivities, Hedging & Risk"] --> B["Methodology<br>Probabilistic Machine Learning<br>Refined Regression + Monte Carlo"] B --> C["Inputs<br>Simulated Market & Counterparty Data"] C --> D["Computation<br>Calculate CVA & Sensitivities"] D --> E["Application<br>Hedging & Risk Assessment"] E --> F["Outcomes<br>Identified Practical Trade-offs<br>Validated Framework"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a unified framework for computing CVA sensitivities, hedging CVA, and assessing CVA risk using probabilistic machine learning on simulated data.
Source sentence
We present a unified framework for computing CVA sensitivities, hedging the CVA, and assessing CVA risk, using probabilistic machine learning meant as refined regression tools on simulated data, validatable by low-cost companion Monte Carlo procedures.
Assumes 4
- The second‑order mixed derivatives of cal‑err with respect to ρ and market variables z exist. statistical
Source sentence
we assume that the matrix of all the second derivatives of cal‑err with respect to one component in ρ and one component in z, denoted by
- Parameters are grouped and only one group is bumped per subset of simulated paths to reduce multicollinearity. execution
Source sentence
To address this issue, we divide the parameters into groups. The m simulated paths of the pricing model are then partitioned into blocks of paths such that only one group of parameters is bumped in each subset.
- Linear regression may fail to identify individual parameter effects when parameters have very different magnitudes and price impacts. statistical
Source sentence
a linear regression may not be able to identify the individual effect of each parameter when all are bumped simultaneously.
- The Jacobian trick for converting model to market sensitivities works only for first‑order sensitivities. model
Source sentence
the Jacobian trick of Section 2.3 to convert model into market sensitivities is only workable for first‑order sensitivities, hence our focus on the latter in this work.
Data
- Simulated pricing model paths synthetic unknown
Source sentence
The m simulated paths of the pricing model are then partitioned into blocks of paths such that only one group of parameters is bumped in each subset.
Costs
Partial
Transaction costs can be incorporated via a deep‑hedging approach, but are not included in the headline results.
Source sentence
Note that one could also easily account for transaction costs in this setup, which is nothing but a deep hedging approach over one time step for computing the HVA trend c and the hedging policy ∆ec compressing the ensuing economic capital and its cost.
Finds 5
- Nested Monte Carlo provides the most accurate twin upper bounds but requires about 2 hours, whereas simulation‑label generation takes ~1 minute, neural‑network training ~30 s, and LS regression 2‑3 s. performance
Source sentence
In terms of the twin upper bounds, the nested CVA has the best accuracy, but (for a given risk horizon t) it takes about 2 hours, versus about one minute of simulation time for generating the labels, plus 30 seconds for training by neural networks and 2-3 seconds for LS regression.
- Neural‑network predictors outperform linear regression for longer maturities (t = 1 and 3.5 years) where CVA non‑linearity is significant, but linear methods are better at short horizons. predictive-accuracy
Source sentence
... falls short at one year, where the nonlinearity of CVAt(x, ρ) becomes significant: for t = 1 and 3.5 years, the neural network outperforms the linear regression significantly.
- Both nested δCVA(t) and neural‑network δCVAθ(t) yield more conservative VaR and ES estimates than any linear or quadratic proxy across all cases. pricing-calibration
Source sentence
the nested δCVA(t) and the neural network δCVAθ(t) provide more conservative VaR and ES estimates than any linear(-quadratic) proxy in all cases.
- Smart bump sensitivities achieve accuracy comparable to standard bump sensitivities while using fewer simulation paths. theoretical
Source sentence
smart bump sensitivities essentially mean standard bump sensitivities with less paths, but with the important implementation caveat mentioned at the end of Section 2.1
- Linear diagonal quadratic LS proxies give a very fast yet almost equally reliable view on CVA risk compared to neural‑network regressions. performance
Source sentence
an LS proxy, linear diagonal quadratic in market bumps, provides an even quicker (as it is regressed without training) and almost equally reliable view on CVA risk as the neural
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 20 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (289)
- An explicit scheme for pathwise XVA computations — → 2401.13314
- MAR 32 - Internal models approach: Backtesting and P&L attribution test requirements (2020)
- MAR 50 - Credit valuation adjustment framework (Version effective as of 15 Dec 2019)
- MAR 50 - Credit valuation adjustment framework (Version effective as of 01 Jan 2023)
- Review of the Credit Valuation Adjustment Risk Framework (2015)
- Matrix computations (2013)
- Positive XVA s (2022)
- The Cost of Hedging XVA (2021)
and 281 more
- Hedging value adjustment: Fact and friction (2021)
- C onsultative document: F undamental R eview of the T rading B ook: A revised market risk framework (2013)
- Regulation (EU) no 648/2012 of the European parliament and of the council of 4 July 2012 on OTC derivatives, central counterparties and trade repositories (2012)
- Hedging Valuation Adjustment and Model Risk (2023)
- Proposal for a Regulation of the European Parliament and of the Council amending Regulation (EU) No 575/2013 as regards requirements for credit risk, credit valuation adjustment risk, operational risk, market risk and the output floor (2021)
- Maximum likelihood estimation for G aussian processes under inequality constraints (2019)
- Recovering volatility from option prices by evolutionary optimization (2005)
- Model uncertainty and its impact on the pricing of derivative instruments (2006)
- Neural networks for option pricing and hedging: a literature review (2020)
- Deep Learning Volatility (2019)
- Pricing with a smile (1994)
- A Divide-and-Conquer B ayesian Approach to Large-Scale Kriging (2017)
- Magic Points in Finance: Empirical Integration for Parametric Option Pricing (2017)
- Constant-Time Predictive Distributions for Gaussian Processes (2018) — link
- Multi-kernel Gaussian Processes (2011) — link
- Multi-task Gaussian Process Prediction (2007) — link
- Arbitrage-free SVI volatility surfaces (2014)
- Deep Local Volatility (2020) — link
- Trade Incremental XVA Computations
- Calibration of the local volatility in a generalized B lack– S choles model using T ikhonov regularization (2003)
- Financial Modeling: A Backward Stochastic Differential Equations Perspective (2013)
- Sequential Design and Spatial Modeling for Portfolio Tail Risk Measurement (2018) — link
- Machine learning for quantitative finance: fast derivative pricing, hedging and fitting (2018)
- Wealth Transfers, Indifference Pricing, and XVA Compression Schemes (2020)
- Model Calibration with neural networks (2017)
- Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond (2001)
- Gaussian Process Covariance Kernels for Pattern Discovery and Extrapolation (2013) — link
- Occam’s Razor (2001)
- Valuing American Options by Simulation: A Simple Least-Squares Approach (2001) — link
- Smoking Adjoints: fast evaluation of Greeks in Monte Carlo calculations (2005)
- Universal Kernels (2006) — link
- Prediction and Regulation by Linear Least-Square Methods (1983)
- LSM Reloaded – Differentiate xVA on your iPad Mini (2017)
- Bayesian Online Multitask Learning of Gaussian Processes (2010) — link
- Multivariate Gaussian and Student$-t$ Process Regression for Multi-output Prediction (2017) — link
- Fast Greeks by algorithmic differentiation (2011)
- Real-time counterparty credit risk management in Monte Carlo (2011)
- Gaussian processes for time-series modelling (2013) — link
- Hard Shape-Constrained Kernel Machines (2020)
- Algorithms for Optimization (2019)
- Multilayer feedforward networks are universal approximators (1989)
- Smooth Function Approximation by Deep Neural Networks with General Activation Functions (2019) — link
- Convertible Bonds in a Defaultable Diffusion Model (2011)
- Local Gaussian process approximation for large computer experiments (2015)
- Portfolio optimization through Kriging methods (2016) — link
- Product Kernel Interpolation for Scalable Gaussian Processes (2018)
- Counterparty Risk and Funding: Immersion and Beyond (2016)
- Machine Learning: A Probabilistic Perspective (2012)
- The volatility surface: a practitioner’s guide (2011)
- Stochastic kriging for efficient nested simulation of expected shortfall (2010)
- Kernels for vector-valued functions: A review (2012)
- Introduction to Gaussian processes (1998)
- Kriging the Local Volatility (2020)
- Counterparty Risk and Funding (2014)
- Modelling, Pricing, and Hedging Counterparty Credit Exposure (2010) — link
- Counterparty credit risk in a multivariate structural model with jumps (2015)
- Bayesian learning for neural networks (1996)
- Efficient XVA Management: Pricing, Hedging, and Attribution using Trade-Level Regression and Global Conditioning (2014)
- Statistical Hedging (2019)
- Variable Selection for Nonparametric Gaussian Process Priors: Models and Computational Strategies (2011) — link
- Gaussian Processes for Machine Learning (2006)
- U ncertainty Quantification for Stochastic Approximation Limits Using Chaos Expansion (2017)
- Kriging of Financial Term Structures (2016)
- A Distribution–Free Theory of Nonparametric Regression (2002)
- Stochastic Approximation Schemes for Economic Capital and Risk Margin Computations (2017)
- Arbitrage free implied volatility surfaces (2010)
- Deep learning volatility: a deep neural network perspective on pricing and calibration in (rough) volatility models (2020) — link
- Gated neural networks for implied volatility surfaces (2020)
- KrigHedge: Gaussian Process Surrogates for Delta Hedging (2020)
- Applied Numerical Linear Algebra (1997) — link
- Deep calibration of rough stochastic volatility models (2018) — link
- Gaussian Process Regression for Derivative Portfolio Modeling and Application to CVA Computations (2020)
- Sequential Design for Optimal Stopping Problems (2015)
- Kriging Metamodels and Experimental Design for Bermudan Option Pricing (2018)
- Pricing options and computing implied volatilities using neural networks (2019)
- Deep smoothing of the implied volatility surface (2019)
- Deep learning in asset pricing (2019)
- Robust estimation of shape-constrained state price density surfaces (2015)
- Delta-hedging vega risk? (2004)
- Deeply learning derivatives (2018) — link
- Machine learning in finance: the case of deep learning for option pricing (2017)
- A nonparametric approach to pricing and hedging derivative securities via learning networks (1994)
- Incorporating functional knowledge in neural networks (2009)
- Incorporating second-order functional knowledge for better option pricing (2001)
- Pricing and hedging derivative securities with neural networks and a homogeneity hint (2000)
- Pricing and hedging derivative securities with neural networks: Bayesian regularization, early stopping, and bagging (2001)
- Kernel regression for vehicle trajectory reconstruction under speed and intervehicular distance constraints (2020)
- Calibration of the local volatility in a trinomial tree using T ikhonov regularization (2002)
- LOCAL VARIANCE GAMMA AND EXPLICIT CALIBRATION TO OPTION PRICES (2017) — link
- Finite-dimensional Gaussian approximation with linear inequality constraints (2018)
- Gaussian processes for machine learning (2006)
- Machine learning for quantitative finance: fast derivative pricing, hedging and fitting (2018)
- Gaussian process emulators for computer experiments with inequality constraints (2017)
- Finite-dimensional approximation of Gaussian processes with inequality constraints and errors measurements (2017)
- Maximum likelihood estimation for Gaussian processes under inequality constraints (2018) — link
- Exact Hamiltonian monte carlo for truncated multivariate gaussians (2014)
- RestrictedMVN: multivariate normal restricted by affine constraints (2016)
- The normal law under linear restrictions: simulation and estimation via minimax tilting (2017)
- The interpretation and estimation of effective sample size (1984)
- A probabilistic approach to nonparametric local volatility (2019) — link
- Gaussian process modeling with inequality constraints (2012)
- Monotone emulation of computer experiments (2015)
- Gaussian processes with monotonicity information (2010)
- Estimating shape constrained functions using Gaussian processes (2016)
- Gaussian processes with linear operator inequality constraints (2019) — link
- f-anogan: Fast unsupervised anomaly detection with generative adversarial networks (2019)
- Sparse representation of multivariate extremes with applications to anomaly detection (2017)
- On anomaly ranking and excess-mass curves (2015)
- Identification of outliers in multivariate data (1996)
- An overview of anomaly detection techniques: Existing solutions and latest technological trends (2007)
- A Bayesian approach to outlier detection and residual analysis (1988)
- Machine learning techniques for anomaly detection: an overview (2013)
- A D arwinian Theory of Model Risk (2021)
- An overview of deep learning based methods for unsupervised and semi-supervised anomaly detection in videos (2018)
- Outlier analysis second edition (2017)
- Outlier detection for high dimensional data (2001)
- Outlier detection for high-dimensional data (2015)
- Feature reduction using principal component analysis for effective anomaly–based intrusion detection on NSL-KDD (2010)
- Semi-supervised novelty detection (2010)
- Anomaly and Novelty detection for robust semi-supervised learning (2020)
- Unsupervised anomaly detection in large databases using B ayesian networks (2008)
- Anomaly detection in finance: editors’ introduction (2018)
- A comprehensive survey of anomaly detection techniques for high dimensional big data (2020)
- Anomaly detection: A survey (2009)
- Variational autoencoder based anomaly detection using reconstruction probability (2015)
- Regularization of Inverse Problems (1996)
- Learning multiple yield curve models (2019)
- Deep Learning in a Generalized HJM -type Framework Through Arbitrage-Free Regularization (2019)
- An Empirical Analysis of the Swaption Cube (2010) — link
- Deep Learning (2017)
- Hands-on machine learning with S cikit- L earn and T ensorFlow: concepts, tools, and techniques to build intelligent systems (2017)
- To recognize shapes, first learn to generate images (2007)
- Learning multiple layers of representation (2007)
- Reducing the dimensionality of data with neural networks (2006)
- Overcoming the Disentanglement vs Reconstruction Trade-off via J acobian Supervision (2018)
- Contractive auto-encoders: Explicit invariance during feature extraction (2011)
- Higher order contractive auto-encoder (2011)
- Tutorial on variational autoencoders (2016) — link
- Auto-Encoding Variational B ayes (2013) — link
- Recent advances in autoencoder-based representation learning (2018) — link
- Deep learning volatility (2019)
- Volatility model calibration with convolutional neural networks (2018)
- Understanding the difficulty of training deep feedforward neural networks (2010)
- Nonlinear PCA : a new hierarchical approach. (2002)
- Curve dynamics with artificial neural networks (2018)
- Adam: A method for stochastic optimization (2015)
- Representation learning: A review and new perspectives (2013)
- Practical recommendations for gradient-based training of deep architectures (2012)
- Stacked convolutional auto-encoders for hierarchical feature extraction (2011)
- Greedy layer-wise training of deep networks (2007)
- A fast learning algorithm for deep belief nets (2006)
- A mathematical theory of communication (1948)
- Information theory, inference and learning algorithms (2003)
- Identification of outliers (1980)
- Hybrid recommender system based on autoencoders (2016)
- Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion (2010)
- Generative adversarial interpolative autoencoding: adversarial training on latent space interpolations encourage convex latent distributions (2018) — link
- Machine learning: a probabilistic perspective (2012)
- A trivariate cloughâ€â€tocher scheme for tetrahedral data (1984)
- Autorec: Autoencoders meet collaborative filtering (2015)
- Low-rank matrix completion: A contemporary survey (2019)
- Matrix completion and low-rank SVD via fast alternating least squares (2015)
- Collaborative filtering with stacked denoising autoencoders and sparse inputs (2015)
- A survey on matrix completion: Perspective of signal processing (2019) — link
- Spectral regularization algorithms for learning large incomplete matrices (2010)
- International Convergence of Capital Measurement and Capital Standards (2005)
- Basel III: A global regulatory framework for more resilient banks and banking systems (2011)
- The standardised approach for measuring counterparty credit risk exposures (2014)
- When capital is a funding source: The XVA Anticipated BSDEs (2017)
- MVA by Replication and Regression (2015)
- XVA: Credit, Funding and Capital Valuation Adjustments (2015)
- A comparison of selection schemes used in genetic algorithms (1995)
- Genetic algorithms: The crossover-mutation debate (2005)
- Genetic Algorithms in Search Optimization and Machine Learning (1989)
- Interest rate models-theory and practice: with smile, inflation and credit (2007)
- An introduction to genetic algorithms (1998)
- Genetic algorithms in search, optimization and machine learning (1989)
- A knowledge-intensive genetic algorithm for supervised learning (1993)
- Evolutionary Algorithms in Theory and Practice: Evolution Strategies, Evolutionary Programming, Genetic Algorithms (1996)
- Mcgraw-hill science (1997)
- The elements of statistical learning (2001)
- Counterparty Risk and Funding: A Tale of Two Puzzles (2014)
- Evolutionary algos for optimising MVA (2017)
- Credit, Funding, Margin, and Capital Valuation Adjustments for Bilateral Portfolios (2017)
- Efficient XVA Management: Pricing, Hedging, and Allocation Using Trade-Level Regression and Global Conditioning (2014)
- Gaussian process regression for derivative portfolio modeling and application to CVA computations (2019)
- Funding and Credit Risk with Locally Elliptical Portfolio Processes: An Application to CCPs (2018)
- CVA with Greeks and AAD (2015)
- AAD and least-square Monte Carlo: Fast Bermudan-style options and XVA Greeks (2017)
- Capital and Funding (2016)
- ISDA SIMM Methodology, Version R 1.3 (2017)
- LSM Reloaded-Differentiate xVA on your iPad Mini (2017)
- Smoking adjoints: Fast monte carlo greeks (2006)
- Smoking Adjoints: fast evaluation of Greeks in Monte Carlo calculations (2005)
- Modern Computational Finance: AAD and Parallel Simulations (2018)
- Parallel simulated annealing algorithms (1996)
- A theoretical comparison of evolutionary algorithms and simulated annealing (1995)
- Parallel search for combinatorial optimization: Genetic algorithms, simulated annealing, tabu search and GRASP (1995)
- Genetic algorithms and simulated annealing: A marriage proposal (1993)
- The xVA Challenge: counterparty credit risk, funding, collateral and capital (2015)
- A case study on the application of a genetic algorithm for optimization of engine parameters (2006)
- Genetic algorithms for the travelling salesman problem: A review of representations and operators (1999)
- Optimizing deep learning hyper-parameters through an evolutionary algorithm (2015)
- The Simple Genetic Algorithm performance: A comparative study on the operators combination (2011)
- A genetic algorithm analysis towards optimization solutions (2014)
- Derivative-free optimization: a review of algorithms and comparison of software implementations (2013)
- Comparison of genetic algorithms for trading strategies (2014)
- Genetic algorithms in economics and finance: Forecasting stock market prices and foreign exchangeâ€â€A review (2002)
- Genetic algorithms and genetic programming in computational finance (2012)
- A Genetic Algorithm for Investment–Consumption Optimization with Value-at-Risk Constraint and Information-Processing Cost (2019)
- Swap Rates and Credit Quality (1996)
- Bilateral counterparty risk with application to CDSs (2010)
- In the Balance (2011)
- PDE Representations of Options with Bilateral Counterparty Risk and Funding Costs (2011)
- In the balance (2011)
- Funding Costs, Funding Strategies (2013)
- Derivatives funding, netting and accounting (2017)
- The FVA Debate: In Theory and Practice (2012)
- PDE Representations of Options with Bilateral Counterparty Risk and Funding Costs (2010)
- Recovering Volatility from Option Prices by Evolutionary Optimization (2005)
- General wrong-way risk and stress calibration of exposure (2012)
- CVA and wrong-way risk (2012)
- Jumping with default: wrong-way risk modelling for CVA (2015)
- Pricing of XVA adjustments: from expected exposures to wrong-way risks (2018)
- Invariance properties in the dynamic Gaussian copula model (2017)
- Genealogical and interacting particle systems with applications (2004)
- Disentangling wrong-way risk: pricing credit valuation adjustment via change of measures (2018)
- Bounding Wrong-Way Risk in CVA Calculation (2018)
- Derivative-free optimization: a review of algorithms and comparison of software implementations (2013)
- Computing machinery and intelligence (2009)
- Adaptation in natural and artificial systems: an introductory analysis with applications to biology, control, and artificial intelligence (1992)
- Algorithms for hyper-parameter optimization (2011)
- Practical bayesian optimization of machine learning algorithms (2012)
- A parsimonious arbitrage-free implied volatility parameterization with application to the valuation of volatility derivatives (2004)
- Three ways to solve partial differential equations with neural networks: A review
- Deeponet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operators (2019) — link
- Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations (2017)
- A Data-driven Market Simulator for Small Data Environments (2020)
- Machine Learning (1997)
- " Portfolio Selection". The Journal of Finance 7 (1): 77-91 (1952)
- Intelligent Asset Management (2019)
- Volatility is rough (2018)
- Deep learning in asset pricing (2020)
- Expected Shortfall is jointly elicitable with value at risk-Implications for backtesting (2016)
- Differential machine learning: the shape of things to come (2020)
- Deep neural networks algorithms for stochastic control problems on finite horizon: numerical applications (2021)
- Reinforcement Learning in Continuous Time and Space: A Stochastic Control Approach. (2020)
- Tail-GAN: Simulation of extreme events with ReLU neural networks (2021)
- Valuing American options by simulation: a simple least-squares approach (2001)
- Artificial intelligence in asset management (2020)
- Is positive sentiment in corporate annual reports informative? Evidence from deep learning (2019)
- News or noise? Using Twitter to identify and understand company-specific news flow (2014)
- On the capital market consequences of alternative data: Evidence from outer space (2018)
- The perceptron: a probabilistic model for information storage and organization in the brain. (1958)
- Learning internal representations by error propagation (1985)
- Handwritten digit recognition with a back-propagation network (1989)
- Nonparametric regression using deep neural networks with ReLU activation function (2020)
- Multilayer feedforward networks with a nonpolynomial activation function can approximate any function (1993)
- Fast and faster convergence of sgd for over-parameterized models and an accelerated perceptron (2019)
- Regulatory learning: How to supervise machine learning models? An application to credit scoring (2018)
- Governance of Artificial Intelligence in Finance (2020)
- A Kalman filter for robust outlier detection (2007)
- On-line outlier detection and data cleaning (2004)
- Algorithms for mining distance-based outliers in large datasets (1998)
- Central clearing valuation adjustment (2017)
- Optimization of conditional value-at-risk (2000)
- How many inner simulations to compute conditional expectations with least-square Monte Carlo? (2023)
- XVA analysis from the balance sheet (2021)
- Pathwise CVA Regressions with Oversimulated Defaults (2023)
- The P&L Attribution Test (2016)
- Expected S hortfall is jointly elicitable with V alue at R isk— I mplications for backtesting (2016)
- Adjoint algorithmic differentiation: Calibration and implicit function theorem (2011)
- Learning From Simulated Data in Finance: XVA s, Risk Measures and Calibration (2022)
- From model to market risks: The implicit function theorem ( IFT ) demystified (2018)
- Comparing Alternate Methods for Calculating CVA Capital Charges under B asel III (2018)
- The loss surfaces of multilayer networks (2015)
- Root-N-Consistent Semiparametric Regression (1988) — link
- Using Least Squares to Approximate Unknown Regression Functions (1980) — link
- Linear Model and Extensions (2024) — link
- Statistical regression and classification: from linear models to machine learning (2017)
- Analyzing complex survey data (2005)
1 of 289 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.