arXiv subjects

Zhi-Hai Zhang

Publications and source records attributed to Zhi-Hai Zhang.

10 recordsLinked to original sources

arXiv2026

Which Model Is Actually Serving You? IRIS: Budgeted Black-Box Auditing of Model Substitution and Routing Dilution in LLM Gateways

Commercial LLM gateways mediate access to hosted models, but the served backend may not match the advertised one: it may substitute a cheaper model on every request or route only a fraction $ε$ of requests to it. Prior black-box auditors often need a privileged signal (log-probabilities, token ranks, or reference samples) or a target-specific probe, fix the query budget in advance, and return a yes/no verdict. We present $\mathrm{IRIS}$, an audit that needs only the returned text: it asks endpoints to generate random numbers or strings, fingerprints the backend, and is the first to combine, in one text-only audit, detection of whole-stream substitution and fractional dilution, attribution of the served backend, routing-fraction ($ε$) estimation, and a query budget it sizes itself. A cheap pilot fits the exponential query-error decay and freezes that budget before any suspect query is issued. On an intra-family Qwen3 ladder $\mathrm{IRIS}$ verifies the backend at $0.99$ AUROC and sharpens attribution as queries accumulate; across a commercial OpenRouter library it catches $ε{=}0.3$ dilution on margin-qualified pairs at $0.85$ mean power ($0.017$ false-positive rate) and recovers $ε$ to within $0.04$ for enrolled diluents; and a live cross-provider audit flags $14$ of $15$ same-model provider pairs by genuine quantization and kernel deviations, corroborated on third-party MET traces. Against comparable black-box auditors, $\mathrm{IRIS}$ matches or beats detection on shared tasks, and adaptive allocation lifts the matched-budget target-hit rate from $73$% to $87$%. Further experiments cover adversarial gateways, knob identifiability, unseen diluents, and false-positive control.

cs.CR↗

arXiv2026

Solving contextual chance-constrained programming under decision-dependent uncertainty

We study contextual chance-constrained programming under decision-dependent uncertainty. In this setting, a decision not only needs to satisfy constraints but also alters the distribution of uncertain outcomes. This dependency makes the problem particularly difficult: because feasibility probabilities vary with decisions, it creates both statistical endogeneity and computational intractability. To address this, we propose a nonparametric approximation method based on Contextual Cluster Weights (CCW). For any given decision and context, CCW constructs a local neighborhood (cluster) of ``similar" historical observations and assigns them equal weight. This approach successfully renders both the objective and chance constraints tractable, while providing uniform-in-decision consistency guarantees. Furthermore, we develop reformulations that use pre-calculated clusters. We show that under a specific nestedness condition, these reformulations yield a convex feasible region, which allows for efficient solving. Experiments, including a case study with JD.com, demonstrate that our method outperforms benchmarks in solution quality, feasibility reliability, and runtime. This framework offers a scalable and data-driven approach for firms to make reliable operational decisions when their actions influence uncertainty. It effectively balances performance, risk, and robustness, while remaining interpretable and implementable in practice.

math.OC↗

arXiv2024

Achieving Robust Data-driven Contextual Decision Making in a Data Augmentation Way

This paper focuses on the contextual optimization problem where a decision is subject to some uncertain parameters and covariates that have some predictive power on those parameters are available before the decision is made. More specifically, we focus on solving the Wasserstein-distance-based distributionally robust optimization (DRO) model for the problem, which maximizes the worst-case expected objective over an uncertainty set including all distributions closed enough to a nominal distribution with respect to the Wasserstein distance. We develop a stochastic gradient descent algorithm based on the idea of data augmentation to solve the model efficiently. The algorithm iteratively a) does a bootstrapping sample from the nominal distribution; b) perturbs the adversarially and c) updates decisions. Accordingly, the computational time of the algorithm is only determined by the number of iterations and the complexity of computing the gradient of a single sample. Except for efficiently solving the model, the algorithm provide additional advantages that the proposed algorithm can cope with any nominal distributions and therefore is extendable to solve the problem in an online setting. We also prove that the algorithm converges to the optimal solution of the DRO model at a rate of a $O(1/\sqrt{T})$, where $T$ is the number of iterations of bootstrapping. Consequently, the performance guarantee of the algorithm is that of the DRO model plus $O(1/\sqrt{T})$. Through extensive numerical experiments, we demonstrate the superior performance of the proposed algorithm to several benchmarks.

math.OC↗

arXiv2024

Generalizing Few Data to Unseen Domains Flexibly Based on Label Smoothing Integrated with Distributionally Robust Optimization

Overfitting commonly occurs when applying deep neural networks (DNNs) on small-scale datasets, where DNNs do not generalize well from existing data to unseen data. The main reason resulting in overfitting is that small-scale datasets cannot reflect the situations of the real world. Label smoothing (LS) is an effective regularization method to prevent overfitting, avoiding it by mixing one-hot labels with uniform label vectors. However, LS only focuses on labels while ignoring the distribution of existing data. In this paper, we introduce the distributionally robust optimization (DRO) to LS, achieving shift the existing data distribution flexibly to unseen domains when training DNNs. Specifically, we prove that the regularization of LS can be extended to a regularization term for the DNNs parameters when integrating DRO. The regularization term can be utilized to shift existing data to unseen domains and generate new data. Furthermore, we propose an approximate gradient-iteration label smoothing algorithm (GI-LS) to achieve the findings and train DNNs. We prove that the shift for the existing data does not influence the convergence of GI-LS. Since GI-LS incorporates a series of hyperparameters, we further consider using Bayesian optimization (BO) to find the relatively optimal combinations of these hyperparameters. Taking small-scale anomaly classification tasks as a case, we evaluate GI-LS, and the results clearly demonstrate its superior performance.

cs.LG↗

arXiv2013

Exact periodic and solitonic states in the spinor condensates

We propose a method to analytically solve the one-dimensional coupled nonlinear Gross-Pitaevskii equations which govern the motion of the spinor Bose-Einstein condensates. In a uniform external potential, the Hamiltonian comprises the kinetic energy, the linear and the quadratic Zeeman energies. Several classes of exact periodic and solitonic solutions, either in real or in complex forms, are obtained for both the F=1 and F=2 condensates. These solutions are general that contain neither approximations nor constraints on the system parameters.

cond-mat.quant-gas↗

arXiv2013

Exact temporal evolution of the two-species Bose-Einstein condensates

We construct exact stationary solutions to the one-dimensional coupled Gross-Pitaevskii equations for the two-species Bose-Einstein condensates with equal intraspecies and interspecies interaction constants. Three types of complex solutions as well as their soliton limits are derived. By making use of the SU(2) unitary symmetry, we further obtain analytical time-evolving solutions. These solutions exhibit spatiotemporal periodicity.

cond-mat.quant-gas↗

arXiv2012

Exact solutions to the spin-2 Gross-Pitaevskii equations

We present several exact solutions to the coupled nonlinear Gross-Pitaevskii equations which describe the motion of the one-dimensional spin-2 Bose-Einstein condensates. The nonlinear density-density interactions are decoupled by making use of the properties of Jacobian elliptical functions. The distinct time factors in each hyperfine state implies a "Lamor" procession in these solutions. Furthermore, exact time-evolving solutions to the time-dependent Gross-Pitaevskii equations are constructed through the spin-rotational symmetry of the Hamiltonian. The spin-polarizations and density distributions in the spin-space are analyzed.

cond-mat.quant-gas↗

arXiv2012

Impurity effects on the quantum coherence of a few-boson system

The impurity effects on the quantum coherence of a few-boson system are studied within the two-site Hubbard model. Periodical collapses and revivals of coherence occur in the presence of either polarized or unpolarized fermionic impurities. The relative strength $U_{BF}/U_{BB}$ of the boson-fermion versus the boson-boson interactions plays a key role in the coherence revivals. As the average filling of the impurity increases, the coherence revivals remain nearly unaffected for $U_{BF}/U_{BB}=z$ ($z\in Z$ is an integer) while the odd revival peaks are damped for $U_{BF}/U_{BB}=z+1/2$, in agreement with the experimental observations. For unpolarized fermionic impurities, the coherence revivals are irrelevant to the strength of the fermion-fermion interactions.

cond-mat.quant-gas↗

arXiv2012

Analytical solutions to the spin-1 Bose-Einstein condensates

We analytically solve the one-dimensional coupled Gross-Pitaevskii equations which govern the motion of F=1 spinor Bose-Einstein condensates. The nonlinear density-density interactions are decoupled by making use of the unique properties of the Jacobian elliptical functions. Several types of complex stationary solutions are deduced. Furthermore, exact non-stationary solutions to the time-dependent Gross-Pitaevskii equations are constructed by making use of the spin-rotational symmetry of the Hamiltonian. The spin-polarizations exhibit kinked configurations. Our method is applicable to other coupled nonlinear systems.

cond-mat.quant-gas↗

arXiv2012

Disorder effects on the quantum coherence of a many-boson system

The effects of disorders on the quantum coherence for many-bosons are studied in a double well model. For the ground state, the disorder enhances the quantum coherence. In the deep Mott regime, dynamical evolution reveals periodical collapses and revivals of the quantum coherence which is robust against the disorder. The average over variations in both the on-site energy and the interaction reveals a beat phenomenon of the coherence-decoherence oscillation in the temporal evolution.

cond-mat.quant-gas↗