arXiv subjects

Peng Zeng

Publications and source records attributed to Peng Zeng.

At least 19 recordsLinked to original sources

arXiv2026

PCXR-Copilot: Device-aware pediatric chest radiographic analysis with segmentation, suppression, and edema-related cues

Patients admitted to pediatric ICU and cardiac ICU often require intensive perioperative support and frequent chest radiography to monitor support devices, extravascular lung water, and body-wall edema. Radiographic interpretation and analysis is often difficult due to superimposed anatomic and iatrogenic structures that obscure clinically relevant signals. While deep learning can provide decision support, translation to pediatric radiographs is still limited by high device burden and scarcity of pediatric datasets with clinically grounded device annotations. To address this gap, we first introduce PediaDeviceSeg-1K, a curated dataset of 1,000 anterior-posterior radiographs of pediatric CICU patients with clinically grounded, fine-grained annotations of common ICU devices. To our best knowledge, this is the first publicly available pediatric CICU CXR resource with clinically grounded device annotations for device-focused benchmarking. Building on PediaDeviceSeg-1K, we further present PCXR-Copilot, a device-aware pipeline that outputs lung and device masks, optional device-suppressed views, and edema-related visualization maps and scalar indices. The full pipeline has been integrated and runs in real time within the TWIST platform at Boston Children's Hospital, demonstrating feasibility for real-world clinical deployment.

cs.CV↗

arXiv2026

Memory Echo in Single-Agent Long-Term Memory: Lifecycle Evaluation and Provenance-Guided Mitigation

Long-term memory supports agents across sessions. Repeated retrieval and write-back can let one historical claim occupy multiple retrieval positions and influence current answers without independent evidence. Existing work examines retrieval, updating, and question answering, leaving same-source record accumulation and post-correction persistence underexplored. Motivated by research in social cognition and memory psychology on repeated exposure, retrieval, and revision, we define \emph{Memory Echo} as the excess retrieval exposure gained by historical information from the same source without additional independent evidence, and introduce MEchoBench to evaluate its emergence, correction resistance, persistence, and effect on answers. To address this phenomenon, we introduce the Family--Provenance--Authority framework (FPA). Specifically, FPA selects one representative per source family, conditions the eligibility of old records for the answer context on verified corrections, and constrains final answers with authoritative evidence. In paired replay on 190 fixed questions across four memory frameworks, FPA raises accuracy from 0.825 to 0.920 with DeepSeek V4 Flash and from 0.782 to 0.916 with Kimi K2.6; mean Decision Contamination Effect falls from 0.106 to 0.030 and from 0.109 to 0.038, respectively. These results show that agent memory can preserve traceable history without allowing repeated historical information to govern current decisions. The code is available at https://github.com/OrcasSam/memory-echo-benchmark.

cs.AI↗

arXiv2026

Relevance Is Not Sufficiency: What Actually Closes the Evidence Gap in Long-Term Memory QA

LLM agents that interact with a user across many sessions accumulate histories that exceed their context window, so they store past interactions in an external memory and answer each question from a small set of retrieved records. Existing memory systems rank records by lexical or embedding relevance, yet the top-ranked memories can each be relevant while jointly omitting a complementary fact that the answer requires, especially for multi-session and temporal questions. Drawing on the distinction between relevance and sufficiency in legal evidence scholarship, we recast memory retrieval as constructing a sufficient memory set. To operationalize this view, we introduce a blinded LLM judgment over the retrieved set, together with Gold Hit and Turn Hit as evidence-coverage proxies. We then propose Budgeted Flat Reconstruction (BFR), which builds sufficient sets over a fixed flat memory store in two stages. Specifically, we first apply Formal Concept Analysis for Memory Selection (FCA-MS) to decompose the question into information requirements and select a compact candidate subset that jointly covers them. Then, we repeatedly acquire unseen records through deeper text search or complementary entity and session views, stopping when the budget is exhausted. Experiments on LoCoMo and LongMemEval-S show that BFR outperforms same-store adaptations of recent agent-memory systems in both answer quality and evidence coverage. Specifically, on LongMemEval-S it raises judged accuracy from 72.4% to 82.2% and Turn Hit to 91.4%.

cs.AI↗

arXiv2026

High-Dimensional Statistical Inference for Sparse Support Vector Machines

Using a replica-symmetric high-dimensional characterization, we develop an inferential framework for sparse support vector machines when the sample size and number of features grow proportionally. The main challenge is the nonsmooth hinge loss, which prevents direct application of debiasing arguments developed for smooth classification losses. We overcome this difficulty by representing the $L_1$-penalized support vector machine (SVM) as a linear program and identifying the hinge-loss subgradient through its dual variables. This yields a computationally accessible debiased estimator whose coordinates are asymptotically Gaussian under the proportional asymptotic regime. The resulting distributional characterization provides confidence intervals and hypothesis tests for individual features and enables false-discovery-rate-controlled variable selection. Extensive simulations examine calibration, power, and variable-selection performance under a range of covariance structures, including strongly correlated designs. An analysis of high-dimensional breast cancer gene-expression data illustrates how the proposed inference can distinguish statistically significant features from variables selected by the original sparse SVM.

stat.ML↗

arXiv2026

Recoverability Is a Subspace Property: A Benchmark for Certified State Estimation from Partial PDE Observations

When reconstructing the hidden state of a partial differential equation (PDE) system from partial observations, aggregate prediction error measures performance on a given data distribution but does not reveal how strongly the observations constrain each predicted direction. We introduce UniPDE-Bench, a direction-wise evaluation protocol that incorporates local observation geometry into state-estimation assessment, providing a reference for prediction recovery and confidence-based selection that is independent of the estimator. The protocol whitens the joint observation Jacobian by the noise covariance and normalizes it by a state metric. Its complete right singular basis represents joint variations of the state, which a relative sensitivity threshold partitions into retained and below-threshold directions. In this common basis, the protocol evaluates recovery and the agreement between prediction claims and the geometric partition, while recovery and abstention curves describe confidence-based selection at different claim coverages. In the simulated tasks and observation configurations studied here, confidence rules whose overall ranking exceeds chance can still exhibit below-random abstention on below-threshold directions at some high claim coverages. By separating empirical recovery from direction-selection quality, the protocol relates prediction performance to local observation sensitivity and provides an evaluation of partially observed state estimators beyond aggregate error.

stat.ME↗

arXiv2026

Observation Design for Certified Control Authority: Projection--Estimability Separation and Active-Face Equivalence

A sound runtime admission gate executes only actions it can certify, and certifies only what its observations support. This paper asks how observations should be designed to maximize the set of actions that can be safely admitted, and shows the question is not a re-vocabulary of classical design problems. First, a projection--estimability separation: decomposing a constraint normal as $c=c_{\mathrm{Range}}+c_{\ker}$ relative to an information matrix, two-point discrimination along $c$ becomes arbitrarily reliable as the budget grows whenever $c_{\mathrm{Range}}\neq 0$, while robust admission of an action with normal $c$ is impossible at every budget whenever $c_{\ker}\neq 0$; such mixed directions are generic at any deficient rank, and at full rank the decoupling is bounded by the Kantorovich ratio and diverges with the condition number. Discrimination-optimal designs, being corner solutions of a linear criterion, land in exactly this regime. Second, an active-face equivalence theorem: the permissiveness-optimal design is $L$-optimal for a target matrix generated endogenously by the action faces that become certification bottlenecks, weighted inversely by their remaining slack; single-face collapse recovers $c$-optimal and goal-oriented design exactly, and a Caratheodory argument yields a bottleneck certificate of at most $r(r+1)/2+1$ faces. Around these we assemble exact certifiability per convex contract mode, for which $κ\approx 3.29$ is derived rather than calibrated, the $\sqrt{r}$ price of contract-agnostic design, an information-to-slack transfer theorem with a curvature-budget corollary, and a two-part audit in which a design meeting every margin requirement still leaves an action face at a certification cost above ten times its testing cost, in every probe library tested.

cs.RO↗

arXiv2026

Generator-Independent Runtime Assurance under Partial Observation

Proposal-based controllers---learned policies, language-model planners, and other black-box \emph{generators}---are increasingly deployed behind runtime verification gates. We ask when the closed-loop safety guarantee decouples from the generator. The prevailing per-candidate certification pattern does not compose: under retry or best-of-$k$ selection a per-candidate false-admission level $α$ can inflate to $1-(1-α)^{k}$. Our main theorem shows that \emph{simultaneous setwise soundness}---certifying a set of admissible proposals containing no nonviable action---is necessary and sufficient for generator-independent \emph{admission soundness}, the worst case over all generators of executing a nonviable proposal equalling the probability of setwise failure; together with a design-time certificate and a no-bypass rule it is sufficient for \emph{contract safety}, with violation bound $Γ+\sum_t\varepsilon_t+η$ invariant under arbitrary, even adversarial, replacement of the generator. A second theorem bounds every admission mechanism under partial observation: for a fixed probing and admission policy, if two state hypotheses whose information laws lie within total-variation distance $δ$ require different safe decisions, then $\abar+β+δ\ge1$. A sequential risk ledger makes the guarantee implementable with time-uniform confidence tubes, and shows that deterministic admission computations concentrate all statistical risk in state estimation. Simplex-style runtime assurance and control-barrier-function filtering are recovered as degenerate cases.

cs.AI↗

arXiv2026

ICO: Enhancing Semantic-Shift Jailbreaks via Iterative Context Optimization

Foundation models have achieved remarkable success across diverse tasks, but they remain vulnerable. To investigate such vulnerabilities, semantic-shift jailbreaks have recently emerged as a promising attack paradigm. They bypass explicit safety mechanisms by replacing harmful terms in original harmful questions with benign alternatives and leveraging contextual information to induce the target model to reinterpret these alternatives as their corresponding harmful concepts. However, existing semantic-shift jailbreaks often achieve limited effectiveness. In this work, we reveal that this limitation arises from overlooking the semantic-shift capability of contexts. Through systematic analysis, we find that contexts exhibit substantially different abilities in inducing semantic shifts: contexts with stronger semantic-shift capabilities are more likely to guide models toward recovering harmful meanings and achieving successful jailbreaks. Based on this finding, we systematically identify and distill the characteristics of effective contexts and propose a black-box context-aware semantic-shift jailbreak framework with Iterative Context Optimization (ICO). In each iteration, ICO leverages these characteristics and feedback from the target model to optimize contexts. Extensive experiments on three datasets and eight target foundation models demonstrate that ICO consistently outperforms eight state-of-the-art baselines, achieving an average attack success rate of 74.6%.

cs.CL↗

arXiv2025

A Bayesian Approach for Robust Longitudinal Envelope Models

The envelope model provides a dimension-reduction framework for multivariate linear regression. However, existing envelope methods typically assume normally distributed random errors and do not accommodate repeated measures in longitudinal studies. To address these limitations, we propose the robust longitudinal envelope model (RoLEM). RoLEM employs a scale mixture of matrix-variate normal distributions to model random errors, allowing it to handle potential outliers, and incorporates flexible correlation structures for repeated measurements. In addition, we introduce new prior and proposal distributions on the Grassmann manifold to facilitate Bayesian inference for RoLEM. Simulation studies and real data analysis demonstrate the superior performance of the proposed method.

stat.ME↗

arXiv2025

Logic-informed reinforcement learning for cross-domain optimization of large-scale cyber-physical systems

Cyber-physical systems (CPS) require the joint optimization of discrete cyber actions and continuous physical parameters under stringent safety logic constraints. However, existing hierarchical approaches often compromise global optimality, whereas reinforcement learning (RL) in hybrid action spaces often relies on brittle reward penalties, masking, or shielding and struggles to guarantee constraint satisfaction. We present logic-informed reinforcement learning (LIRL), which equips standard policy-gradient algorithms with projection that maps a low-dimensional latent action onto the admissible hybrid manifold defined on-the-fly by first-order logic. This guarantees feasibility of every exploratory step without penalty tuning. Experimental evaluations have been conducted across multiple scenarios, including industrial manufacturing, electric vehicle charging stations, and traffic signal control, in all of which the proposed method outperforms existing hierarchical optimization approaches. Taking a robotic reducer assembly system in industrial manufacturing as an example, LIRL achieves a 36.47\% to 44.33\% reduction at most in the combined makespan-energy objective compared to conventional industrial hierarchical scheduling methods. Meanwhile, it consistently maintains zero constraint violations and significantly surpasses state-of-the-art hybrid-action reinforcement learning baselines. Thanks to its declarative logic-based constraint formulation, the framework can be seamlessly transferred to other domains such as smart transportation and smart grid, thereby paving the way for safe and real-time optimization in large-scale CPS.

cs.LG↗

arXiv2024

Quench switching of Mn2As

We demonstrate that epitaxial thin film antiferromagnet Mn2As exhibits the quench-switching effect, which was previously reported only in crystallographically similar antiferromagnetic CuMnAs thin films. Quench switching in Mn2As shows stronger increase in resistivity, reaching hundreds of percent at 5K, and significantly longer retention time of the metastable high-resistive state before relaxation towards the low-resistive uniform magnetic state. Qualitatively, Mn2As and CuMnAs show analogous parametric dependence of the magnitude and relaxation of the quench-switching signal. Quantitatively, relaxation dynamics in both materials show direct proportionality to the Néel temperature. This confirms that the quench switching has magnetic origin in both materials. The presented results suggest that the antiferromagnets crystalizing in the Cu2Sb structure are well suited for exploring and exploiting the intriguing physics of highly non-uniform magnetic states associated with the quench switching.

cond-mat.mtrl-sci↗

arXiv2024

Statistical Inference in Classification of High-dimensional Gaussian Mixture

We consider the classification problem of a high-dimensional mixture of two Gaussians with general covariance matrices. Using the replica method from statistical physics, we investigate the asymptotic behavior of a general class of regularized convex classifiers in the high-dimensional limit, where both the sample size $n$ and the dimension $p$ approach infinity while their ratio $α=n/p$ remains fixed. Our focus is on the generalization error and variable selection properties of the estimators. Specifically, based on the distributional limit of the classifier, we construct a de-biased estimator to perform variable selection through an appropriate hypothesis testing procedure. Using $L_1$-regularized logistic regression as an example, we conducted extensive computational experiments to confirm that our analytical findings are consistent with numerical simulations in finite-sized systems. We also explore the influence of the covariance structure on the performance of the de-biased estimator.

stat.ML↗

arXiv2024

Construction Methods Based on Minimum Weight Distribution for Polar Codes with Successive Cancellation List Decoding

Minimum weight distribution (MWD) is an important metric to calculate the first term of union bound called minimum weight union bound (MWUB). In this paper, we first prove the maximum likelihood (ML) performance approaches MWUB as signal-to-noise ratio (SNR) goes to infinity and provide the deviation when MWD and SNR are given. Then, we propose a nested reliability sequence, namely MWD sequence, to construct polar codes independently of channel information. In the sequence, synthetic channels are sorted by partial MWD which is used to evaluate the influence of information bit on MWD and we prove the MWD sequence is the optimum sequence evaluated by MWUB for polar codes obeying partial order. Finally, we introduce an entropy constraint to establish a relationship between list size and MWUB and propose a heuristic construction method named entropy constraint bit-swapping (ECBS) algorithm, where we initialize information set by the MWD sequence and gradually swap information bit and frozen bit to satisfy the entropy constraint. The simulation results show the MWD sequence is more suitable for constructing polar codes with short code length than the polar sequence in 5G and the ECBS algorithm can improve MWD to show better performance as list size increases.

cs.IT↗

arXiv2024

Development of a Nb-based semiconductor-superconductor hybrid platform

Semiconductor-superconductor hybrid materials are used as a platform to realise Andreev bound states, which hold great promise for quantum applications. These states require transparent interfaces between the semiconductor and superconductor, which are typically realised by in-situ deposition of an Al superconducting layer. Here we present a hybrid material based on an InAs two-dimensional electron gas (2DEG) combined with in-situ deposited Nb and NbTi superconductors, which offer a larger operating range in temperature and magnetic field due to their larger superconducting gap. We overcome the inherent difficulty associated with the formation of an amorphous interface between III-V semiconductors and Nb-based superconductors by introducing a 7 nm Al interlayer. The Al interlayer provides an epitaxial connection between an in-situ magnetron sputtered Nb or NbTi thin film and a shallow InAs 2DEG. This metal-to-metal epitaxy is achieved by optimization of the material stack and results in an induced superconducting gap of approximately 1 meV, determined from transport measurements of superconductor-semiconductor Josephson junctions. This induced gap is approximately five times larger than the values reported for Al-based hybrid materials and indicates the formation of highly-transparent interfaces that are required in high-quality hybrid material platforms.

cond-mat.supr-con↗

arXiv2023

Performance Analysis for Polar Codes under Successive Cancellation List Decoding with Fixed List Size

In this paper, we first indicate that the block error event of polar codes under successive cancellation list (SCL) decoding is composed of path loss (PL) error event and path selection (PS) error event, where the PL error event is that correct codeword is lost during the SCL decoding and the PS error event is that correct codeword is reserved in the decoded list but not selected as the decoded codeword. Then, we simplify the PL error event by assuming the all-zero codeword is transmitted and derive the probability lower bound via the joint probability density of the log-likelihood ratios of information bits. Meanwhile, the union bound calculated by the minimum weight distribution is used to evaluate the probability of the PS error event. With the performance analysis, we design a greedy bit-swapping (BS) algorithm to construct polar codes by gradually swapping information bit and frozen bit to reduce the performance lower bound of SCL decoding. The simulation results show that the BLER performance of SCL decoding is close to the lower bound in the medium to high signal-to-noise ratio region and we can optimize the lower bound to improve the BLER performance of SCL decoding by the BS algorithm.

cs.IT↗

arXiv2023

Extended ADMM for general penalized quantile regression with linear constraints in big data

Quantile regression (QR) can be used to describe the comprehensive relationship between a response and predictors. Prior domain knowledge and assumptions in application are usually formulated as constraints of parameters to improve the estimation efficiency. This paper develops methods based on multi-block ADMM to fit general penalized QR with linear constraints of regression coefficients. Different formulations to handle the linear constraints and general penalty are explored and compared. The most efficient one has explicit expressions for each parameter and avoids nested-loop iterations in some existing algorithms. Additionally, parallel ADMM algorithm for big data is also developed when data are stored in a distributed fashion. The stopping criterion and convergence of the algorithm are established. Extensive numerical experiments and a real data example demonstrate the computational efficiency of the proposed algorithms. The details of theoretical proofs and different algorithm variations are presented in Appendix.

stat.CO↗

arXiv2023

Development of Nb-GaAs based superconductor semiconductor hybrid platform by combining in-situ dc magnetron sputtering and molecular beam epitaxy

We present Nb thin films deposited in-situ on GaAs by combining molecular beam epitaxy and magnetron sputtering within an ultra-high vacuum cluster. Nb films deposited at varying power, and a reference film from a commercial system, are compared. The results show clear variation between the in-situ and ex-situ deposition which we relate to differences in magnetron sputtering conditions and chamber geometry. The Nb films have critical temperatures of around $9 \textrm{K}$. and critical perpendicular magnetic fields of up to $B_{c2} = 1.4 \textrm{T}$ at $4.2 \textrm{K}$. From STEM images of the GaAs-Nb interface we find the formation of an amorphous interlayer between the GaAs and the Nb for both the ex-situ and in-situ deposited material.

cond-mat.mtrl-sci↗

arXiv2023

Differentiable Genetic Programming for High-dimensional Symbolic Regression

Symbolic regression (SR) is the process of discovering hidden relationships from data with mathematical expressions, which is considered an effective way to reach interpretable machine learning (ML). Genetic programming (GP) has been the dominator in solving SR problems. However, as the scale of SR problems increases, GP often poorly demonstrates and cannot effectively address the real-world high-dimensional problems. This limitation is mainly caused by the stochastic evolutionary nature of traditional GP in constructing the trees. In this paper, we propose a differentiable approach named DGP to construct GP trees towards high-dimensional SR for the first time. Specifically, a new data structure called differentiable symbolic tree is proposed to relax the discrete structure to be continuous, thus a gradient-based optimizer can be presented for the efficient optimization. In addition, a sampling method is proposed to eliminate the discrepancy caused by the above relaxation for valid symbolic expressions. Furthermore, a diversification mechanism is introduced to promote the optimizer escaping from local optima for globally better solutions. With these designs, the proposed DGP method can efficiently search for the GP trees with higher performance, thus being capable of dealing with high-dimensional SR. To demonstrate the effectiveness of DGP, we conducted various experiments against the state of the arts based on both GP and deep neural networks. The experiment results reveal that DGP can outperform these chosen peer competitors on high-dimensional regression benchmarks with dimensions varying from tens to thousands. In addition, on the synthetic SR problems, the proposed DGP method can also achieve the best recovery rate even with different noisy levels. It is believed this work can facilitate SR being a powerful alternative to interpretable ML for a broader range of real-world problems.

cs.NE↗