arXiv subjects

Yujie Zhao

Publications and source records attributed to Yujie Zhao.

At least 37 records · Page 2Linked to original sources

arXiv2025

TwinAligner: Visual-Dynamic Alignment Empowers Physics-aware Real2Sim2Real for Robotic Manipulation

The robotics field is evolving towards data-driven, end-to-end learning, inspired by multimodal large models. However, reliance on expensive real-world data limits progress. Simulators offer cost-effective alternatives, but the gap between simulation and reality challenges effective policy transfer. This paper introduces TwinAligner, a novel Real2Sim2Real system that addresses both visual and dynamic gaps. The visual alignment module achieves pixel-level alignment through SDF reconstruction and editable 3DGS rendering, while the dynamic alignment module ensures dynamic consistency by identifying rigid physics from robot-object interaction. TwinAligner improves robot learning by providing scalable data collection and establishing a trustworthy iterative cycle, accelerating algorithm development. Quantitative evaluations highlight TwinAligner's strong capabilities in visual and dynamic real-to-sim alignment. This system enables policies trained in simulation to achieve strong zero-shot generalization to the real world. The high consistency between real-world and simulated policy performance underscores TwinAligner's potential to advance scalable robot learning. Code and data will be released on https://twin-aligner.github.io

cs.RO↗

arXiv2025

PRO-V-R1: Reasoning Enhanced Programming Agent for RTL Verification

Register-Transfer Level (RTL) verification is a primary bottleneck, consuming 60-70% of development time. While Large Language Models (LLMs) show promise for RTL automation, their performance and research focus have overwhelmingly centered on RTL generation rather than verification. Current methods for RTL verification rely on large scale proprietary models (e.g., GPT-4o) to generate Python-based functional references, incurring a high cost and raising data-privacy risks. To date, an end-to-end open-source solution for autonomous verification remains absent. We introduce PRO-V-R1, the first trainable open-source agentic framework for autonomous RTL verification. Our contributions are threefold: (1) we design PRO-V sys, a modular agentic system that couples LLM-based reasoning with programmatic tool use for RTL verification; (2) we establish a data construction pipeline that leverages existing RTL datasets to build simulation-validated, expert-level trajectories tailored for supervised fine-tuning (SFT) RTL verification agents; and (3) we implement an efficient reinforcement learning (RL) algorithm that uses verification-specific rewards derived from program-tool feedback to optimize the end-to-end verification workflow. Our empirical evaluation demonstrates PRO-V-R1 achieves a 57.7% functional correctness rate and 34.0% in robust fault detection, significantly outperforming the base model's 25.7% and 21.8% (respectively) from the state-of-the-art (SOTA) automatic verification system. This configuration also outperforms large-scale proprietary LLMs in functional correctness and shows comparable robustness for fault detection.

cs.AI↗

arXiv2025

Benchmarking Scientific Understanding and Reasoning for Video Generation using VideoScience-Bench

The next frontier for video generation lies in developing models capable of zero-shot reasoning, where understanding real-world scientific laws is crucial for accurate physical outcome modeling under diverse conditions. However, existing video benchmarks are physical commonsense-based, offering limited insight into video models' scientific reasoning capability. We introduce VideoScience-Bench, a benchmark designed to evaluate undergraduate-level scientific understanding in video models. Each prompt encodes a composite scientific scenario that requires understanding and reasoning across multiple scientific concepts to generate the correct phenomenon. The benchmark comprises 200 carefully curated prompts spanning 14 topics and 103 concepts in physics and chemistry. We conduct expert-annotated evaluations across seven state-of-the-art video models in T2V and I2V settings along five dimensions: Prompt Consistency, Phenomenon Congruency, Correct Dynamism, Immutability, and Spatio-Temporal Continuity. Using a VLM-as-a-Judge to assess video generations, we observe strong correlation with human assessments. To the best of our knowledge, VideoScience-Bench is the first benchmark to evaluate video models not only as generators but also as reasoners, requiring their generations to demonstrate scientific understanding consistent with expected physical and chemical phenomena. Our data and evaluation code are available at: \href{https://github.com/hao-ai-lab/VideoScience}{github.com/hao-ai-lab/VideoScience}.

cs.CV↗

arXiv2025

OrcaLoca: An LLM Agent Framework for Software Issue Localization

Recent developments in Large Language Model (LLM) agents are revolutionizing Autonomous Software Engineering (ASE), enabling automated coding, problem fixes, and feature improvements. However, localization -- precisely identifying software problems by navigating to relevant code sections -- remains a significant challenge. Current approaches often yield suboptimal results due to a lack of effective integration between LLM agents and precise code search mechanisms. This paper introduces OrcaLoca, an LLM agent framework that improves accuracy for software issue localization by integrating priority-based scheduling for LLM-guided action, action decomposition with relevance scoring, and distance-aware context pruning. Experimental results demonstrate that OrcaLoca becomes the new open-source state-of-the-art (SOTA) in function match rate (65.33%) on SWE-bench Lite. It also improves the final resolved rate of an open-source framework by 6.33 percentage points through its patch generation integration.

cs.SE↗

arXiv2025

LangSplatV2: High-dimensional 3D Language Gaussian Splatting with 450+ FPS

In this paper, we introduce LangSplatV2, which achieves high-dimensional feature splatting at 476.2 FPS and 3D open-vocabulary text querying at 384.6 FPS for high-resolution images, providing a 42 $\times$ speedup and a 47 $\times$ boost over LangSplat respectively, along with improved query accuracy. LangSplat employs Gaussian Splatting to embed 2D CLIP language features into 3D, significantly enhancing speed and learning a precise 3D language field with SAM semantics. Such advancements in 3D language fields are crucial for applications that require language interaction within complex scenes. However, LangSplat does not yet achieve real-time inference performance (8.2 FPS), even with advanced A100 GPUs, severely limiting its broader application. In this paper, we first conduct a detailed time analysis of LangSplat, identifying the heavyweight decoder as the primary speed bottleneck. Our solution, LangSplatV2 assumes that each Gaussian acts as a sparse code within a global dictionary, leading to the learning of a 3D sparse coefficient field that entirely eliminates the need for a heavyweight decoder. By leveraging this sparsity, we further propose an efficient sparse coefficient splatting method with CUDA optimization, rendering high-dimensional feature maps at high quality while incurring only the time cost of splatting an ultra-low-dimensional feature. Our experimental results demonstrate that LangSplatV2 not only achieves better or competitive query accuracy but is also significantly faster. Codes and demos are available at our project page: https://langsplat-v2.github.io.

cs.CV↗

arXiv2025

A Review of Statistical Methods for Handling Nonignorable Missing Data using Instrument Approach

Nonignorable missing data, where the probability of missingness depends on unobserved values, presents a significant challenge in statistical analysis. Traditional methods often rely on strong parametric assumptions that are difficult to verify and may lead to biased estimates if misspecified. Recent advances have introduced the concept of a nonresponse instrument or shadow variable as a powerful tool to enhance model identifiability without requiring full parametric specification. This paper provides a comprehensive review of statistical methods that leverage instrumental variables to address nonignorable missingness, focusing on two predominant semiparametric frameworks: one with a parametric data model and a nonparametric propensity model, and the other with a parametric propensity model and a nonparametric data model. We discuss key developments, methodological insights, and remaining challenges in this rapidly evolving field.

stat.ME↗

arXiv2025

Group Sequential Design for Non-Proportional Hazards: Logrank, Weighted Logrank, and MaxCombo Methods

Non-proportional hazards (NPH) are often observed in clinical trials with time-to-event endpoints. A common example is a long-term clinical trial with a delayed treatment effect in immunotherapy for cancer. When designing clinical trials with time-to-event endpoints, it is crucial to consider NPH scenarios to gain a complete understanding of design operating characteristics. In this paper, we focus on group sequential design for three NPH methods: the average hazard ratio, the weighted logrank test, and the MaxCombo combination test. For each of these approaches, we provide analytic forms of design characteristics that facilitate sample size calculation and bound derivation for group sequential designs. Examples are provided to illustrate the proposed methods. To facilitate statisticians in designing and comparing group sequential designs under NPH, we have implemented the group sequential design methodology in the gsDesign2 R package at https://cran.r-project.org/web/packages/gsDesign2/.

stat.ME↗

arXiv2025

Identification of Partial-Differential-Equations-Based Models from Noisy Data via Splines

We propose a two-stage method called \textit{Spline Assisted Partial Differential Equation based Model Identification (SAPDEMI)} to identify partial differential equation (PDE)-based models from noisy data. In the first stage, we employ the cubic splines to estimate unobservable derivatives. The underlying PDE is based on a subset of these derivatives. This stage is computationally efficient: its computational complexity is a product of a constant with the sample size; this is the lowest possible order of computational complexity. In the second stage, we apply the Least Absolute Shrinkage and Selection Operator (Lasso) to identify the underlying PDE-based model. Statistical properties are developed, including the model identification accuracy. We validate our theory through various numerical examples and a real data case study. The case study is based on a National Aeronautics and Space Administration (NASA) data set.

stat.ME↗

arXiv2025

Pose-Robust Calibration Strategy for Point-of-Gaze Estimation on Mobile Phones

Although appearance-based point-of-gaze (PoG) estimation has improved, the estimators still struggle to generalize across individuals due to personal differences. Therefore, person-specific calibration is required for accurate PoG estimation. However, calibrated PoG estimators are often sensitive to head pose variations. To address this, we investigate the key factors influencing calibrated estimators and explore pose-robust calibration strategies. Specifically, we first construct a benchmark, MobilePoG, which includes facial images from 32 individuals focusing on designated points under either fixed or continuously changing head poses. Using this benchmark, we systematically analyze how the diversity of calibration points and head poses influences estimation accuracy. Our experiments show that introducing a wider range of head poses during calibration improves the estimator's ability to handle pose variation. Building on this insight, we propose a dynamic calibration strategy in which users fixate on calibration points while moving their phones. This strategy naturally introduces head pose variation during a user-friendly and efficient calibration process, ultimately producing a better calibrated PoG estimator that is less sensitive to head pose variations than those using conventional calibration strategies. Codes and datasets are available at our project page.

cs.CV↗

arXiv2025

Simulation of pulsed dynamic nuclear polarization in the steady state

In pulsed dynamic nuclear polarization (DNP), enhancement of the polarization of bulk nuclei requires the repeated application of a microwave pulse sequence. So far, analysis of a one-time transfer of electron spin polarization to a dipolar-coupled nuclear spin has guided the design of DNP pulse sequences. This has obvious shortcomings, such as an inability to predict the optimal repetition time. In an actual pulsed DNP experiment, a balance is reached between the polarization arriving from the unpaired electrons and nuclear relaxation. In this article, we explore three algorithms to compute this stroboscopic steady state: (1) explicit time evolution by propagator squaring, (2) generation of an effective propagator using the matrix logarithm, and (3) direct calculation of the steady state with the Newton-Raphson method. Algorithm (2) is numerically unstable for this purpose. Algorithm (1) and (3) are both stable; algorithm (3) is the most efficient. We compare the steady-state simulations to existing experimental results at 0.34 T and 1.2 T and to the first experimental observation of X-inverse-X (XiX) DNP at 3.4 T: agreement is good, and improves further when electron-proton distance and electron Rabi frequency distributions are accounted for. We demonstrate that the trajectory of the spin system during one-time application of a microwave pulse sequence differs from the steady orbit. This has implications for DNP pulse sequence design.

physics.chem-ph↗

arXiv2025

RA-PbRL: Provably Efficient Risk-Aware Preference-Based Reinforcement Learning

Reinforcement Learning from Human Feedback (RLHF) has recently surged in popularity, particularly for aligning large language models and other AI systems with human intentions. At its core, RLHF can be viewed as a specialized instance of Preference-based Reinforcement Learning (PbRL), where the preferences specifically originate from human judgments rather than arbitrary evaluators. Despite this connection, most existing approaches in both RLHF and PbRL primarily focus on optimizing a mean reward objective, neglecting scenarios that necessitate risk-awareness, such as AI safety, healthcare, and autonomous driving. These scenarios often operate under a one-episode-reward setting, which makes conventional risk-sensitive objectives inapplicable. To address this, we explore and prove the applicability of two risk-aware objectives to PbRL : nested and static quantile risk objectives. We also introduce Risk-AwarePbRL (RA-PbRL), an algorithm designed to optimize both nested and static objectives. Additionally, we provide a theoretical analysis of the regret upper bounds, demonstrating that they are sublinear with respect to the number of episodes, and present empirical results to support our findings. Our code is available in https://github.com/aguilarjose11/PbRLNeurips.

cs.LG↗

arXiv2024

MAGE: A Multi-Agent Engine for Automated RTL Code Generation

The automatic generation of RTL code (e.g., Verilog) through natural language instructions has emerged as a promising direction with the advancement of large language models (LLMs). However, producing RTL code that is both syntactically and functionally correct remains a significant challenge. Existing single-LLM-agent approaches face substantial limitations because they must navigate between various programming languages and handle intricate generation, verification, and modification tasks. To address these challenges, this paper introduces MAGE, the first open-source multi-agent AI system designed for robust and accurate Verilog RTL code generation. We propose a novel high-temperature RTL candidate sampling and debugging system that effectively explores the space of code candidates and significantly improves the quality of the candidates. Furthermore, we design a novel Verilog-state checkpoint checking mechanism that enables early detection of functional errors and delivers precise feedback for targeted fixes, significantly enhancing the functional correctness of the generated RTL code. MAGE achieves a 95.7% rate of syntactic and functional correctness code generation on VerilogEval-Human 2 benchmark, surpassing the state-of-the-art Claude-3.5-sonnet by 23.3 %, demonstrating a robust and reliable approach for AI-driven RTL design workflows.

cs.AR↗

arXiv2024

Charge management system based on disturbance observer sliding mode control for space inertial sensors

Precision space inertial sensors are imperative for Earth geodesy missions, gravitational wave observations, and fundamental physics experiments in space. In these missions, free-falling test masses(TMs) are susceptible to parasitic electrostatic forces and torques, with significant contributions from the interaction between stray electric fields and TM charge. These effects can make up a sizable fraction of the noise budget. Thus, a charge management system(CMS) is essential in high-precise space-based missions. However, the operating environment for space charge control is full of uncertainties and disturbances. TM charge tracking precision is negatively affected by many physical parameters such as external charging rate, quantum yield, UV light power, etc. Those parameters are rarely measured and supposed to vary because of changes in solar activity, temperature, aging of electronic components and so on. The unpredictability and variability of these parameters affects the CMS performance in long-term space missions and must be evaluated or eliminated. This paper presents a simple physics-based model of the discharging process with high charging/discharging rate based on the geometry of inertial sensors. After that, a disturbance observer sliding mode control (DOSMC) is proposed for the CMS with parametric uncertainties and unknown disturbance to maintain the TM charge below a certain level and improve its robustness. The simulation results show that the DOSMC is able to force the system trajectory coincides with the sliding line, which depends neither on the parameters or disturbances. In this way, the DOSMC can effectively ignore the parameter perturbation and external disturbances. The control precision can reach 0.1 mV, which is superior to that of a classic proportional-integral-derivative controller and conventional sliding mode control.

physics.ins-det↗

arXiv2024

Actuation system of the inertial sensor for high-precision space missions using torsion pendulum

Precision space inertial sensors are imperative to Earth geodesy missions, gravitational wave observations and several fundamental physics experiments in space. In these missions, the residual acceleration noise of the test mass(TM) caused by the forces from inertial sensor components and environment is supposed to be kept below a certain level. As a number of forces contributing to residual acceleration are related to actuation system, developing a precise actuation system to exclude any erroneous force and obtain an ultra sensitive value for TM acceleration noise is necessary and essential. However, it is difficult to test the actuation system on ground. In this paper, a torsion pendulum is established to test the influence of actuation system on TM torque noise and a closed-loop control system combined torsion pendulum and parts of actuation modules is designed to assess the performance of actuation control algorithm. The experimental results show that the parameters in an actuation system will introduce additional torque noise and the maximum noise can reach as much as 10^{-13}Nm /Hz^{1/2} at 1 mHz. The stable tracking error for the closed-loop system is about 10^{-7}, indicating that the combination system achieves good tracking performance and robustness for TM rotation control in different conditions of inertial sensors.

physics.space-ph↗

arXiv2024

3D-Carbon: An Analytical Carbon Modeling Tool for 3D and 2.5D Integrated Circuits

Environmental sustainability is crucial for Integrated Circuits (ICs) across their lifecycle, particularly in manufacturing and use. Meanwhile, ICs using 3D/2.5D integration technologies have emerged as promising solutions to meet the growing demands for computational power. However, there is a distinct lack of carbon modeling tools for 3D/2.5D ICs. Addressing this, we propose 3D-Carbon, an analytical carbon modeling tool designed to quantify the carbon emissions of 3D/2.5D ICs throughout their life cycle. 3D-Carbon factors in both potential savings and overheads from advanced integration technologies, considering practical deployment constraints like bandwidth. We validate 3D-Carbon's accuracy against established baselines and illustrate its utility through case studies in autonomous vehicles. We believe that 3D-Carbon lays the initial foundation for future innovations in developing environmentally sustainable 3D/2.5D ICs. Our open-source code is available at https://github.com/UMN-ZhaoLab/3D-Carbon.

cs.AR↗

arXiv2023

Adjusted inference for multiple testing procedure in group sequential designs

Adjustment of statistical significance levels for repeated analysis in group sequential trials has been understood for some time. Similarly, methods for adjustment accounting for testing multiple hypotheses are common. There is limited research on simultaneously adjusting for both multiple hypothesis testing and multiple analyses of one or more hypotheses. We address this gap by proposing adjusted-sequential p-values that reject an elementary hypothesis when its adjusted-sequential p-values are less than or equal to the family-wise Type I error rate (FWER) in a group sequential design. We also propose sequential p-values for intersection hypotheses as a tool to compute adjusted sequential p-values for elementary hypotheses. We demonstrate the application using weighted Bonferroni tests and weighted parametric tests, comparing adjusted sequential p-values to a desired FWER for inference on each elementary hypothesis tested.

stat.ME↗

arXiv2023

Number of Repetitions in Re-randomization Tests

In covariate-adaptive or response-adaptive randomization, the treatment assignment and outcome can be correlated. Under this situation, re-randomization tests are a straightforward and attractive method to provide valid statistical inference. In this paper, we investigate the number of repetitions in the re-randomization tests. This is motivated by the group sequential design in clinical trials, where the nominal significance bound can be very small at an interim analysis. Accordingly, re-randomization tests lead to a very large number of required repetitions, which may be computationally intractable. To reduce the number of repetitions, we propose an adaptive procedure and compare it with multiple approaches under pre-defined criteria. Monte Carlo simulations are conducted to show the performance of different approaches in a limited sample size. We also suggest strategies to reduce total computation time and provide practical guidance in preparing, executing and reporting before and after data are unblinded at an interim analysis, so one can complete the computation within a reasonable time frame.

stat.ME↗

arXiv2023

Accelerate the Warm-up Stage in the Lasso Computation via a Homotopic Approach

In optimization, it is known that when the objective functions are strictly convex and well-conditioned, gradient-based approaches can be extremely effective, e.g., achieving the exponential rate of convergence. On the other hand, the existing Lasso-type estimator in general cannot achieve the optimal rate due to the undesirable behavior of the absolute function at the origin. A homotopic method is to use a sequence of surrogate functions to approximate the $\ell_1$ penalty that is used in the Lasso-type of estimators. The surrogate functions will converge to the $\ell_1$ penalty in the Lasso estimator. At the same time, each surrogate function is strictly convex, which enables a provable faster numerical rate of convergence. In this paper, we demonstrate that by meticulously defining the surrogate functions, one can prove a faster numerical convergence rate than any existing methods in computing for the Lasso-type of estimators. Namely, the state-of-the-art algorithms can only guarantee $O(1/ε)$ or $O(1/\sqrtε)$ convergence rates, while we can prove an $O([\log(1/ε)]^2)$ for the newly proposed algorithm. Our numerical simulations show that the new algorithm also performs better empirically.

stat.ML↗