arXiv subjects

Tianyi Hu

Publications and source records attributed to Tianyi Hu.

At least 19 recordsLinked to original sources

arXiv2026

TransMeme: A Multi-Agent Framework for Cross-Cultural Meme Transcreation

Internet memes are a pervasive form of multimodal online communication; however, such communication often involves users from diverse linguistic and cultural backgrounds. Therefore, adapting memes across cultures and languages is a central challenge for enabling mutual understanding in online communication. Unlike ordinary translation or standalone text rewriting, cross-cultural meme transcreation must jointly preserve communicative intent, adapt culture-dependent meaning for the target audience, and maintain coherence between text and image. In this work, we first provide an explicit task analysis of cross-cultural meme transcreation and identify three core challenges: culture-specific knowledge understanding, intent and tone preservation, and multimodal consistency. Based on this analysis, we propose a multi-agent framework with specialized agents that are coordinated to address these challenges through cultural adaptation, target text rewriting, revision, and conditional visual adjustment. The framework strengthens target text adaptation with coordinated feedback to handle difficult cases that require deeper cultural or visual intervention. We evaluate the framework on bidirectional Chinese-English meme transcreation using both human evaluation and LLM-as-a-Judge. Our method consistently outperforms all baselines across both evaluation settings. In human evaluation, it achieves the best performance on all four dimensions and delivers a 33.1% average improvement over the strongest baseline, while in LLM-as-a-Judge, it attains the highest Top-1 ranking rate (60% versus 26% for the second-best baseline). Further analysis indicates that each component contributes to the performance. Our error analysis suggests that the remaining bottlenecks lie in humor reconstruction and image-text alignment rather than simple cultural knowledge gaps, pointing to future work on humor transfer.

cs.AI↗

arXiv2026

SHARE: Towards Head-Mounted AR with User-Centric SLAM in Shared Human-Robot Workspaces

Human-Robot Collaboration (HRC) in shared physical spaces using Augmented Reality (AR) interfaces is powered by Simultaneous Localization and Mapping (SLAM). Existing multi-agent SLAM systems rely on an edge server to combine visual findings of multiple resource-constrained agents, perform computation, and schedule updates to their local maps. However, the edge treats all agents uniformly and ignores the fundamentally different latency requirements of heterogeneous HRC agents: robots and head-mounted AR users. This uniform resource allocation often results in high lag for user manipulation, as it does not meet the stringent latency requirements of AR. In this work, we design, implement, and evaluate SHARE, a user-centric SLAM system that strategically prioritizes AR user experience while maintaining accurate tracking performance for robots. SHARE builds a first-of-its-kind experience model for HRC agents and adaptively adjusts transmission priorities to match it. To reduce end-to-end latency, SHARE leverages the redundancy of visual features acquired by agents in shared human-robot workspaces to reduce computation time induced by edge-based processing. Real-world deployment with commercial AR headsets and a ground robot achieves 13.22 ms average latency for AR users (43.3% reduction from baseline) while maintaining sub-2-centimeter tracking accuracy. User studies further reveal statistically significant improvements in user perception.

cs.HC↗

arXiv2026

CHASM: Unveiling Covert Advertisements on Chinese Social Media

Current benchmarks for evaluating large language models (LLMs) in social media moderation completely overlook a serious threat: covert advertisements, which disguise themselves as regular posts to deceive and mislead consumers into making purchases, leading to significant ethical and legal concerns. In this paper, we present the CHASM, a first-of-its-kind dataset designed to evaluate the capability of Multimodal Large Language Models (MLLMs) in detecting covert advertisements on social media. CHASM is a high-quality, anonymized, manually curated dataset consisting of 4,992 instances, based on real-world scenarios from the Chinese social media platform Rednote. The dataset was collected and annotated under strict privacy protection and quality control protocols. It includes many product experience sharing posts that closely resemble covert advertisements, making the dataset particularly challenging.The results show that under both zero-shot and in-context learning settings, none of the current MLLMs are sufficiently reliable for detecting covert advertisements.Our further experiments revealed that fine-tuning open-source MLLMs on our dataset yielded noticeable performance gains. However, significant challenges persist, such as detecting subtle cues in comments and differences in visual and textual structures.We provide in-depth error analysis and outline future research directions. We hope our study can serve as a call for the research community and platform moderators to develop more precise defenses against this emerging threat.

cs.LG↗

arXiv2026

Rhythms of Recovery: Patient-Centered Virtual Reality Exergame for Physical Rehabilitation in the Intensive Care Unit

Early mobilization is a structured protocol designed to facilitate motor recovery in intensive care unit (ICU) patients with ICU-acquired weakness. This process is typically implemented by an interdisciplinary team of nurses, physical therapists, and other healthcare professionals. However, its application is often constrained by the patients' critical conditions, limited mobility, and the challenges of coordinating care within resource-intensive ICU environments. In this study, we developed a patient-centered virtual reality (VR) exergame through an interdisciplinary design process involving clinicians and therapists, tailored to the constraints of critical care. The exergame incorporates progressive mobility levels that mirror early mobilization practices, and includes an embodied avatar to provide guidance and motivation. Using Meta Quest 3 body tracking, the system captures and visualizes patients' movements, thereby providing motivational engagement and quantifiable mobility metrics. We evaluated the exergame in two stages: a dual-user study involving healthy participants and healthcare professionals or students (N = 13), and a subsequent study with cardiothoracic ICU patients (N = 18) to assess feasibility, design validity, and clinical acceptance. Across both studies, participants reported high enjoyment and engagement without discomfort or stress. Furthermore, patients demonstrated increases in movement speed, range of motion, and workspace volume of the upper body across game levels. Physiological monitoring further indicated that the exergame elicited exertion without inducing excessive cardiovascular responses. These findings highlight the feasibility of VR exergames as a clinically acceptable and engaging adjunct to early mobilization in critical care, offering a novel pathway to improve rehabilitation outcomes for ICU patients.

cs.HC↗

arXiv2026

SeeUPO: Sequence-Level Agentic-RL with Convergence Guarantees

Reinforcement learning (RL) has emerged as the predominant paradigm for training large language model (LLM)-based AI agents. However, existing backbone RL algorithms lack verified convergence guarantees in agentic scenarios, especially in multi-turn settings, which can lead to training instability and failure to converge to optimal policies. In this paper, we systematically analyze how different combinations of policy update mechanisms and advantage estimation methods affect convergence properties in single/multi-turn scenarios. We find that REINFORCE with Group Relative Advantage Estimation (GRAE) can converge to the globally optimal under undiscounted conditions, but the combination of PPO & GRAE breaks PPO's original monotonic improvement property. Furthermore, we demonstrate that mainstream backbone RL algorithms cannot simultaneously achieve both critic-free and convergence guarantees in multi-turn scenarios. To address this, we propose SeeUPO (Sequence-level Sequential Update Policy Optimization), a critic-free approach with convergence guarantees for multi-turn interactions. SeeUPO models multi-turn interaction as sequentially executed multi-agent bandit problems. Through turn-by-turn sequential policy updates in reverse execution order, it ensures monotonic improvement and convergence to global optimal solution via backward induction. Experiments on AppWorld and BFCL v4 demonstrate SeeUPO's substantial improvements over existing backbone algorithms: relative gains of 43.3%-54.6% on Qwen3-14B and 24.1%-41.9% on Qwen2.5-14B (averaged across benchmarks), along with superior training stability.

cs.AI↗

arXiv2026

DIVERGE: Diversity-Enhanced RAG for Open-Ended Information Seeking

Existing retrieval-augmented generation (RAG) systems often assume that each query has a single correct answer. This assumption overlooks open-ended information-seeking scenarios where multiple plausible answers are valuable, and where diversity is important for creativity, fairness, and inclusive access to information. We show that standard RAG systems fail to fully use diverse retrieved contexts: simply increasing retrieval diversity does not necessarily lead to diverse generations. To address this limitation, we propose Diverge, a plug-and-play agentic RAG framework that improves the diversity--quality trade-off through iterative, reflection-guided exploration of diverse viewpoints and diversity-aware retrieval support. We further introduce evaluation metrics for characterizing the diversity-quality trade-off in open-ended question answering. Experiments across multiple real-world datasets and backbone LLMs show that Diverge achieves the best trade-off among competitive baselines, increasing diversity by $\sim2\times$ without noticeable quality degradation. These results reveal a systematic limitation of current RAGs and show the value of explicit diversity modeling.

cs.CL↗

arXiv2026

Less Noise, More Voice: Reinforcement Learning for Reasoning via Instruction Purification

Reinforcement Learning with Verifiable Rewards (RLVR) has advanced LLM reasoning, but remains constrained by inefficient exploration under limited rollout budgets, leading to low sampling success and unstable training in complex tasks. We find that many exploration failures arise not from problem difficulty, but from a small number of prompt tokens that introduce interference. Building on this insight, we propose the Less Noise Sampling Framework (LENS), which first prompts by identifying and removing interference tokens. then transfers successful rollouts from the purification process to supervise policy optimization on the original noisy prompts, enabling the model to learn to ignore interference in the real-world, noisy prompting settings. Experimental results show that LENS significantly outperforms GRPO, delivering higher performance and faster convergence, with a 3.88% average gain and over 1.6$\times$ speedup on math reasoning, and a 1.83% gain on scientific and general reasoning. Our work highlights the critical role of pruning interference tokens in improving rollout efficiency, offering a new perspective for RLVR research.

cs.LG↗

arXiv2026

JADE: Bridging the Strategic-Operational Gap in Dynamic Agentic RAG

The evolution of Retrieval-Augmented Generation (RAG) has shifted from static retrieval pipelines to dynamic, agentic workflows where a central planner orchestrates multi-turn reasoning. However, existing paradigms face a critical dichotomy: they either optimize modules jointly within rigid, fixed-graph architectures, or empower dynamic planning while treating executors as frozen, black-box tools. We identify that this \textit{decoupled optimization} creates a ``strategic-operational mismatch,'' where sophisticated planning strategies fail to materialize due to unadapted local executors, often leading to negative performance gains despite increased system complexity. In this paper, we propose \textbf{JADE} (\textbf{J}oint \textbf{A}gentic \textbf{D}ynamic \textbf{E}xecution), a unified framework for the joint optimization of planning and execution within dynamic, multi-turn workflows. By modeling the system as a cooperative multi-agent team unified under a single shared backbone, JADE enables end-to-end learning driven by outcome-based rewards. This approach facilitates \textit{co-adaptation}: the planner learns to operate within the capability boundaries of the executors, while the executors evolve to align with high-level strategic intent. Empirical results demonstrate that JADE transforms disjoint modules into a synergistic system, yielding remarkable performance improvements via joint optimization and enabling a flexible balance between efficiency and effectiveness through dynamic workflow orchestration.

cs.AI↗

arXiv2025

Heterogeneity in Multi-Agent Reinforcement Learning

Heterogeneity is a fundamental property in multi-agent reinforcement learning (MARL), which is closely related not only to the functional differences of agents, but also to policy diversity and environmental interactions. However, the MARL field currently lacks a rigorous definition and deeper understanding of heterogeneity. This paper systematically discusses heterogeneity in MARL from the perspectives of definition, quantification, and utilization. First, based on an agent-level modeling of MARL, we categorize heterogeneity into five types and provide mathematical definitions. Second, we define the concept of heterogeneity distance and propose a practical quantification method. Third, we design a heterogeneity-based multi-agent dynamic parameter sharing algorithm as an example of the application of our methodology. Case studies demonstrate that our method can effectively identify and quantify various types of agent heterogeneity. Experimental results show that the proposed algorithm, compared to other parameter sharing baselines, has better interpretability and stronger adaptability. The proposed methodology will help the MARL community gain a more comprehensive and profound understanding of heterogeneity, and further promote the development of practical algorithms.

cs.MA↗

arXiv2025

Simulating surfactant effects in phase-transforming fluids

Surfactants are critical in natural processes and engineering, but measuring their concentrations in non-equilibrium conditions and in the presence of flow is difficult. Therefore, computational methods are a key tool for improving our understanding. Predicting the effect of surfactants on liquid-vapor transformations is particularly challenging due to (1) simultaneous mass transfer, non-equilibrium thermodynamics and Marangoni stresses, and (2) the phenomenological assumptions underlying many liquid-vapor phase-change models. Starting from the Navier-Stokes-Korteweg equations, a first-principles approach to liquid-vapor phase transformations, we developed a model of liquid-vapor flows with surfactants. We performed simulations of bubbles under equilibrium and liquid-vapor interface oscillations to demonstrate that the model successfully reproduces surfactant-mediated reductions in surface tension. We also investigated the mechanisms whereby surfactant affects bubble coalescence and condensation. Overall, this work provides a new framework for studying the effect of surfactants on liquid-vapor transformations and suggests multiple areas for future research, including the impact of complex surface chemistries on flow around bubbles and the acoustic response of bubbles with surfactants.

physics.flu-dyn↗

arXiv2025

Phase-Field/Discontinuity Capturing operator for direct van der Waals simulation (DVS)

Discontinuity capturing (DC) operators are commonly employed to numerically solve problems involving sharp gradients in the solution. Despite their success, the application of DC operators to the direct van der Waals simulation (DVS) remains challenging. The DVS framework models non-equilibrium phase transitions by admitting interfacial regions in which the derivative of pressure with respect to density is negative. In these regions, we demonstrate that classical DC operators may violate the free energy dissipation law and produce unphysical wave structures. To address this limitation, we propose the phase-field/discontinuity capturing (PF/DC) operator. Numerical results show that PF/DC yields stable and accurate solutions in both bulk fluids and interfacial regions. Finally, we apply the proposed method to simulate cavitating flow over a three-dimensional bluff body, obtaining excellent agreement with experimental data and significant improvements over results produced using classical DC operators.

physics.comp-ph↗

arXiv2025

Voting with the Graph: Stable RLAIF via Topological Consistency Maximization

Reinforcement Learning from AI Feedback (RLAIF) relies on LLM judges as preference measurement instruments, yet these instruments are fundamentally limited by random measurement errors -- stochastic fluctuations that manifest as preference cycles (e.g., $A \succ B \succ C \succ A$), occurring in 5-9% of evaluations across state-of-the-art models. While repeated sampling mitigates noise by averaging multiple judgments, it treats each comparison in isolation and fails to exploit the structural constraints that distinguish systematic signals from random noise. We introduce Topological Consensus Rewards (TCR), a framework that leverages transitivity as a denoising mechanism via topological majority voting: systematic signals reinforce each other through transitive chains, while random errors cluster into topologically exposed cycles. TCR approximates the Maximum Acyclic Subgraph to filter stochastic noise from preference signals. We also propose Cycle Incidence Rate (CIR) as a diagnostic metric that measures the proportion of samples containing preference cycles. Under our noise model, these cycles arise primarily from stochastic measurement errors rather than genuine intransitivity. Experiments on Arena-Hard, MT-Bench, and WritingBench demonstrate that TCR consistently outperforms pairwise baselines and classical ranking algorithms, while exhibiting robust performance across different judge models.

cs.AI↗

arXiv2025

Do LLMs Understand Wine Descriptors Across Cultures? A Benchmark for Cultural Adaptations of Wine Reviews

Recent advances in large language models (LLMs) have opened the door to culture-aware language tasks. We introduce the novel problem of adapting wine reviews across Chinese and English, which goes beyond literal translation by incorporating regional taste preferences and culture-specific flavor descriptors. In a case study on cross-cultural wine review adaptation, we compile the first parallel corpus of professional reviews, containing 8k Chinese and 16k Anglophone reviews. We benchmark both neural-machine-translation baselines and state-of-the-art LLMs with automatic metrics and human evaluation. For the latter, we propose three culture-oriented criteria -- Cultural Proximity, Cultural Neutrality, and Cultural Genuineness -- to assess how naturally a translated review resonates with target-culture readers. Our analysis shows that current models struggle to capture cultural nuances, especially in translating wine descriptions across different cultures. This highlights the challenges and limitations of translation models in handling cultural content.

cs.CL↗

arXiv2025

XR Reality Check: What Commercial Devices Deliver for Spatial Tracking

Inaccurate spatial tracking in extended reality (XR) devices leads to virtual object jitter, misalignment, and user discomfort, fundamentally limiting immersive experiences and natural interactions. In this work, we introduce a novel testbed that enables simultaneous, synchronized evaluation of multiple XR devices under identical environmental and kinematic conditions. Leveraging this platform, we present the first comprehensive empirical benchmarking of five state-of-the-art XR devices across 16 diverse scenarios. Our results reveal substantial intra-device performance variation, with individual devices exhibiting up to 101\% increases in error when operating in featureless environments. We also demonstrate that tracking accuracy strongly correlates with visual conditions and motion dynamics. We also observe significant inter-device disparities, with performance differences of up to 2.8$\times$, which are closely linked to hardware specifications such as sensor configurations and dedicated processing units. Finally, we explore the feasibility of substituting a motion capture system with the Apple Vision Pro as a practical ground truth reference. While the Apple Vision Pro delivers highly accurate relative pose error estimates ($R^2 = 0.830$), its absolute pose error estimation remains limited ($R^2 = 0.387$), highlighting both its potential and its constraints for rigorous XR evaluation. This work establishes the first standardized framework for comparative XR tracking evaluation, providing the research community with reproducible methodologies, comprehensive benchmark datasets, and open-source tools that enable systematic analysis of tracking performance across devices and conditions, thereby accelerating the development of more robust spatial sensing technologies for XR systems.

eess.SY↗

arXiv2025

Will You Be Aware? Eye Tracking-Based Modeling of Situational Awareness in Augmented Reality

Augmented Reality (AR) systems, while enhancing task performance through real-time guidance, pose risks of inducing cognitive tunneling-a hyperfocus on virtual content that compromises situational awareness (SA) in safety-critical scenarios. This paper investigates SA in AR-guided cardiopulmonary resuscitation (CPR), where responders must balance effective compressions with vigilance to unpredictable hazards (e.g., patient vomiting). We developed an AR app on a Magic Leap 2 that overlays real-time CPR feedback (compression depth and rate) and conducted a user study with simulated unexpected incidents (e.g., bleeding) to evaluate SA, in which SA metrics were collected via observation and questionnaires administered during freeze-probe events. Eye tracking analysis revealed that higher SA levels were associated with greater saccadic amplitude and velocity, and with reduced proportion and frequency of fixations on virtual content. To predict SA, we propose FixGraphPool, a graph neural network that structures gaze events (fixations, saccades) into spatiotemporal graphs, effectively capturing dynamic attentional patterns. Our model achieved 83.0% accuracy (F1=81.0%), outperforming feature-based machine learning and state-of-the-art time-series models by leveraging domain knowledge and spatial-temporal information encoded in ET data. These findings demonstrate the potential of eye tracking for SA modeling in AR and highlight its utility in designing AR systems that ensure user safety and situational awareness.

cs.LG↗

arXiv2025

Culinary Crossroads: A RAG Framework for Enhancing Diversity in Cross-Cultural Recipe Adaptation

In cross-cultural recipe adaptation, the goal is not only to ensure cultural appropriateness and retain the original dish's essence, but also to provide diverse options for various dietary needs and preferences. Retrieval Augmented Generation (RAG) is a promising approach, combining the retrieval of real recipes from the target cuisine for cultural adaptability with large language models (LLMs) for relevance. However, it remains unclear whether RAG can generate diverse adaptation results. Our analysis shows that RAG tends to overly rely on a limited portion of the context across generations, failing to produce diverse outputs even when provided with varied contextual inputs. This reveals a key limitation of RAG in creative tasks with multiple valid answers: it fails to leverage contextual diversity for generating varied responses. To address this issue, we propose CARRIAGE, a plug-and-play RAG framework for cross-cultural recipe adaptation that enhances diversity in both retrieval and context organization. To our knowledge, this is the first RAG framework that explicitly aims to generate highly diverse outputs to accommodate multiple user preferences. Our experiments show that CARRIAGE achieves Pareto efficiency in terms of diversity and quality of recipe adaptation compared to closed-book LLMs.

cs.CL↗

arXiv2025

Evaluation Hallucination in Multi-Round Incomplete Information Lateral-Driven Reasoning Tasks

Multi-round incomplete information tasks are crucial for evaluating the lateral thinking capabilities of large language models (LLMs). Currently, research primarily relies on multiple benchmarks and automated evaluation metrics to assess these abilities. However, our study reveals novel insights into the limitations of existing methods, as they often yield misleading results that fail to uncover key issues, such as shortcut-taking behaviors, rigid patterns, and premature task termination. These issues obscure the true reasoning capabilities of LLMs and undermine the reliability of evaluations. To address these limitations, we propose a refined set of evaluation standards, including inspection of reasoning paths, diversified assessment metrics, and comparative analyses with human performance.

cs.CL↗

arXiv2025

CoMoE: Contrastive Representation for Mixture-of-Experts in Parameter-Efficient Fine-tuning

In parameter-efficient fine-tuning, mixture-of-experts (MoE), which involves specializing functionalities into different experts and sparsely activating them appropriately, has been widely adopted as a promising approach to trade-off between model capacity and computation overhead. However, current MoE variants fall short on heterogeneous datasets, ignoring the fact that experts may learn similar knowledge, resulting in the underutilization of MoE's capacity. In this paper, we propose Contrastive Representation for MoE (CoMoE), a novel method to promote modularization and specialization in MoE, where the experts are trained along with a contrastive objective by sampling from activated and inactivated experts in top-k routing. We demonstrate that such a contrastive objective recovers the mutual-information gap between inputs and the two types of experts. Experiments on several benchmarks and in multi-task settings demonstrate that CoMoE can consistently enhance MoE's capacity and promote modularization among the experts.

cs.LG↗