Searcharxiv⌕ Search

arXiv subjects

Tian Yu

Publications and source records attributed to Tian Yu.

At least 19 recordsLinked to original sources

Drive the Thoughts: Runtime Monitoring of VLA Reasoning-Trajectory Consistency

While machine learning models help autonomous vehicles (AVs) navigate complex environments, their black-box nature complicates validation and verification (V&V). Vision-language-action (VLA) models offer a unique opportunity: alongside trajectories, they generate an explicit chain-of-thought (CoT) explaining their decisions. We posit that this CoT can serve as a candidate specification to cross-check trajectories and expose unsafe behavior. To assess this, we curate DriveAlignBench, a dataset of 150 CoT-trajectory pairs from NVIDIA's Alpamayo 1.5 VLA, manually annotated for CoT reliability, CoT-trajectory consistency, and trajectory safety. We find that 33.3% of CoTs are unreliable, and among reliable CoTs, 26% have inconsistent trajectories. To detect these cases, we propose integrating a CoT-trajectory consistency check into a runtime monitor. This check is nontrivial because CoTs express open-vocabulary, scene-relative maneuvers, while trajectories are low-level ego-motion sequences whose semantics depend on road geometry and motion context. To bridge this gap, we develop a family of automated consistency monitors. The strongest, the lane-aware monitor, adds lane-relative evidence to the input of the judge model scoring consistency. This raises F1 across all four evaluated judges from 0.56-0.62 with raw waypoints to 0.65-0.82, showing that evidence representation improves detection across judges. Moreover, for reliable CoTs, inconsistency is associated with unsafe behavior: 27% of human-labeled inconsistent trajectories are unsafe compared to 3% of consistent ones, and the lane-aware monitor flags most of these unsafe cases. We release DriveAlignBench, the monitor implementations, and per-run results at https://github.com/less-lab-uva/drive-the-thoughts.

cs.SE↗

Time-Consistent Evaluation of Hierarchical Code Graphs for SWE Agents

Evaluating repository-aware software engineering (SWE) agents requires strict separation between pre-existing repository knowledge and future code evolution, yet standard benchmarks frequently suffer from temporal contamination and prompt-level solution leakage. Building on the time-consistent benchmark protocol of Sun (2026), we enforce strict zero-leakage isolation across DragonFly in C++ and React in JavaScript/TypeScript, locking repository snapshots and knowledge graphs at historical cutoff commits $T_0$ and evaluating 100 post-$T_0$ pull requests across four prompt granularities: Minimal, Concise, Contextual, and Guided. Under this zero-leakage protocol, we conduct a matched evaluation of three context-augmentation regimes: (1) an unaugmented Baseline, (2) a flat syntax-level Tree-sitter AST graph, and (3) ServiceExplorer, a deterministic two-level hierarchical dependency graph capturing Level-1 system/package containers and Level-2 architectural modules. Across 7,200 paired evaluations spanning six foundation models across the MAI-Code Flash, Claude, and GPT-Sol families, ServiceExplorer consistently outperforms both Baseline and Tree-sitter AST across every model, repository, and prompt granularity ($+15.6\%$ to $+86.2\%$ relative $F_1$ lift, $p < 10^{-4}$), while tripling exact full-PR localization on DragonFly. Crucially, equipping an efficiency-tier or prior-generation model with ServiceExplorer consistently surpasses the unaugmented Baseline $F_1$ of the next-generation model across all evaluated families.

cs.SE↗

BioFact-MoE: Biologically Factorized Mixture of Experts for Vision-Language Prognostic Modeling in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is biologically heterogeneous, shaped by the interplay between hepatic functional reserve and tumor-related oncologic factors; thus, similar survival outcomes may reflect fundamentally different underlying biological processes. Prognostic modeling in HCC is informed by rich multimodal information from multiparametric MRI and radiology reports from routine clinical practice. Existing prognostic vision-language models (VLMs) learn a single entangled latent representation that blends hepatic and tumor-related factors, limiting both accuracy and biological interpretability. We present BioFact-MoE, a biologically factorized Mixture of Experts (MoE) framework that explicitly decomposes liver and tumor factors via biologically supervised experts within a residual MoE survival architecture. On a HCC cohort of N=588 patients (pretrained on 4,582 3D MRI image-report pairs), BioFact-MoE consistently improves survival prediction over all baselines across time horizons, achieving 12-, 18-, and 24-month AUCs of 75.33%, 75.85%, and 73.96%. Beyond scalar risk prediction, gated expert weights enable phenotype-aware risk stratification. Pathway-informed gating uncovers clinically meaningful treatment-associated survival heterogeneity. In held-out validation, hepatic and tumor embeddings show selective associations with liver function and tumor burden markers, respectively (p<0.05), without supervision. The code is available at https://github.com/jy-639/BioFact-MoE.

cs.CV↗

ATime-Consistent Benchmark for Repository-Level Software Engineering Evaluation

Evaluation of repository-aware software engineering systems is often confounded by synthetic task design, prompt leakage, and temporal contamination between repository knowledge and future code changes. We present a time-consistent benchmark methodology that snapshots a repository at time T0, constructs repository-derived code knowledge using only artifacts available before T0, and evaluates on engineering tasks derived from pull requests merged in the future interval (T0, T1]. Each historical pull request is transformed into a natural-language task through an LLM-assisted prompt-generation pipeline, and the benchmark is formalized as a matched A/B comparison in which the same software engineering agent is evaluated with and without repository-derived code knowledge while all other variables are held constant. We also report a baseline characterization study on two open-source repositories, DragonFly and React, using three Claude-family models and four prompt granularities. Across both repositories, file-level F1 increases monotonically from minimal to guided prompts, reaching 0.8081 on DragonFly and 0.8078 on React for the strongest tested model. These results show that prompt construction is a first-order benchmark variable. More broadly, the benchmark highlights that temporal consistency and prompt control are core validity requirements for repository-aware software engineering evaluation.

cs.SE↗

DECADE: A Temporally-Consistent Unsupervised Diffusion Model for Enhanced Rb-82 Dynamic Cardiac PET Image Denoising

Rb-82 dynamic cardiac PET imaging is widely used for the clinical diagnosis of coronary artery disease (CAD), but its short half-life results in high noise levels that degrade dynamic frame quality and parametric imaging. The lack of paired clean-noisy training data, rapid tracer kinetics, and frame-dependent noise variations further limit the effectiveness of existing deep learning denoising methods. We propose DECADE (A Temporally-Consistent Unsupervised Diffusion model for Enhanced Rb-82 CArdiac PET DEnoising), an unsupervised diffusion framework that generalizes across early- to late-phase dynamic frames. DECADE incorporates temporal consistency during both training and iterative sampling, using noisy frames as guidance to preserve quantitative accuracy. The method was trained and evaluated on datasets acquired from Siemens Vision 450 and Siemens Biograph Vision Quadra scanners. On the Vision 450 dataset, DECADE consistently produced high-quality dynamic and parametric images with reduced noise while preserving myocardial blood flow (MBF) and myocardial flow reserve (MFR). On the Quadra dataset, using 15%-count images as input and full-count images as reference, DECADE outperformed UNet-based and other diffusion models in image quality and K1/MBF quantification. The proposed framework enables effective unsupervised denoising of Rb-82 dynamic cardiac PET without paired training data, supporting clearer visualization while maintaining quantitative integrity.

cs.CV↗

UniSync: Towards Generalizable and High-Fidelity Lip Synchronization for Challenging Scenarios

Lip synchronization aims to generate realistic talking videos that match given audio, which is essential for high-quality video dubbing. However, current methods have fundamental drawbacks: mask-based approaches suffer from local color discrepancies, while mask-free methods struggle with global background texture misalignment. Furthermore, most methods struggle with diverse real-world scenarios such as stylized avatars, face occlusion, and extreme lighting conditions. In this paper, we propose UniSync, a unified framework designed for achieving high-fidelity lip synchronization in diverse scenarios. Specifically, UniSync uses a mask-free pose-anchored training strategy to keep head motion and eliminate synthesis color artifacts, while employing mask-based blending consistent inference to ensure structural precision and smooth blending. Notably, fine-tuning on compact but diverse videos empowers our model with exceptional domain adaptability, handling complex corner cases effectively. We also introduce the RealWorld-LipSync benchmark to evaluate models under real-world demands, which covers diverse application scenarios including both human faces and stylized avatars. Extensive experiments demonstrate that UniSync significantly outperforms state-of-the-art methods, advancing the field towards truly generalizable and production-ready lip synchronization.

cs.CV↗

Multi-Agent Synergy-Driven Iterative Visual Narrative Synthesis

Automated generation of high-quality media presentations is challenging, requiring robust content extraction, narrative planning, visual design, and overall quality optimization. Existing methods often produce presentations with logical inconsistencies and suboptimal layouts, thereby struggling to meet professional standards. To address these challenges, we introduce RCPS (Reflective Coherent Presentation Synthesis), a novel framework integrating three key components: (1) Deep Structured Narrative Planning; (2) Adaptive Layout Generation; (3) an Iterative Optimization Loop. Additionally, we propose PREVAL, a preference-based evaluation framework employing rationale-enhanced multi-dimensional models to assess presentation quality across Content, Coherence, and Design. Experimental results demonstrate that RCPS significantly outperforms baseline methods across all quality dimensions, producing presentations that closely approximate human expert standards. PREVAL shows strong correlation with human judgments, validating it as a reliable automated tool for assessing presentation quality.

cs.CL↗

Geometry and Mechanics of Non-Euclidean Curved-Crease Origami

Recently there have been extensive theoretical, numerical and experimental works on curved-fold origami. However, we notice that a unified and complete geometric framework for describing the geometry and mechanics of curved-fold origami, especially those with nontrivial Gaussian curvature at the crease (non-Euclidean crease), is still absent. Herein we provide a unified geometric framework that describes the shape of a generic curved-fold origami composed of two general strips. The explicit description indicates that four configurations emerge, determined by its spatial crease and configuration branch. Within this geometric framework, we derive the equilibrium equations and study the mechanical response of the curved-crease origami, focusing on Euler's buckling behavior. Both linear stability analysis and finite element simulation indicate that the overlaid configuration exhibits a lower buckling threshold. To further capture the large deformation behavior efficiently, we develop a bistrip model based on the anisotropic Kirchhoff rod theory, which predicts the main features successfully. This work bridges the geometry and mechanics of curved-crease origami, offering insights for applications in robotics, actuators, and deployable space structures.

cond-mat.soft↗

Hybrid Robot Learning for Automatic Robot Motion Planning in Manufacturing

Industrial robots are widely used in diverse manufacturing environments. Nonetheless, how to enable robots to automatically plan trajectories for changing tasks presents a considerable challenge. Further complexities arise when robots operate within work cells alongside machines, humans, or other robots. This paper introduces a multi-level hybrid robot motion planning method combining a task space Reinforcement Learning-based Learning from Demonstration (RL-LfD) agent and a joint-space based Deep Reinforcement Learning (DRL) based agent. A higher level agent learns to switch between the two agents to enable feasible and smooth motion. The feasibility is computed by incorporating reachability, joint limits, manipulability, and collision risks of the robot in the given environment. Therefore, the derived hybrid motion planning policy generates a feasible trajectory that adheres to task constraints. The effectiveness of the method is validated through sim ulated robotic scenarios and in a real-world setup.

cs.RO↗

Auto-RAG: Autonomous Retrieval-Augmented Generation for Large Language Models

Iterative retrieval refers to the process in which the model continuously queries the retriever during generation to enhance the relevance of the retrieved knowledge, thereby improving the performance of Retrieval-Augmented Generation (RAG). Existing work typically employs few-shot prompting or manually constructed rules to implement iterative retrieval. This introduces additional inference overhead and overlooks the remarkable reasoning capabilities of Large Language Models (LLMs). In this paper, we introduce Auto-RAG, an autonomous iterative retrieval model centered on the LLM's powerful decision-making capabilities. Auto-RAG engages in multi-turn dialogues with the retriever, systematically planning retrievals and refining queries to acquire valuable knowledge. This process continues until sufficient external information is gathered, at which point the results are presented to the user. To this end, we develop a method for autonomously synthesizing reasoning-based decision-making instructions in iterative retrieval and fine-tuned the latest open-source LLMs. The experimental results indicate that Auto-RAG is capable of autonomous iterative interaction with the retriever, effectively leveraging the remarkable reasoning and decision-making abilities of LLMs, which lead to outstanding performance across six benchmarks. Further analysis reveals that Auto-RAG can autonomously adjust the number of iterations based on the difficulty of the questions and the utility of the retrieved knowledge, without requiring any human intervention. Moreover, Auto-RAG expresses the iterative retrieval process in natural language, enhancing interpretability while providing users with a more intuitive experience\footnote{Code is available at \url{https://github.com/ictnlp/Auto-RAG}.

cs.CL↗

Truth-Aware Context Selection: Mitigating Hallucinations of Large Language Models Being Misled by Untruthful Contexts

Although Large Language Models (LLMs) have demonstrated impressive text generation capabilities, they are easily misled by untruthful contexts provided by users or knowledge augmentation tools, leading to hallucinations. To alleviate LLMs from being misled by untruthful context and take advantage of knowledge augmentation, we propose Truth-Aware Context Selection (TACS), a lightweight method to adaptively recognize and mask untruthful context from the inputs. TACS begins by performing truth detection on the input context, leveraging the parameterized knowledge within the LLM. Subsequently, it constructs a corresponding attention mask based on the truthfulness of each position, selecting the truthful context and discarding the untruthful context. Additionally, we introduce a new evaluation metric, Disturbance Adaption Rate, to further study the LLMs' ability to accept truthful information and resist untruthful information. Experimental results indicate that TACS can effectively filter untruthful context and significantly improve the overall quality of LLMs' responses when presented with misleading information.

cs.CL↗

Double-eigenvalue bifurcation and multistability in serpentine strips with tunable buckling behaviors

Serpentine structures, composed of straight and circular strips, have garnered significant attention as potential designs for flexible electronics due to their remarkable stretchability. When subjected to stretching, these serpentine strips buckle out of plane, and previous studies have identified two distinct buckling modes whose order of appearance may interchange in serpentine structures with a single cell. In this study, we employ anisotropic rod theory to model serpentine strips as a multi-segment boundary value problem (BVP), with continuity conditions enforced at the interface between the straight and curved strips. We solve the BVP using methods of continuation, and our results reveal that: 1) the exchange of the two buckling modes in a single-cell serpentine strip is induced by a double-eigenvalue and associated secondary bifurcations, which also alter the stability of the two buckling modes; 2) a variety of stable states with reversible symmetry can be manually obtained in tabletop models and are found to be disconnected from the planar branch in numerical continuation. Furthermore, we demonstrate that modulating the strip thickness across different cells leads to the initiation of buckling in the thinnest section, thereby allowing for the tuning of buckling modes in serpentine strips. In structures with two cells, the sequence of the two buckling modes can also be controlled by designing serpentine strips with nonuniform height. This work could enhance the mechanical design of serpentine-interconnect-based flexible structures and could have applications in multistable actuators and mechanical memory devices.

cond-mat.soft↗

TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space

Large Language Models (LLMs) sometimes suffer from producing hallucinations, especially LLMs may generate untruthful responses despite knowing the correct knowledge. Activating the truthfulness within LLM is the key to fully unlocking LLM's knowledge potential. In this paper, we propose TruthX, an inference-time intervention method to activate the truthfulness of LLM by identifying and editing the features within LLM's internal representations that govern the truthfulness. TruthX employs an auto-encoder to map LLM's representations into semantic and truthful latent spaces respectively, and applies contrastive learning to identify a truthful editing direction within the truthful space. During inference, by editing LLM's internal representations in truthful space, TruthX effectively enhances the truthfulness of LLM. Experiments show that TruthX improves the truthfulness of 13 advanced LLMs by an average of 20% on TruthfulQA benchmark. Further analyses suggest that TruthX can control LLM to produce truthful or hallucinatory responses via editing only one vector in LLM's internal representations.

cs.CL↗

Tractography with T1-weighted MRI and associated anatomical constraints on clinical quality diffusion MRI

Diffusion MRI (dMRI) streamline tractography, the gold standard for in vivo estimation of brain white matter (WM) pathways, has long been considered indicative of macroscopic relationships with WM microstructure. However, recent advances in tractography demonstrated that convolutional recurrent neural networks (CoRNN) trained with a teacher-student framework have the ability to learn and propagate streamlines directly from T1 and anatomical contexts. Training for this network has previously relied on high-resolution dMRI. In this paper, we generalize the training mechanism to traditional clinical resolution data, which allows generalizability across sensitive and susceptible study populations. We train CoRNN on a small subset of the Baltimore Longitudinal Study of Aging (BLSA), which better resembles clinical protocols. Then, we define a metric, termed the epsilon ball seeding method, to compare T1 tractography and traditional diffusion tractography at the streamline level. Under this metric, T1 tractography generated by CoRNN reproduces diffusion tractography with approximately two millimeters of error.

eess.IV↗

Exploiting dynamic bifurcation in elastic ribbons for mode skipping and selection

In this paper, we systematically study the dynamic snap-through behavior of a pre-deformed elastic ribbon by combining theoretical analysis, discrete numerical simulations, and experiments. By rotating one of its clamped ends with controlled angular speed, we observe two snap-through transition paths among the multiple stable configurations of a ribbon in three-dimensional (3D) space, which is different from the classical snap-through of a two-dimensional (2D) bistable beam. Our theoretical model for the static bifurcation analysis is derived based on the Kirchhoff equations, and dynamical numerical simulations are conducted using the Discrete Elastic Rods (DER) algorithm. The planar beam model is also employed for the asymptotic analysis of dynamic snap-through behaviors. The results show that, since the snap-through processes of both planar beams and 3D ribbons are governed by the saddle-node bifurcation, the same scaling law for the delay applies. We further demonstrate that, in elastic ribbons, by controlling the rotating velocity at the end, distinct snap-through pathways can be realized by selectively skipping specific modes, moreover, particular final modes can be strategically achieved. Through a parametric study using numerical simulations, we construct general phase diagrams for both mode skipping and selection of snapping ribbons. The work serves as a benchmark for future investigations on dynamic snap-through of thin elastic structures and provides guidelines for the novel design of intelligent mechanical systems.

cond-mat.soft↗

ConFL: Constraint-guided Fuzzing for Machine Learning Framework

As machine learning gains prominence in various sectors of society for automated decision-making, concerns have risen regarding potential vulnerabilities in machine learning (ML) frameworks. Nevertheless, testing these frameworks is a daunting task due to their intricate implementation. Previous research on fuzzing ML frameworks has struggled to effectively extract input constraints and generate valid inputs, leading to extended fuzzing durations for deep execution or revealing the target crash. In this paper, we propose ConFL, a constraint-guided fuzzer for ML frameworks. ConFL automatically extracting constraints from kernel codes without the need for any prior knowledge. Guided by the constraints, ConFL is able to generate valid inputs that can pass the verification and explore deeper paths of kernel codes. In addition, we design a grouping technique to boost the fuzzing efficiency. To demonstrate the effectiveness of ConFL, we evaluated its performance mainly on Tensorflow. We find that ConFL is able to cover more code lines, and generate more valid inputs than state-of-the-art (SOTA) fuzzers. More importantly, ConFL found 84 previously unknown vulnerabilities in different versions of Tensorflow, all of which were assigned with new CVE ids, of which 3 were critical-severity and 13 were high-severity. We also extended ConFL to test PyTorch and Paddle, 7 vulnerabilities are found to date.

cs.SE↗

Avoiding localization instabilities in rotary pleating

Rotary pleating is a widely used process for making filters out of nonwoven fabric sheets. This involves indirect elastic-plastic bending of pre-weakened creases by continuously injecting material into an accordion-shaped pack. This step can fail through a localization instability that creates a kink in a pleat facet instead of in the desired crease location. In the present work, we consider the effects of geometric and material parameters on the rotary pleating process. We formulate the process as a multi-point variable-arc-length boundary value problem for planar inextensible rods, with hinge connections. Both the facets (rods) and creases (hinges) obey nonlinear moment-curvature or moment-angle constitutive laws. Some unexpected aspects of the sleeve boundary condition at the point of material injection, common to many continuous sheet processes, are noted. The process, modeled as quasistatic, features multiple equilibria which we explore by numerical continuation. The presence of, presumably stable, kinked equilibria is taken as a conservative sign of potential pleating failure. Failure may also occur due to localization at the injection point. We may thus obtain "pleatability surfaces" that separate the parameter space into regions where mechanical pleating will succeed or fail. Successful pleating depends primarily on the distance between the injection point and the pleated pack. Other factors, such as the crease stiffness and strength relative to that of the facets, also have an influence. Our approach can be adapted to study other pleating and forming processes, the deployment and collapse of folded structures, or multi-stability in compliant structures.

cond-mat.soft↗

Continuous modeling of creased annuli with tunable bistable and looping behaviors

Creases are purposely introduced to thin structures for designing deployable origami, artistic geometries, and functional structures with tunable nonlinear mechanics. Modeling the mechanics of creased structures is challenging because creases introduce geometric discontinuity and often have complex mechanical responses due to the local material damage. In this work, we propose a continuous description of the sharp geometry of creases and apply it to the study of \emph{creased annuli}, made by introducing radial creases to annular strips with the creases annealed to behave elastically. We find creased annuli have generic bistability and can be folded into various compact shapes, depending on the crease pattern and the overcurvature of the flat annulus. We use a regularized Dirac delta function (RDDF) to describe the geometry of a crease, with the finite spike of the RDDF capturing the localized curvature. Together with anisotropic rod theory, we solve the nonlinear mechanics of creased annuli, with its stability determined by the standard conjugate point test. We find excellent agreement between precision tabletop models, numerical predictions from our analytical framework, and modeling results from finite element simulations. We further show that by varying the rest curvature of the thin strip, dynamic switches between different states of creased annuli can be achieved, which could inspire the design of novel deployable and morphable structures. We believe our smooth description of discontinuous geometries will benefit to the mechanical modeling and design of a wide spectrum of engineering structures that embrace geometric and material discontinuities.

cond-mat.soft↗