arXiv subjects

Zhe Yu

Publications and source records attributed to Zhe Yu.

At least 37 records · Page 2Linked to original sources

arXiv2026

From Retinal Evidence to Safe Decisions: RETINA-SAFE and ECRT for Hallucination Risk Triage in Medical LLMs

Hallucinations in medical large language models (LLMs) remain a safety-critical issue, particularly when available evidence is insufficient or conflicting. We study this problem in diabetic retinopathy (DR) decision settings and introduce RETINA-SAFE, an evidence-grounded benchmark aligned with retinal grading records, comprising 12,522 samples. RETINA-SAFE is organized into three evidence-relation tasks: E-Align (evidence-consistent), E-Conflict (evidence-conflicting), and E-Gap (evidence-insufficient). We further propose ECRT (Evidence-Conditioned Risk Triage), a two-stage white-box detection framework: Stage 1 performs Safe/Unsafe risk triage, and Stage 2 refines unsafe cases into contradiction-driven versus evidence-gap risks. ECRT leverages internal representation and logit shifts under CTX/NOCTX conditions, with class-balanced training for robust learning. Under evidence-grouped (not patient-disjoint) splits across multiple backbones, ECRT provides strong Stage-1 risk triage and explicit subtype attribution, improves Stage-1 balanced accuracy by +0.15 to +0.19 over external uncertainty and self-consistency baselines and by +0.02 to +0.07 over the strongest adapted supervised baseline, and consistently exceeds a single-stage white-box ablation on Stage-1 balanced accuracy. These findings support white-box internal signals grounded in retinal evidence as a practical route to interpretable medical LLM risk triage.

cs.AI↗

arXiv2026

LatentAudit: Real-Time White-Box Faithfulness Monitoring for Retrieval-Augmented Generation with Verifiable Deployment

Retrieval-augmented generation (RAG) mitigates hallucination but does not eliminate it: a deployed system must still decide, at inference time, whether its answer is actually supported by the retrieved evidence. We introduce LatentAudit, a white-box auditor that pools mid-to-late residual-stream activations from an open-weight generator and measures their Mahalanobis distance to the evidence representation. The resulting quadratic rule requires no auxiliary judge model, runs at generation time, and is simple enough to calibrate on a small held-out set. We show that residual-stream geometry carries a usable faithfulness signal, that this signal survives architecture changes and realistic retrieval failures, and that the same rule remains amenable to public verification. On PubMedQA with Llama-3-8B, LatentAudit reaches 0.942 AUROC with 0.77,ms overhead. Across three QA benchmarks and five model families (Llama-2/3, Qwen-2.5/3, Mistral), the monitor remains stable; under a four-way stress test with contradictions, retrieval misses, and partial-support noise, it reaches 0.9566--0.9815 AUROC on PubMedQA and 0.9142--0.9315 on HotpotQA. At 16-bit fixed-point precision, the audit rule preserves 99.8% of the FP16 AUROC, enabling Groth16-based public verification without revealing model weights or activations. Together, these results position residual-stream geometry as a practical basis for real-time RAG faithfulness monitoring and optional verifiable deployment.

cs.AI↗

arXiv2026

UniCA: Unified Covariate Adaptation for Time Series Foundation Model

Time Series Foundation Models (TSFMs) have achieved remarkable success through large-scale pretraining. However, their design primarily targets real-valued series, limiting their ability to handle general forecasting tasks involving diverse and often heterogeneous covariates -- such as categorical variables and multimodal data (e.g., images, text) -- which are typically task-specific and difficult to leverage during pretraining. To address this gap, we propose Unified Covariate Adaptation (UniCA), a framework to bridge TSFMs with general covariate-aware forecasting. UniCA first performs covariate homogenization to transform heterogeneous covariates into high-level homogeneous series representations and then fuses them via a unified attention-based fusion mechanism. UniCA is compatible and universal for adaptation with both homogeneous and heterogeneous covariates, incorporating extra covariate information while preserving the generalization ability of TSFMs.Extensive experiments on multiple unimodal and multimodal covariate-aware forecasting benchmarks demonstrate the superiority of UniCA, highlighting the promise of covariate-aware TSFM adaptation in real-world forecasting scenarios.Code: https://github.com/hanlu-nju/UniCA.

cs.LG↗

arXiv2026

Modeling Image-Caption Rating from Comparative Judgments

Image caption rating is becoming increasingly important because computer-generated captions are used extensively for descriptive annotation. However, rating the accuracy of captions in describing images is time-consuming and subjective in nature. In contrast, it is often easier for people to compare (between two pairs) which image-caption pair better matches each other. In this study, we propose a machine learning framework that models such comparative judgments instead of direct ratings. The model can then be applied to rank unseen image-caption pairs in the same way as a regression model trained on direct ratings. Inspired by a state-of-the-art regression approach, we extracted visual and text features using a pre-trained ViLBERT model and tweaked the learning parameters of the baseline model to improve the model performance. This new regression model (with Kendall's $τ_c=0.812$) outperformed the baseline model (with Kendall's $τ_c=0.758$) on the VICR dataset. The same model structure was applied to the comparative learning framework. Trained on comparative judgments (image-caption pair A better matches each other than image-caption pair B), the comparative learning model achieved a performance similar (with Kendall's $τ_c=0.804$) to that of the regression model. In addition, a small-scale human subject study was conducted to compare the cost and quality of direct ratings, pairwise comparisons, and same-image comparisons. The results showed that comparative judgments yielded faster results and greater agreement among human annotators than direct ratings. These results suggest that collecting comparative judgments instead of direct ratings as training data labels is promising for lower annotation costs and greater consistency. The model trained on such comparative judgments can perform as well as the model trained on direct ratings.

cs.CV↗

arXiv2026

Efficient Story Point Estimation With Comparative Learning

Story points are unitless, project-specific effort estimates that help developers plan their sprints. Traditionally, developers have collaboratively estimated story points using planning poker or other manual techniques. Machine learning can reduce this burden, but only with sufficient context from the historical decisions made by the project team. That is, state-of-the-art models, such as GPT2SP and FastText-SVM, only make accurate (within-project) predictions when they are trained on data from the same project. The goal of this study is to streamline story point estimation by evaluating a comparative learning-based framework for calibrating project-specific story point prediction models. Instead of assigning a specific story point value to every backlog item, developers are presented with pairs of items and asked to indicate which item requires more effort. Using these comparative judgments, a machine learning model was trained to predict the story point estimates. We empirically evaluated our technique using data from 23,313 manual estimates across 16 projects. The model trained on comparative judgments achieved, on average, a 0.34 Spearman's rank correlation coefficient between its predictions and the ground truth story points. This is similar to, if not better than, the performance of a state-of-the-art regression model trained on ground truth story points. Through human subject experiments, the advantages of comparative judgments were validated - higher confidence, lower annotation time, and comparable agreement were observed for comparative judgments compared to direct ratings. In summary, the proposed comparative learning approach is more efficient than regression-based approaches, given its better performance, lower required annotation time, and higher training data reliability.

cs.AI↗

arXiv2026

Modeling Art Evaluations from Comparative Judgments: A Deep Learning Approach to Predicting Aesthetic Preferences

Modeling human aesthetic judgments in visual art presents significant challenges due to individual preference variability and the high cost of obtaining labeled data. To reduce cost of acquiring such labels, we propose to apply a comparative learning framework based on pairwise preference assessments rather than direct ratings. This approach leverages the Law of Comparative Judgment, which posits that relative choices exhibit less cognitive burden and greater cognitive consistency than direct scoring. We extract deep convolutional features from painting images using ResNet-50 and develop both a deep neural network regression model and a dual-branch pairwise comparison model. We explored four research questions: (RQ1) How does the proposed deep neural network regression model with CNN features compare to the baseline linear regression model using hand-crafted features? (RQ2) How does pairwise comparative learning compare to regression-based prediction when lacking access to direct rating values? (RQ3) Can we predict individual rater preferences through within-rater and cross-rater analysis? (RQ4) What is the annotation cost trade-off between direct ratings and comparative judgments in terms of human time and effort? Our results show that the deep regression model substantially outperforms the baseline, achieving up to $328\%$ improvement in $R^2$. The comparative model approaches regression performance despite having no access to direct rating values, validating the practical utility of pairwise comparisons. However, predicting individual preferences remains challenging, with both within-rater and cross-rater performance significantly lower than average rating prediction. Human subject experiments reveal that comparative judgments require $60\%$ less annotation time per item, demonstrating superior annotation efficiency for large-scale preference modeling.

cs.CV↗

arXiv2026

Comparative Separation: Evaluating Separation on Comparative Judgment Test Data

This research seeks to benefit the software engineering society by proposing comparative separation, a novel group fairness notion to evaluate the fairness of machine learning software on comparative judgment test data. Fairness issues have attracted increasing attention since machine learning software is increasingly used for high-stakes and high-risk decisions. It is the responsibility of all software developers to make their software accountable by ensuring that the machine learning software do not perform differently on different sensitive groups -- satisfying the separation criterion. However, evaluation of separation requires ground truth labels for each test data point. This motivates our work on analyzing whether separation can be evaluated on comparative judgment test data. Instead of asking humans to provide the ratings or categorical labels on each test data point, comparative judgments are made between pairs of data points such as A is better than B. According to the law of comparative judgment, providing such comparative judgments yields a lower cognitive burden for humans than providing ratings or categorical labels. This work first defines the novel fairness notion comparative separation on comparative judgment test data, and the metrics to evaluate comparative separation. Then, both theoretically and empirically, we show that in binary classification problems, comparative separation is equivalent to separation. Lastly, we analyze the number of test data points and test data pairs required to achieve the same level of statistical power in the evaluation of separation and comparative separation, respectively. This work is the first to explore fairness evaluation on comparative judgment test data. It shows the feasibility and the practical benefits of using comparative judgment test data for model evaluations.

cs.SE↗

arXiv2026

Impact of particle production during inflation on the CMB detection

This work focuses on particle production described by a nonminimally coupled model during inflation. In this model, three parameters determine the characteristic frequency and strength of the induced gravitational waves (GWs). Considering the impact of particle production on inflation, we identify the parameter values that generate the strongest GWs without violating the slow-roll mechanism at the CMB scale. However, even with such extreme parameters, the power spectrum of induced GWs is only about $0.3\%$ of that of vacuum GWs. This contribution remains insignificant when identifying the primary source of the detected CMB B-mode polarization. Furthermore, when our results are integrated with the constraints driven by P+ACT+LB+BK18, the contribution of induced GWs at CMB scales becomes negligible. In contrast, their impact on the scalar spectral index $n_s$ proves significant. For a range of parameter values, the Starobinsky inflation model yields predictions for $n_s$ that are consistent with the measurements obtained from P+ACT+LB+BK18.

astro-ph.CO↗

arXiv2025

Isolated singularities for elliptic equations with convolution terms in a punctured ball

The purpose of this article is two-fold. First, we investigate the inequality $$ -Δu+V(x) u\geq f\quad\mbox{ in } B_1\setminus\{0\}\subset \mathbb{R}^N , N \geq 2, $$ where $f\in L^1_{loc}(B_1)$. If $V\geq 0$ is radially symmetric, we provide optimal conditions for which any solution $0\leq u\in \mathcal{C}^2(B_1\setminus\{0\})$ of the above inequality satisfies $u, Δu, V(x)u\in L^1_{loc}(B_1)$. This extends a result of H. Brezis and P.-L. Lions (1982), originally established for constant potentials $V$. Second, we investigate the equation $$\displaystyle -Δu + λV(x) u = (K_{α, β} * u^p) u^q \quad\text{in } B_1 \setminus \{0\},$$ where $0\leq V\in \mathcal{C}^{0, ν}( \overline B_1\setminus\{0\})$, $0<ν<1$, $λ, p, q>0$ and $$K_{α, β}(x) = |x|^{-α}\log^β\frac{2e}{|x|}, \quad\text{where } 0 \leq α< N, β\in \mathbb{R}.$$ For $N \geq 3$, we establish sharp conditions on the exponents $α, β, p, q$ under which singular solutions exist and exhibit the asymptotic behavior $u(x) \simeq |x|^{2-N}$ near the origin. For $N = 2$, we provide a classification of the existence and boundedness of solutions based on the local behavior of the potential $V(x)$ near the origin.

math.AP↗

arXiv2025

FairReweighing: Density Estimation-Based Reweighing Framework for Improving Separation in Fair Regression

There has been a prevalence of applying AI software in both high-stakes public-sector and industrial contexts. However, the lack of transparency has raised concerns about whether these data-informed AI software decisions secure fairness against people of all racial, gender, or age groups. Despite extensive research on emerging fairness-aware AI software, up to now most efforts to solve this issue have been dedicated to binary classification tasks. Fairness in regression is relatively underexplored. In this work, we adopted a mutual information-based metric to assess separation violations. The metric is also extended so that it can be directly applied to both classification and regression problems with both binary and continuous sensitive attributes. Inspired by the Reweighing algorithm in fair classification, we proposed a FairReweighing pre-processing algorithm based on density estimation to ensure that the learned model satisfies the separation criterion. Theoretically, we show that the proposed FairReweighing algorithm can guarantee separation in the training data under a data independence assumption. Empirically, on both synthetic and real-world data, we show that FairReweighing outperforms existing state-of-the-art regression fairness solutions in terms of improving separation while maintaining high accuracy.

cs.LG↗

arXiv2025

The parameterized quasinormal modes for modified Teukolsky equations

We introduce the modified Teukolsky equation within a parameterized framework, analogous to the case of small deviations of potential in spherical symmetry. Both the radial and angular equations acquire modifications described by two independent sets of parameters. We derive the parameterized framework of the quasinormal mode spectra using the continued fraction method. The results are cross-validated with the two-dimensional pseudo-spectral method, demonstrating excellent agreement and ensuring self-consistency. This work establishes a robust foundation for a theory-agnostic interpretation of gravitational-wave ringdown signals, providing a practical tool for probing potential deviations from General Relativity in the strong-field regime.

gr-qc↗

arXiv2025

Mano Technical Report

Graphical user interfaces (GUIs) are the primary medium for human-computer interaction, yet automating GUI interactions remains challenging due to the complexity of visual elements, dynamic environments, and the need for multi-step reasoning. Existing methods based on vision-language models (VLMs) often suffer from limited resolution, domain mismatch, and insufficient sequential decisionmaking capability. To address these issues, we propose Mano, a robust GUI agent built upon a multi-modal foundation model pre-trained on extensive web and computer system data. Our approach integrates a novel simulated environment for high-fidelity data generation, a three-stage training pipeline (supervised fine-tuning, offline reinforcement learning, and online reinforcement learning), and a verification module for error recovery. Mano demonstrates state-of-the-art performance on multiple GUI benchmarks, including Mind2Web and OSWorld, achieving significant improvements in success rate and operational accuracy. Our work provides new insights into the effective integration of reinforcement learning with VLMs for practical GUI agent deployment, highlighting the importance of domain-specific data, iterative training, and holistic reward design.

cs.MM↗

arXiv2025

Rescattering of non-minimal coupling scalar particles during inflation

We investigate the rescattering effects arising from non-minimally coupled scalar particles $χ$ that are suddenly produced during inflation. The coupling term $ξR χ^2$ significantly enhances resonant particle production compared to minimal coupling scenarios. Consequently, the produced $χ$ particles rescattering off the homogeneous inflaton condensate $ϕ$, generating abundant $δϕ$ quanta within very short time intervals. This process leads to characteristic enhancements in the power spectrum of primordial curvature perturbations at scales corresponding to the moments of particle production. When this occurs at small scales, the power spectrum amplitude can reach as high as $\mathcal{O}(10^{-2})$. Furthermore, analysis of the equilateral bispectrum shows that this mechanism also induces substantial non-Gaussian features.

astro-ph.CO↗

arXiv2025

Cross-Border Legal Adaptation of Autonomous Vehicle Design based on Logic and Non-monotonic Reasoning

This paper focuses on the legal compliance challenges of autonomous vehicles in a transnational context. We choose the perspective of designers and try to provide supporting legal reasoning in the design process. Based on argumentation theory, we introduce a logic to represent the basic properties of argument-based practical (normative) reasoning, combined with partial order sets of natural numbers to express priority. Finally, through case analysis of legal texts, we show how the reasoning system we provide can help designers to adapt their design solutions more flexibly in the cross-border application of autonomous vehicles and to more easily understand the legal implications of their decisions.

cs.AI↗

arXiv2025

A Joint Planning Model for Fixed and Mobile Electric Vehicle Charging Stations Considering Flexible Capacity Strategy

The widespread adoption of electric vehicles (EVs) has significantly increased demand on both transportation and power systems, posing challenges to their stable operation. To support the growing need for EV charging, both fixed charging stations (FCSs) and mobile charging stations (MCSs) have been introduced, serving as key interfaces between the power grid and traffic network. Recognizing the importance of collaborative planning across these sectors, this paper presents a two-stage joint planning model for FCSs and MCSs, utilizing an improved alternating direction method of multipliers (ADMM) algorithm. The primary goal of the proposed model is to transform the potential negative impacts of large-scale EV integration into positive outcomes, thereby enhancing social welfare through collaboration among multiple stakeholders. In the first stage, we develop a framework for evaluating FCS locations, incorporating assessments of EV hosting capacity and voltage stability. The second stage introduces a joint planning model for FCSs and MCSs, aiming to minimize the overall social costs of the EV charging system while maintaining a reliable power supply. To solve the planning problem, we employ a combination of mixed-integer linear programming, queueing theory, and sequential quadratic programming. The improved ADMM algorithm couples the siting and sizing decisions consistently by introducing coupling constraints, and supports a distributed optimization framework that coordinates the interests of EV users, MCS operators, and distribution system operators. Additionally, a flexible capacity planning strategy that accounts for the multi-period development potential of EVCS is proposed to reduce both the complexity and the investment required for FCS construction. Finally, a case study with comparative experiments demonstrates the effectiveness of the proposed models and solution methods.

eess.SY↗

arXiv2025

Holographic thermal correlators from recursions

We express holographic thermal correlators using a recurrence relation of $\{a_n\}$ at $n\to\infty$, building on recent advances in the connection formula for the Heun equation. We consider two gravitational solutions that correspond to distinct states in different subsectors of $\mathcal{N}=4$ super-Yang-Mills theory at finite temperature and density. The first is the Reissner-Nordström-AdS$_5$ black hole, which has finite entropy at zero temperature, and the second is a charged dilatonic black hole in AdS$_5$, which has zero entropy at zero temperature. In both cases, we perturb the system with a charged scalar field and express the perturbation equation in terms of the Heun equation. We find interesting moving patterns of the poles of the correlators including eigenvalue repulsions. We discuss the relation between the recurrence relation and the Virasoro conformal block as two equivalent approaches to write the connection formula for the Heun equation.

hep-th↗

arXiv2025

Supermassive Black Holes with High Accretion Rates in Active Galactic Nuclei. XII. Reverberation Mapping Results for 15 PG Quasars from a Long-Duration High-Cadence Campaign

We present the first results from long-term high-cadence spectroscopic monitoring of 15 PG quasars with relatively strong Fe II emission as a part of a broader reverberation mapping campaign performed with the Calar Alto Observatory 2.2m telescope. The $V$-band, 5100 Å continuum, and H$β$ broad emission line light curves were measured for a set of quasars for between dozens to more than a hundred epochs from May 2017 to July 2020. Accurate time lags between the variations of the H$β$ broad line fluxes and the optical continuum strength are obtained for all 15 quasars, ranging from $17.0_{-3.2}^{+2.5}$ to $95.9_{-23.9}^{+7.1}$ days in the rest frame. The virial masses of the central supermassive black holes are derived for all 15 quasars, ranging between $0.50_{-0.19}^{+0.18}$ and $19.17_{-2.73}^{+2.98}$ in units of $10^7 M_\odot$. For 11 of the objects in our sample, this is the first reverberation analysis published. Of the rest, two objects have been the subject of previous reverberation studies, but we determine time lags for these that are only half as long as found in the earlier investigations, which had only been able to sample much more sparsely. The remaining two objects have previously been monitored with high sampling rates. Our results here are consistent with the earlier findings in the sense that the time lag and the line width vary inversely consistent with virialization.

astro-ph.GA↗

arXiv2024

Approaching Code Search for Python as a Translation Retrieval Problem with Dual Encoders

Code search is vital in the maintenance and extension of software systems. Past works have used separate language models for the natural language and programming language artifacts on models with multiple encoders and different loss functions. Similarly, this work approaches code search for Python as a translation retrieval problem while the natural language queries and the programming language are treated as two types of languages. By using dual encoders, these two types of language sequences are projected onto a shared embedding space, in which the distance reflects the similarity between a given pair of query and code. However, in contrast to previous work, this approach uses a unified language model, and a dual encoder structure with a cosine similarity loss function. A unified language model helps the model take advantage of the considerable overlap of words between the artifacts, making the learning much easier. On the other hand, the dual encoders trained with cosine similarity loss helps the model learn the underlining patterns of which terms are important for predicting linked pairs of artifacts. Evaluation shows the proposed model achieves performance better than state-of-the-art code search models. In addition, this model is much less expensive in terms of time and complexity, offering a cheaper, faster, and better alternative.

cs.SE↗