arXiv subjects

Wei Zhu

Publications and source records attributed to Wei Zhu.

At least 55 records · Page 3Linked to original sources

arXiv2026

TaFall: Balance-Informed Fall Detection via Passive Thermal Sensing

Falls are a major cause of injury and mortality among older adults, yet most incidents occur in private indoor environments where monitoring must balance effectiveness with privacy. Existing privacy-preserving fall detection approaches, particularly those based on radio frequency sensing, often rely on coarse motion cues, which limits reliability in real-world deployments. We introduce TaFall, a balance-informed fall detection system based on low-cost, privacy-preserving thermal array sensing. The key insight is that TaFall models a fall as a process of balance degradation and detects falls by estimating pose-driven biomechanical balance dynamics. To enable this capability from low-resolution thermal array maps, we propose (i) an appearance-motion fusion model for robust pose reconstruction, (ii) physically grounded balance-aware learning, and (iii) pose-bridged pretraining to improve robustness. TaFall achieves a detection rate of 98.26% with a false alarm rate of 0.65% on our dataset with over 3,000 fall instances from 35 participants across diverse indoor environments. In 27 day deployments across four homes, TaFall attains an ultra-low false alarm rate of 0.00126% and a pilot bathroom study confirms robustness under moisture and thermal interference. Together, these results establish TaFall as a reliable and privacy-preserving approach to fall detection in everyday living environments.

cs.CV↗

arXiv2026

Critical Majorana fermion at a topological quantum Hall bilayer transition

Quantum Hall bilayers are a uniquely tunable platform that can realize continuous transitions between distinct topological phases of matter. One prominent example is the transition between the Halperin state and the Moore--Read Pfaffian, long predicted to host a critical theory of Majorana fermions but so far not verified in unbiased microscopic simulations. Using the fuzzy sphere regularization, we identify the low-energy spectrum at this transition with the 3D gauged Majorana conformal field theory. We show that the transition is driven by the closing of the neutral fermion gap, and we directly extract the operator content in both integer and half-integer spin sectors. Our results resolve the long-standing question of the nature of a topological phase transition in a setting relevant to quantum Hall experiments, while also providing the first realization of a fermionic theory on the fuzzy sphere, previously limited to bosonic theories.

cond-mat.str-el↗

arXiv2026

PROMO: Promptable Outfitting for Efficient High-Fidelity Virtual Try-On

Virtual Try-on (VTON) has become a core capability for online retail, where realistic try-on results provide reliable fit guidance, reduce returns, and benefit both consumers and merchants. Diffusion-based VTON methods achieve photorealistic synthesis, yet often rely on intricate architectures such as auxiliary reference networks and suffer from slow sampling, making the trade-off between fidelity and efficiency a persistent challenge. We approach VTON as a structured image editing problem that demands strong conditional generation under three key requirements: subject preservation, faithful texture transfer, and seamless harmonization. Under this perspective, our training framework is generic and transfers to broader image editing tasks. Moreover, the paired data produced by VTON constitutes a rich supervisory resource for training general-purpose editors. We present PROMO, a promptable virtual try-on framework built upon a Flow Matching DiT backbone with latent multi-modal conditional concatenation. By leveraging conditioning efficiency and self-reference mechanisms, our approach substantially reduces inference overhead. On standard benchmarks, PROMO surpasses both prior VTON methods and general image editing models in visual fidelity while delivering a competitive balance between quality and speed. These results demonstrate that flow-matching transformers, coupled with latent multi-modal conditioning and self-reference acceleration, offer an effective and training-efficient solution for high-quality virtual try-on.

cs.CV↗

arXiv2026

Two Low Mass-Ratio Microlensing Planets and Two Types of Central-Resonant Degeneracy

We present observations and analysis of two low planet/host mass-ratio ($q$) microlensing planets discovered in high-magnification events. KMT-2025-BLG-0811Lb has $q \sim 4.5 \times 10^{-5}$, and a Bayesian analysis favors a super-Earth/mini-Neptune orbiting an M- or K-dwarf host at a projected separation of $\sim 3$ au. KMT-2025-BLG-0912Lb has $q = 2.6 \times 10^{-4}$ and likely hosts a super-Earth/mini-Neptune around either a low-mass M dwarf or a brown dwarf at $\sim 1$ au. Even with an observing cadence of $Γ> 30~{\rm hr}^{-1}$ during the planetary signal, KMT-2025-BLG-0811 still exhibits the "central-resonant" degeneracy. Reviewing nine such events, we find that the "central-resonant" degeneracy can be divided into two distinct types that occupy separate regions in the plane of $q$ and normalized source radius ($ρ$). Type~I events have similar $q$ but substantially different $ρ$ and are more difficult to resolve from the light curves. For Type~II events, the "resonant" solutions have relatively lower $q$ and larger $ρ$. Our review provides guidance for searching for the alternative solution once one solution has been identified.

astro-ph.EP↗

arXiv2026

KMT-2025-BLG-1314 and KMT-2025-BLG-1392: two microlensing planetary/brown-dwarf candidates analyzed with differentiable code

Analysis of binary-lens microlensing events typically requires intensive computation because of the multimodal and complex posterior distributions. With the recent development of the JAX-based differentiable binary-lensing modeling package microlux, we present an analysis of two microlensing events with planet/brown-dwarf candidates, KMT-2025-BLG-1314 and KMT-2025-BLG-1392. Both events exhibit the "Close/Wide" degeneracy, and KMT-2025-BLG-1314 suffers from the "Planet/Binary" degeneracy and a recently recognized "Point/Finite" degeneracy among the planetary solutions. For KMT-2025-BLG-1314, the binary mass ratio is $\log q \sim -3.5$ for the planetary solutions and $\log q > -1.5$ for the binary solutions, while for KMT-2025-BLG-1392, we find $\log q \sim -1.3$. We show that for the analysis of KMT-2025-BLG-1314, Hamiltonian Monte Carlo (HMC), enabled by microlux, provides robust parameter inference and outperforms traditional Markov chain Monte Carlo (MCMC) methods in the presence of bimodal posteriors.

astro-ph.EP↗

arXiv2026

BreathNet: Generalizable Audio Deepfake Detection via Breath-Cue-Guided Feature Refinement

As deepfake audio becomes more realistic and diverse, developing generalizable countermeasure systems has become crucial. Existing detection methods primarily depend on XLS-R front-end features to improve generalization. Nonetheless, their performance remains limited, partly due to insufficient attention to fine-grained information, such as physiological cues or frequency-domain features. In this paper, we propose BreathNet, a novel audio deepfake detection framework that integrates fine-grained breath information to improve generalization. Specifically, we design BreathFiLM, a feature-wise linear modulation mechanism that selectively amplifies temporal representations based on the presence of breathing sounds. BreathFiLM is trained jointly with the XLS-R extractor, in turn encouraging the extractor to learn and encode breath-related cues into the temporal features. Then, we use the frequency front-end to extract spectral features, which are then fused with temporal features to provide complementary information introduced by vocoders or compression artifacts. Additionally, we propose a group of feature losses comprising Positive-only Supervised Contrastive Loss (PSCL), center loss, and contrast loss. These losses jointly enhance the discriminative ability, encouraging the model to separate bona fide and deepfake samples more effectively in the feature space. Extensive experiments on five benchmark datasets demonstrate state-of-the-art (SOTA) performance. Using the ASVspoof 2019 LA training set, our method attains 1.99% average EER across four related eval benchmarks, with particularly strong performance on the In-the-Wild dataset, where it achieves 4.70% EER. Moreover, under the ASVspoof5 evaluation protocol, our method achieves an EER of 4.94% on this latest benchmark.

cs.SD↗

arXiv2026

Pursuing Best Industrial Practices for Retrieval-Augmented Generation in the Medical Domain

While retrieval augmented generation (RAG) has been swiftly adopted in industrial applications based on large language models (LLMs), there is no consensus on what are the best practices for building a RAG system in terms of what are the components, how to organize these components and how to implement each component for the industrial applications, especially in the medical domain. In this work, we first carefully analyze each component of the RAG system and propose practical alternatives for each component. Then, we conduct systematic evaluations on three types of tasks, revealing the best practices for improving the RAG system and how LLM-based RAG systems make trade-offs between performance and efficiency.

cs.CL↗

arXiv2026

ACL: Aligned Contrastive Learning Improves BERT and Multi-exit BERT Fine-tuning

Despite its success in self-supervised learning, contrastive learning is less studied in the supervised setting. In this work, we first use a set of pilot experiments to show that in the supervised setting, the cross-entropy loss objective (CE) and the contrastive learning objective often conflict with each other, thus hindering the applications of CL in supervised settings. To resolve this problem, we introduce a novel \underline{A}ligned \underline{C}ontrastive \underline{L}earning (ACL) framework. First, ACL-Embed regards label embeddings as extra augmented samples with different labels and employs contrastive learning to align the label embeddings with its samples' representations. Second, to facilitate the optimization of ACL-Embed objective combined with the CE loss, we propose ACL-Grad, which will discard the ACL-Embed term if the two objectives are in conflict. To further enhance the performances of intermediate exits of multi-exit BERT, we further propose cross-layer ACL (ACL-CL), which is to ask the teacher exit to guide the optimization of student shallow exits. Extensive experiments on the GLUE benchmark results in the following takeaways: (a) ACL-BRT outperforms or performs comparably with CE and CE+SCL on the GLUE tasks; (b) ACL, especially CL-ACL, significantly surpasses the baseline methods on the fine-tuning of multi-exit BERT, thus providing better quality-speed tradeoffs for low-latency applications.

cs.CL↗

arXiv2026

A Generative Model for Joint Multiple Intent Detection and Slot Filling

In task-oriented dialogue systems, spoken language understanding (SLU) is a critical component, which consists of two sub-tasks, intent detection and slot filling. Most existing methods focus on the single-intent SLU, where each utterance only has one intent. However, in real-world scenarios users usually express multiple intents in an utterance, which poses a challenge for existing dialogue systems and datasets. In this paper, we propose a generative framework to simultaneously address multiple intent detection and slot filling. In particular, an attention-over-attention decoder is proposed to handle the variable number of intents and the interference between the two sub-tasks by incorporating an inductive bias into the process of multi-task learning. Besides, we construct two new multi-intent SLU datasets based on single-intent utterances by taking advantage of the next sentence prediction (NSP) head of the BERT model. Experimental results demonstrate that our proposed attention-over-attention generative model achieves state-of-the-art performance on two public datasets, MixATIS and MixSNIPS, and our constructed datasets.

cs.CL↗

arXiv2026

MRAG: Benchmarking Retrieval-Augmented Generation for Bio-medicine

While Retrieval-Augmented Generation (RAG) has been swiftly adopted in scientific and clinical QA systems, a comprehensive evaluation benchmark in the medical domain is lacking. To address this gap, we introduce the Medical Retrieval-Augmented Generation (MRAG) benchmark, covering various tasks in English and Chinese languages, and building a corpus with Wikipedia and Pubmed. Additionally, we develop the MRAG-Toolkit, facilitating systematic exploration of different RAG components. Our experiments reveal that: (a) RAG enhances LLM reliability across MRAG tasks. (b) the performance of RAG systems is influenced by retrieval approaches, model sizes, and prompting strategies. (c) While RAG improves usefulness and reasoning quality, LLM responses may become slightly less readable for long-form questions. We will release the MRAG-Bench's dataset and toolkit with CCBY-4.0 license upon acceptance, to facilitate applications from both academia and industry.

cs.CL↗

arXiv2026

Evaluating ChatGPT on Medical Information Extraction Tasks: Performance, Explainability and Beyond

Large Language Models (LLMs) like ChatGPT have demonstrated amazing capabilities in comprehending user intents and generate reasonable and useful responses. Beside their ability to chat, their capabilities in various natural language processing (NLP) tasks are of interest to the research community. In this paper, we focus on assessing the overall ability of ChatGPT in 4 different medical information extraction (MedIE) tasks across 6 benchmark datasets. We present the systematically analysis by measuring ChatGPT's performance, explainability, confidence, faithfulness, and uncertainty. Our experiments reveal that: (a) ChatGPT's performance scores on MedIE tasks fall behind those of the fine-tuned baseline models. (b) ChatGPT can provide high-quality explanations for its decisions, however, ChatGPT is over-confident in its predcitions. (c) ChatGPT demonstrates a high level of faithfulness to the original text in the majority of cases. (d) The uncertainty in generation causes uncertainty in information extraction results, thus may hinder its applications in MedIE tasks.

cs.CL↗

arXiv2026

Incompressible quantum liquid on the four-dimensional sphere

The study of quantum Hall effect (QHE) is a foundation of topological physics, inspiring extensive explorations of its high-dimensional generalizations. Notably, the four dimensional (4D) QHE has been experimentally realized in synthetic quantum systems, including cold atoms, photonic lattices, and metamaterials. However, the many-body effect in the 4D QHE system remains poorly understood. In this study, we explore this problem by formulating the microscopic wavefunctions inspired by Laughlin's seminal work. Employing a generalized pseudo-potential framework, we derive an exact microscopic Hamiltonian consisting of two-body projectors that annihilate the microscopic wavefunctions. Diagonalizations on a small size system show that the quasi-hole states remain zero energy while the quasi-particle states exhibit a finite gap, in consistency with an incompressible state. Furthermore, the pairing distribution is calculated to substantiate the liquid-like nature of the wavefunction. Our work provides a preliminary understanding to the fractional topological states in high dimension.

cond-mat.str-el↗

arXiv2026

SYMPHONY: Synergistic Multi-agent Planning with Heterogeneous Language Model Assembly

Recent advancements have increasingly focused on leveraging large language models (LLMs) to construct autonomous agents for complex problem-solving tasks. However, existing approaches predominantly employ a single-agent framework to generate search branches and estimate rewards during Monte Carlo Tree Search (MCTS) planning. This single-agent paradigm inherently limits exploration capabilities, often resulting in insufficient diversity among generated branches and suboptimal planning performance. To overcome these limitations, we propose Synergistic Multi-agent Planning with Heterogeneous langauge model assembly (SYMPHONY), a novel multi-agent planning framework that integrates a pool of heterogeneous language model-based agents. By leveraging diverse reasoning patterns across agents, SYMPHONY enhances rollout diversity and facilitates more effective exploration. Empirical results across multiple benchmark tasks show that SYMPHONY achieves strong performance even when instantiated with open-source LLMs deployable on consumer-grade hardware. When enhanced with cloud-based LLMs accessible via API, SYMPHONY demonstrates further improvements, outperforming existing state-of-the-art baselines and underscoring the effectiveness of heterogeneous multi-agent coordination in planning tasks.

cs.AI↗

arXiv2026

Task-Aware LLM Council with Adaptive Decision Pathways for Decision Support

Large language models (LLMs) have shown strong capabilities across diverse decision-making tasks. However, existing approaches often overlook the specialization differences among available models, treating all LLMs as uniformly applicable regardless of task characteristics. This limits their ability to adapt to varying reasoning demands and task complexities. In this work, we propose Task-Aware LLM Council (TALC), a task-adaptive decision framework that integrates a council of LLMs with Monte Carlo Tree Search (MCTS) to enable dynamic expert selection and efficient multi-step planning. Each LLM is equipped with a structured success memory profile derived from prior task trajectories, enabling semantic matching between current reasoning context and past successes. At each decision point, TALC routes control to the most contextually appropriate model and estimates node value using a dual-signal mechanism that fuses model-based evaluations with historical utility scores. These signals are adaptively weighted based on intra-node variance and used to guide MCTS selection, allowing the system to balance exploration depth with planning confidence. Experiments on WebShop, HumanEval, and the Game of 24 demonstrate that TALC achieves superior task success rates and improved search efficiency compared to strong baselines, validating the benefits of specialization-aware routing and adaptive planning.

cs.AI↗

arXiv2026

ShapLoRA: Allocation of Low-rank Adaption on Large Language Models via Shapley Value Inspired Importance Estimation

Low-rank adaption (LoRA) is a representative method in the field of parameter-efficient fine-tuning (PEFT), and is key to Democratizating the modern large language models (LLMs). The vanilla LoRA is implemented with uniform ranks, and the recent literature have found that properly allocating ranks on the LLM backbones results in performance boosts. However, the previous rank allocation methods have limitations since they rely on inexplanable and unreliable importance measures for the LoRA ranks. To address the above issues, we propose the ShapLoRA framework. Inspired by the explanable attribution measure Shapley Value, we combine the sensitivity-based measures with the idea of coalitions in the collaborative games among LoRA ranks, and propose a more explainable importance measure called Shapley sensitivity. In addition, we optimize the workflow of the existing works by: (a) calculating Shapley sensitivity on a separate validation set; (b) Setting up the allocating-retraining procedures for fair comparisons. We have conducted experiments on various challenging tasks, and the experimental results demonstrate that our ShapLoRA method can outperform the recent baselines with comparable tunable parameters.\footnote{Codes and fine-tuned models will be open-sourced to facilitate future research.

cs.CL↗

arXiv2026

Six New Circumbinary Disk Occultation (CBO) Candidates from the Zwicky Transient Facility

Circumbinary disk occultation (CBO) systems, in which a misaligned circumbinary disk periodically obscures the central binary, provide unique probes of disk structure and dynamics. However, fewer than ten candidates with measured orbital periods were previously known. In this work, we identify six new CBO candidates, designated ZTF-CBO-1 through ZTF-CBO-6, through a systematic search of Zwicky Transient Facility (ZTF) photometry. These systems exhibit deep ($\gtrsim$1 mag) periodic dimming events with orbital periods ranging from $\sim$30 to $\sim$530 days and occultation durations spanning 32%-77% of their orbital periods. Such large duty cycles rule out the interpretation of circum-companion material occultation. Spectral energy distributions of ZTF-CBO-1 and ZTF-CBO-2 reveal infrared excess indicative of a dust component. TESS observations of ZTF-CBO-1 show no hours- to days-timescale variability during ingress and egress, indicating a smooth disk edge. These discoveries nearly double the known CBO sample, enabling meaningful population-level studies.

astro-ph.SR↗

arXiv2026

EmoBipedNav: Emotion-aware Social Navigation for Bipedal Robots with Deep Reinforcement Learning

This study presents an emotion-aware navigation framework -- EmoBipedNav -- using deep reinforcement learning (DRL) for bipedal robots walking in socially interactive environments. The inherent locomotion constraints of bipedal robots challenge their safe maneuvering capabilities in dynamic environments. When combined with the intricacies of social environments, including pedestrian interactions and social cues, such as emotions, these challenges become even more pronounced. To address these coupled problems, we propose a two-stage pipeline that considers both bipedal locomotion constraints and complex social environments. Specifically, social navigation scenarios are represented using sequential LiDAR grid maps (LGMs), from which we extract latent features, including collision regions, emotion-related discomfort zones, social interactions, and the spatio-temporal dynamics of evolving environments. The extracted features are directly mapped to the actions of reduced-order models (ROMs) through a DRL architecture. Furthermore, the proposed framework incorporates full-order dynamics and locomotion constraints during training, effectively accounting for tracking errors and restrictions of the locomotion controller while planning the trajectory with ROMs. Comprehensive experiments demonstrate that our approach exceeds both model-based planners and DRL-based baselines. The hardware videos and open-source code are available at https://gatech-lidar.github.io/emobipednav.github.io/.

cs.RO↗

arXiv2025

Deformation and Stress Evolution during Laser Powder Bed Fusion of Semi-Crystalline Polyamide-12

Laser powder bed fusion (L-PBF) of semi-crystalline polymers such as polyamide-12 (PA12) has found increasing use in various industrial applications. However, achieving high dimensional accuracy remains a significant challenge. Despite the seemingly straightforward layer-by-layer manufacturing concept, the L-PBF process involves complex thermal histories and strongly coupled multiphysics, making the evolution of stress and deformation mechanisms still not fully understood. To address this, a comprehensive three-dimensional thermo-mechanical modeling framework is developed to simulate the L-PBF process of PA12. The model for the first time incorporates transient heat transfer, phase transformation induced volumetric shrinkage, thermoviscoelasticity, and a modified non-isothermal crystallization kinetics. To alleviate the computational burden of part-scale simulations, a dual-mesh strategy is employed to efficiently couple thermal and mechanical fields without compromising numerical accuracy, which also enables the framework to handle L-PBF simulations of arbitrarily complex three-dimensional geometries. Particular attention is paid to the role of mechanical and thermal boundary conditions. Specifically, the underlying powder bed is modeled as a fictitious viscous medium, providing support while permitting upward displacement. Additionally, a radiative heat loss boundary condition, which more closely approximates the actual physical process, is applied to the top powder surface. The incorporation of this radiation effect significantly enhances the crystallization rate and improves the agreement with experimentally measured warpage. The model is validated against experimental warpage data under various preheating temperatures. Furthermore, strain decoupling analysis for the first time reveals that displacement induced by phase transformation is approximately 10 times greater than that caused by thermal expansion, highlighting the dominant role of crystallizationinduced shrinkage in warpage formation. Numerical tests also indicate that warpage is highly sensitive to the preheating target temperature of the PA12 powder bed, while the temperature of the newly recoated powder within the tested range has a limited effect. This work provides a predictive modeling foundation for future optimization of polymer L-PBF processes at part-scale, particularly in controlling deformation and improving dimensional accuracy.

physics.med-ph↗