arXiv subjects

Yijian Huang

Publications and source records attributed to Yijian Huang.

4 recordsLinked to original sources

arXiv2026

Dissecting Embodied Abilities in Multimodal Language Models through Skill-level Evaluation and Diagnosis

Understanding the capability bottlenecks of embodied multimodal large language models (MLLMs) is crucial for improving embodied agents. However, existing embodied benchmarks mainly focus on task-level evaluation and fail to provide actionable insights into the underlying causes of model failures. To address this limitation, we introduce BEAR, a benchmark that decomposes embodied tasks into 14 atomic skills for fine-grained skill-level evaluation. BEAR comprises 4,469 interleaved image-video-text samples spanning 14 skills across 6 categories, ranging from low-level perception to high-level planning. We evaluate 20 MLLMs on BEAR under a hierarchical skill-level diagnosis framework and uncover two key findings: (1) perceptual capabilities are major bottlenecks behind reasoning failures, and (2) current models suffer from unstable spatiotemporal modeling that remains largely unexposed in prior benchmarks. Motivated by these findings, we further propose BEAR-Agent, a multimodal conversational agent that augments MLLMs with visual and spatial reasoning tools. BEAR-Agent substantially improves performance across embodied skills, achieving a relative improvement of 17.5% on GPT-5 over the base model on BEAR, while also outperforming strong baselines in both simulation and real-world robotic experiments. Project page: https://bear-official66.github.io/

cs.CV↗

arXiv2026

Tightening Control in Neyman--Pearson Linear Classification

Neyman--Pearson classification prioritizes one class by constraining its accuracy above a prespecified level, and then takes the accuracy of the other class as the utility objective. This paradigm is well suited for disease screening and diagnosis, among other applications. Statistical learning under this framework is complicated since classifier performance determines its acceptability. Furthermore, no learned classifier that is consistent for the oracle classifier can guarantee satisfaction of the control constraint in finite samples. Classical learning theory targets a control-relaxed empirical utility maximization (EUM) classifier. However, even the EUM classifier fails to achieve the desired control level on average. We conjecture that this under-control phenomenon is a manifestation of the over-optimism bias well known in standard statistical learning, and develop asymptotic theory to confirm it. Motivated by this insight, we propose refined learning procedures under two accuracy control strategies for the prioritized class: one controlling accuracy in expectation and the other with high probability. We further develop training-data-based methods to predict and infer class-specific accuracies of the resulting classifiers. Simulation studies demonstrate favorable finite-sample performance, and we illustrate the proposed methods with an application to cancer detection.

math.ST↗

arXiv2026

Cross-Audit Projection for Model Risk Prediction

For training-data-based model risk prediction, $K$-fold cross-validation~(CV) is widely used to mitigate the well-known over-optimism of the empirical risk and is often regarded as reliable. However, for binary classification via empirical risk minimization, our numerical studies reveal a surprising phenomenon: $K$-fold CV may perform poorly in estimating class-specific risks, even worse than the empirical estimator. We perform a higher-order asymptotic analysis showing that $K$-fold CV may converge at a slower rate, whereas the empirical estimator exhibits a second-order asymptotic bias that explains its over-optimism. These findings motivate a novel two-step procedure for model risk prediction, termed cross-audit projection (CAP). The cross-audit step adopts the same resampling scheme as $K$-fold CV to estimate over-optimism in subsamples, while the asymptotic-theory-informed projection step adjusts for the reduced sample size in bias correction of the empirical risk. The resulting CAP estimator is first-order asymptotically equivalent to the empirical risk while achieving second-order asymptotic unbiasedness. An accompanying inference procedure is also developed. Simulation studies support theoretical advantages of CAP and demonstrate favorable finite-sample performance. An application to breast cancer detection further illustrates the proposed method.

stat.ME↗

arXiv2010

Quantile calculus and censored regression

Quantile regression has been advocated in survival analysis to assess evolving covariate effects. However, challenges arise when the censoring time is not always observed and may be covariate-dependent, particularly in the presence of continuously-distributed covariates. In spite of several recent advances, existing methods either involve algorithmic complications or impose a probability grid. The former leads to difficulties in the implementation and asymptotics, whereas the latter introduces undesirable grid dependence. To resolve these issues, we develop fundamental and general quantile calculus on cumulative probability scale in this article, upon recognizing that probability and time scales do not always have a one-to-one mapping given a survival distribution. These results give rise to a novel estimation procedure for censored quantile regression, based on estimating integral equations. A numerically reliable and efficient Progressive Localized Minimization (PLMIN) algorithm is proposed for the computation. This procedure reduces exactly to the Kaplan--Meier method in the $k$-sample problem, and to standard uncensored quantile regression in the absence of censoring. Under regularity conditions, the proposed quantile coefficient estimator is uniformly consistent and converges weakly to a Gaussian process. Simulations show good statistical and algorithmic performance. The proposal is illustrated in the application to a clinical study.

math.ST↗