arXiv subjects

Ruodu Wang

Publications and source records attributed to Ruodu Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Quadratic-form Optimal Transport

We introduce the framework of quadratic-form optimal transport (QOT), whose transport cost has the form $\iint c\,\mathrm{d}π\otimes\mathrm{d}π$ for some coupling $π$ between two marginals. Interesting examples of quadratic-form transport cost and their optimization include inequality measurement, the variance of a bivariate function, covariance, Kendall's tau, the Gromov--Wasserstein distance, quadratic assignment problems, and quadratic regularization of classic optimal transport. QOT leads to substantially different mathematical structures compared to classic transport problems and many technical challenges. We illustrate the fundamental properties of QOT and provide several cases where explicit solutions are obtained. For a wide class of cost functions, including the rectangular cost functions, the QOT problem is solved by a new coupling called the diamond transport, whose copula is supported on a diamond in the unit square.

math.PR↗

arXiv2026

Adaptive Window Selection for Financial Risk Forecasting

Risk forecasts in financial regulation and internal management are calculated through historical data. The unknown structural changes of financial data pose a substantial challenge in selecting an appropriate look-back window for risk modeling and forecasting. We develop a data-driven online learning method, called the bootstrap-based adaptive window selection (BAWS), that adaptively determines the window size in a sequential manner. A central component of BAWS is to compare the realized scores against a data-dependent threshold based on the bootstrap method. We provide an asymptotic justification for the bootstrap threshold, covering non-smooth scores such as the VaR check loss and the joint VaR--ES score, with an extension to stationary weakly dependent data via the moving block bootstrap. A single-break analysis further shows that BAWS rejects overlong windows crossing sufficiently large breaks. The proposed method is applicable to the forecasting of risk measures that are elicitable individually or jointly, such as the Value-at-Risk (VaR) and the pair of VaR and the corresponding Expected Shortfall. Through simulation studies and an empirical analysis, we demonstrate that BAWS often improves upon the standard rolling window approach and the recently developed method of stability-based adaptive window selection, especially when there are structural changes in the data-generating process.

q-fin.RM↗

arXiv2026

Online monotone density estimation and log-optimal calibration

We study the problem of online monotone density estimation, where density estimators must be constructed in a predictable manner from sequentially observed data. We propose two online estimators: an online analogue of the classical Grenander estimator, and an expert aggregation estimator inspired by exponential weighting methods from the online learning literature. In the well-specified stochastic setting, where the underlying density is monotone, we show that the expected cumulative log-likelihood gap between the online estimators and the true density admits an $O(n^{1/3})$ bound. We further establish a $\sqrt{n\log{n}}$ pathwise regret bound for the expert aggregation estimator relative to the best offline monotone estimator chosen in hindsight, under minimal regularity assumptions on the observed sequence. As an application of independent interest, we show that the problem of constructing log-optimal p-to-e calibrators for sequential hypothesis testing can be formulated as an online monotone density estimation problem. We adapt the proposed estimators to build empirically adaptive p-to-e calibrators and establish their optimality. Numerical experiments illustrate the theoretical results.

stat.ML↗

arXiv2026

Newsvendor under Ambiguity and Misspecification

Problem definition: We consider a newsvendor problem with unknown demand distribution, where we distinguish ambiguity under which the newsvendor does not differentiate demand distributions of common characteristics and misspecification under which such characteristics might be misspecified. Methodology/results: The newsvendor hedges against ambiguity and misspecification by maximizing the worst-case expected profit regularized by a distribution's distance to an ambiguity set. Focusing on the popular mean-variance ambiguity set and optimal-transport cost for the misspecification, we show that the decision criterion of misspecification aversion possesses insightful interpretations as distributional transforms. We derive the closed-form optimal order quantity that generalizes the solution of the Scarf model under only ambiguity aversion. We establish the finite-sample performance guarantee, which consists of two parts: in-sample optimal value and out-of-sample effect of misspecification that can be further decoupled into estimation error and distribution shift. We also extend the framework to multiple products, distributional characteristics specified via optimal transport, and misspecification measured by total variation distance. Managerial implications: The closed-form solution highlights the impact of misspecification aversion: the optimal order quantity under misspecification aversion can decrease as the price or variance increases, reversing the monotonicity of that under only ambiguity aversion. Hence, ambiguity and misspecification, as different layers of distributional uncertainty, can result in distinct operational consequences. The finite-sample performance guarantee theoretically justifies the necessity of incorporating misspecification aversion in a non-stationary environment, which is also well demonstrated in our experiments with real-world data.

math.OC↗

arXiv2026

E-backtesting

In the recent Basel Accords, the Expected Shortfall (ES) replaces the Value-at-Risk (VaR) as the standard risk measure for market risk in the banking sector, making it the most important risk measure in financial regulation. One of the most challenging tasks in risk modeling practice is to backtest ES forecasts provided by financial institutions. To design a model-free backtesting procedure for ES, we make use of the recently developed techniques of e-values and e-processes. Backtest e-statistics are introduced to formulate e-processes for risk measure forecasts, and unique forms of backtest e-statistics for VaR and ES are characterized using recent results on identification functions. For a given backtest e-statistic, a few criteria for optimally constructing the e-processes are studied. The proposed method can be naturally applied to many other risk measures and statistical quantities. We conduct extensive simulation studies and data analysis to illustrate the advantages of the model-free backtesting method, and compare it with the ones in the literature.

q-fin.RM↗

arXiv2026

Online LLM watermark detection via e-processes

Watermarking for large language models (LLMs) has emerged as an effective tool for distinguishing AI-generated text from human-written content. Statistically, watermark schemes induce dependence between generated tokens and a pseudo-random sequence, reducing watermark detection to a hypothesis testing problem on independence. We develop a unified framework for LLM watermark detection based on e-processes, providing anytime-valid guarantees for online testing. We propose various methods to construct empirically adaptive e-processes that can enhance the detection power. The proposed methods are applicable to any sequential testing problem where independent pivotal statistics are available. In addition, theoretical results are established to characterize the power properties of the proposed procedures. Some experiments demonstrate that the proposed framework achieves competitive performance compared to existing watermark detection methods.

stat.ME↗

arXiv2026

Submodular risk measures

We study submodularity for law-invariant functionals, with particular attention to convex risk measures. Expected losses are modular, and certainty equivalents are submodular exactly when the loss function is convex. Law-invariant coherent risk measures are submodular exactly when they are coherent distortion risk measures, including Expected Shortfall (ES), and several deviation measures are also submodular. Beyond positive homogeneity, submodularity is restrictive for convex risk measures. We give a complete characterization for shortfall risk measures via the Arrow--Pratt measure of risk aversion, show that optimized certainty equivalents are always submodular, and prove that adjusted Expected Shortfall (AES) is submodular only when it reduces to ES. An empirical illustration for daily US equity returns finds no ES submodularity violations, many Value-at-Risk (VaR) violations, and relatively few AES violations.

q-fin.RM↗

arXiv2026

Diversification Preferences and Risk Attitudes

Portfolio diversification is a cornerstone of modern finance, while risk aversion is central to decision theory; both concepts are long-standing and foundational. We investigate their connections by studying how different forms of diversification correspond to notions of risk aversion. We focus on the classical distinctions between weak and strong risk aversion, and consider diversification preferences for pairs of risks that are identically distributed, comonotonic, antimonotonic, independent, or exchangeable, as well as their intersections. Under a weak continuity condition and without assuming completeness of preferences, diversification for antimonotonic and identically distributed pairs implies weak risk aversion, and diversification for exchangeable pairs is equivalent to strong risk aversion. The implication from diversification for independent pairs to weak risk aversion requires a stronger continuity. We further provide results and examples that clarify the relationships between various diversification preferences and risk attitudes, in particular justifying the one-directional nature of many implications.

econ.TH↗

arXiv2026

Tiny but uniform improvements of adaptive BH procedures via compound e-values

After the seminal Benjamini-Hochberg (BH) procedure for controlling the false discovery rate (FDR) was proposed, dozens of papers have attempted to improve its power by adapting to the unknown proportion of nulls. We observe that most null proportion estimates are simply compound e-values in disguise, and thus most adaptive FDR procedures can be interpreted as instances of the e-weighted BH (ep-BH) procedure of Ignatiadis, Wang, and Ramdas [2024], i.e., the BH procedure weighted by compound e-values. This lens helps us show that most existing procedures are inadmissible, and we provide uniform improvements to them. While the improvements are small in practice, they still come for free (without additional assumptions), and help unify the literature. We also use our "leave-one-out ep-BH method" to design a new method with finite-sample FDR control for the simultaneous t-test setting.

stat.ME↗

arXiv2026

Wasserstein Distributionally Robust Quantile Regression

We study distributionally robust quantile regression using type-$p$ Wasserstein ambiguity sets. We derive a closed-form expression for the worst-case quantile regression loss under general $p$-Wasserstein uncertainty. We further give a uniqueness result showing that for $p>1$, the check loss yields the only class of convex loss functions for which such an additive Wasserstein regularization holds. Our analysis also uncovers qualitative differences between the regimes $p=1$ and $p>1$. When $p>1$, the slope coefficients coincide with those of the regularized formulation, while the intercept undergoes a radius-dependent adjustment; the value $p$ affects only this intercept correction, whereas the choice of transport norm influences both. Finally, we establish finite-sample out-of-sample risk guarantees of order $O(N^{-1/2})$ under mild moment conditions. Numerical experiments illustrate the theoretical findings and the practical implications of the proposed formulation.

math.ST↗

arXiv2026

Conformal e-prediction in the presence of confounding

This note extends conformal e-prediction to cover the case where there is observed confounding between the random object $X$ and its label $Y$. We consider both the case where the observed data is IID and a case where some dependence between observations is permitted.

math.ST↗

arXiv2026

Counter-monotonic Risk Sharing with Heterogeneous Distortion Risk Measures

We study risk sharing among agents with preferences modeled by heterogeneous distortion risk measures, who are not necessarily risk averse. Pareto optimality for agents using risk measures is often studied through the lens of inf-convolutions, because allocations that attain the inf-convolution are Pareto optimal, and the converse holds true under translation invariance. Our main focus is on groups of agents who exhibit varying levels of risk seeking. Under mild assumptions, we derive explicit solutions for the unconstrained inf-convolution and the counter-monotonic inf-convolution, which can be represented by a generalization of distortion risk measures.

q-fin.RM↗

arXiv2026

Allocation Mechanisms in Decentralized Exchange Markets with Frictions

The classical theory of efficient allocations of an aggregate endowment in a pure-exchange economy has hitherto primarily focused on the Pareto-efficiency of allocations, under the implicit assumption that transfers between agents are frictionless, and hence costless to the economy. In this paper, we argue that certain transfers cause frictions that result in costs to the economy. We show that these frictional costs are tantamount to a form of subadditivity of the cost of transferring endowments between agents. We suggest an axiomatic study of allocation mechanisms, that is, the mechanisms that transform feasible allocations into other feasible allocations, in the presence of such transfer costs. Among other results, we provide an axiomatic characterization of those allocation mechanisms that admit representations as robust (worst-case) linear allocation mechanisms, as well as those mechanisms that admit representations as worst-case conditional expectations. We call the latter Robust Conditional Mean Allocation mechanisms, and we relate our results to the literature on (decentralized) risk sharing within a pool of agents.

cs.GT↗

arXiv2026

On convex order and supermodular order without finite mean

Many results on the convex order in the literature were stated for random variables with finite mean. For instance, a fundamental result in dependence modeling is that the sum of a pair of random random variables is upper bounded in convex order by that of its comonotonic version and lower bounded by that of its counter-monotonic version, and all existing proofs of this result require the random variables' expectations to be finite. We show that the above result remains true even when discarding the finite-mean assumption, and obtain several other results on the comparison of infinite-mean random variables via the convex order. To our surprise, we find two deceivingly similar definitions of the convex order, both of which exist widely in the literature, and they are not equivalent for random variables with infinite mean. This subtle discrepancy in definitions also applies to the supermodular order, and it gives rise to some incorrect statements, often found in the literature.

math.PR↗

arXiv2026

Lambda Expected Shortfall

The Lambda Value-at-Risk (Lambda-VaR) is a generalization of the Value-at-Risk (VaR), which has been actively studied in quantitative finance. Over the past two decades, the Expected Shortfall (ES) has become one of the most important risk measures alongside VaR because of its various desirable properties in the practice of optimization, risk management, and financial regulation. Analogously to the intimate relation between ES and VaR, we introduce the Lambda Expected Shortfall (Lambda-ES), as a generalization of ES and a counterpart to Lambda-VaR. Our definition of Lambda-ES has an explicit formula and many convenient properties, and we show that it is the smallest quasi-convex and law-invariant risk measure dominating Lambda-VaR under mild assumptions. We examine further properties of Lambda-ES, its dual representation, and related optimization problems.

q-fin.MF↗

arXiv2025

Risk-insurance parity

Risk aversion and insurance are two prominent and interconnected concepts in economics and finance. To explore their fundamental connection, we introduce risk-insurance parity, which associates various classes of insurance contracts with different notions of risk aversion. We show that the classic notions -- both weak and strong -- of risk aversion can be characterized by propensity to different classes of insurance contracts, generalizing recent results on propensity to full, proportional, and deductible-limit contracts in the literature. We obtain full characterizations of the classes of insurance indemnity functions that correspond to weak and strong risk aversion. Risk-insurance parity allows us to define two new notions of risk aversion, between weak and strong, characterized by insurance propensity to deductible-only and limit-only contracts respectively.

econ.TH↗

arXiv2025

Aggregate then evaluate

We distinguish two frameworks for decisions under ambiguity: evaluate-then-aggregate (ETA) and aggregate-then-evaluate (ATE). Given a statistic that represents the decision maker's pure-risk preferences (such as expected utility) and an ambiguous act, an ETA model first evaluates the act under each plausible probabilistic model using this statistic and then aggregates the resulting evaluations according to ambiguity attitudes. In contrast, an ATE model first aggregates ambiguity by assigning the act a single representative distribution and then evaluates that distribution using the statistic. These frameworks differ in the order in which risk and ambiguity are processed, and they coincide when there is no ambiguity. While most existing ambiguity models fall within the ETA framework, our study focuses on the ATE framework, which is conceptually just as compelling and has been relatively neglected in the literature. We develop a Choquet ATE model, which generalizes the Choquet expected utility model by allowing arbitrary pure-risk preferences. We provide an axiomatization of this model in a Savage setting with an exogenous source of unambiguous events. The Choquet ATE framework allows us to analyze a wide range of ambiguity attitudes and their interplay with risk attitudes.

econ.TH↗

arXiv2025

Choquet rank-dependent utility with an exogenous unambiguous source

We axiomatize the Choquet rank-dependent utility model within a Savage framework with an exogenous source of pure risk. This model is a decision model under ambiguity, serving as a conceptual generalization of the Choquet expected utility model. The model unifies risk and ambiguity and reduces to the rank-dependent utility for pure risks. Our axiomatization uses two main axioms for biseparable preferences, along with some regularity axioms. A benefit of this axiomatization is that the fairly weak regularity axioms guarantee the existence of matching probabilities. Further, we discuss ambiguity attitudes for the CRDU model. We characterize these attitudes by properties of the associated matching probabilities and show that the supermodularity of the matching probability provides a robust representation. Finally, we show that under an additional Property, this model has a different representation using act-dependent distortion functions.

econ.TH↗