arXiv subjects

Tao Li

Publications and source records attributed to Tao Li.

At least 145 records · Page 8Linked to original sources

arXiv2025

UniBiomed: A Universal Foundation Model for Grounded Biomedical Image Interpretation

The integration of AI-assisted biomedical image analysis into clinical practice demands AI-generated findings that are not only accurate but also interpretable to clinicians. However, existing biomedical AI models generally lack the ability to simultaneously generate diagnostic findings and localize corresponding biomedical objects. This limitation makes it challenging for clinicians to correlate AI-generated findings with visual evidence (e.g., tiny lesions) in images and interpret the results of AI models. To address this challenge, we introduce UniBiomed, the first universal foundation model for grounded biomedical image interpretation, which is capable of generating accurate diagnostic findings and simultaneously segmenting the corresponding biomedical targets. UniBiomed is based on a novel integration of Multi-modal Large Language Model and Segment Anything Model, which can effectively unify diverse biomedical tasks in universal training for advancing grounded interpretation. To develop UniBiomed, we curate a large-scale dataset comprising over 27 million triplets of images, region annotations, and text descriptions across ten biomedical imaging modalities. Extensive validation on 70 internal and 14 external datasets demonstrated the state-of-the-art performance of UniBiomed in diverse biomedical tasks, including image segmentation, disease recognition, region-aware diagnosis, vision question answering, and report generation. In summary, UniBiomed is a powerful and versatile biomedical foundation model, unlocking the untapped grounded interpretation capability for optimizing AI-assisted biomedical image analysis.

cs.CV↗

arXiv2025

SCOUT: A Defense Against Data Poisoning Attacks in Fine-Tuned Language Models

Backdoor attacks create significant security threats to language models by embedding hidden triggers that manipulate model behavior during inference, presenting critical risks for AI systems deployed in healthcare and other sensitive domains. While existing defenses effectively counter obvious threats such as out-of-context trigger words and safety alignment violations, they fail against sophisticated attacks using contextually-appropriate triggers that blend seamlessly into natural language. This paper introduces three novel contextually-aware attack scenarios that exploit domain-specific knowledge and semantic plausibility: the ViralApp attack targeting social media addiction classification, the Fever attack manipulating medical diagnosis toward hypertension, and the Referral attack steering clinical recommendations. These attacks represent realistic threats where malicious actors exploit domain-specific vocabulary while maintaining semantic coherence, demonstrating how adversaries can weaponize contextual appropriateness to evade conventional detection methods. To counter both traditional and these sophisticated attacks, we present \textbf{SCOUT (Saliency-based Classification Of Untrusted Tokens)}, a novel defense framework that identifies backdoor triggers through token-level saliency analysis rather than traditional context-based detection methods. SCOUT constructs a saliency map by measuring how the removal of individual tokens affects the model's output logits for the target label, enabling detection of both conspicuous and subtle manipulation attempts. We evaluate SCOUT on established benchmark datasets (SST-2, IMDB, AG News) against conventional attacks (BadNet, AddSent, SynBkd, StyleBkd) and our novel attacks, demonstrating that SCOUT successfully detects these sophisticated threats while preserving accuracy on clean inputs.

cs.CR↗

arXiv2025

Seasonal thermal stress analysis of defective mass concrete sidewalls based on the average forming temperature method

Thermal cracking in urban underground sidewalls is frequently observed when structures are cast in summer and enter service in winter, as seasonal temperature gradients act under structural restraint. To quantify the local stress field associated with pre-existing cracks, an orthogonal finite-element simulation matrix of 16 combinations is constructed. Distributions of maximum principal stress () at the surface crack tip and along the upper half of the crack bottom are evaluated using steady-state thermal loading and a linear-elastic constitutive model. Across all cases, pronounced tensile stress concentration occurs at both locations: the maximum ranges from 19.2 to 34.1 MPa at the crack surface end and from 17.2 to 29.4 MPa at the crack bottom. These concentrated values are consistently higher than the stress level at the same locations in an otherwise identical uncracked wall, clarifying how seasonal temperature gradients under restraint amplify local stresses around existing defects. The quantitative ranges reported here provide a basis for risk screening and for formulating practical mitigation measures (e.g., joint spacing and insulation strategies) in the design and operation of urban underground enclosure walls. In addition, three-dimensional simulations of randomly distributed internal voids show that adopting average forming temperature increases the peak tensile stress on void surfaces from 3.42 to 4.40 MPa at 10 deg C and from 5.98 to 6.96 MPa at -5 deg C, further highlighting the risk amplification effect of AFT under cold service conditions.

physics.comp-ph↗

arXiv2025

Formation of abiogenic hydrocarbons in supercritical fluids under Earth's upper mantle conditions

The formation of hydrocarbons in Earth's interior has traditionally been considered to have biogenic origins; however, growing evidence suggests that some hydrocarbons may instead originate abiotically in the deep carbon cycle. It is widely expected that the Fisher-Tropsch-type (FTT) process, which typically refers to the conversion of inorganic carbon to organic matter in geological settings,may also happen in Earth's interior, but the absence of industrial catalysts and aqueous conditions in deep environments suggest that the FTT process can be very different from that in the chemical industry. Here, we performed extensive \textit{ab initio} molecular dynamics (AIMD) simulations ($>$ 2.4 ns) to investigate the FTT synthesis in dry mixture and in aqueous solutions at 10-13 GPa and 1000-1400 K.We found that large hydrocarbon-related species containing C, O, and H($>$C$_2$) are abiotically synthesized via the polymerization of CO without any catalyst. Supercritical water, commonly found in the deep Earth, does not prevent organic molecule formation but restricts product size and carbon reduction.Our studies reveal a previously unrecognized abiogenic route for hydrocarbon synthesis in mantle geofluids. These carbon-containing fluids could potentially migrate from depth to shallower crustal reservoirs, thereby influencing Earth's surface carbon budget.

physics.comp-ph↗

arXiv2025

How do trout regulate patterns of muscle contraction to optimize propulsive efficiency during steady swimming

Understanding efficient fish locomotion offers insights for biomechanics, fluid dynamics, and engineering. Traditional studies often miss the link between neuromuscular control and whole-body movement. To explore energy transfer in carangiform swimming, we created a bio-inspired digital trout. This model combined multibody dynamics, Hill-type muscle modeling, and a high-fidelity fluid-structure interaction algorithm, accurately replicating a real trout's form and properties. Using deep reinforcement learning, the trout's neural system achieved hierarchical spatiotemporal control of muscle activation. We systematically examined how activation strategies affect speed and energy use. Results show that axial myomere coupling-with activation spanning over 0.5 body lengths-is crucial for stable body wave propagation. Moderate muscle contraction duration ([0.1,0.3] of a tail-beat cycle) lets the body and fluid act as a passive damping system, cutting energy use. Additionally, the activation phase lag of myomeres shapes the body wave; if too large, it causes antagonistic contractions that hinder thrust. These findings advance bio-inspired locomotion understanding and aid energy-efficient underwater system design.

physics.flu-dyn↗

arXiv2025

Tomographic beta-gamma spectroscopy of nuclear beta decay

Nuclear $β$ decay, a sensitive probe of nuclear structure and weak interactions, has become a precision test bed for physics beyond the Standard Model, driven by recent advances in spectrometric techniques. Here we introduce tomographic $β$-$γ$ spectroscopy (TBGS) of nuclear $β$ decay, a method that detects the energies of $β$, $γ$, and internal conversion electrons while simultaneously reconstructing the energy deposition vertices. Using the PandaX-4T detector operated as a TBGS, we obtain a precise and unbiased decay scheme of $^{214}$Pb, a key background isotope in searches for dark matter and Majorana neutrinos. For the first time, transitions of $^{214}$Pb to both the ground and excited states of $^{214}$Bi are measured concurrently, revealing discrepancies in branching ratios of up to 4.7$σ$ relative to previous evaluations. Combined with state-of-the-art theoretical spectral-shape calculations, these results establish a new benchmark for background modelling in rare-event searches and highlight the potential of TBGS as a versatile tool for fundamental physics and nuclear applications.

nucl-ex↗

arXiv2025

EEGAgent: A Unified Framework for Automated EEG Analysis Using Large Language Models

Scalable and generalizable analysis of brain activity is essential for advancing both clinical diagnostics and cognitive research. Electroencephalography (EEG), a non-invasive modality with high temporal resolution, has been widely used for brain states analysis. However, most existing EEG models are usually tailored for individual specific tasks, limiting their utility in realistic scenarios where EEG analysis often involves multi-task and continuous reasoning. In this work, we introduce EEGAgent, a general-purpose framework that leverages large language models (LLMs) to schedule and plan multiple tools to automatically complete EEG-related tasks. EEGAgent is capable of performing the key functions: EEG basic information perception, spatiotemporal EEG exploration, EEG event detection, interaction with users, and EEG report generation. To realize these capabilities, we design a toolbox composed of different tools for EEG preprocessing, feature extraction, event detection, etc. These capabilities were evaluated on public datasets, and our EEGAgent can support flexible and interpretable EEG analysis, highlighting its potential for real-world clinical applications.

cs.LG↗

arXiv2025

Closely competing valence bond crystal orders in the ground state of the spin-$\frac{1}{2}$ antiferromagnetic Heisenberg model on the pyrochlore lattice: a large scale unrestricted variational study

The spin-$\frac{1}{2}$ antiferromagnetic Heisenberg model on the pyrochlore lattice(PAFH) is arguably the most well known strongly frustrated quantum magnet in three spatial dimension. However, due to the rapid scaling of Hilbert space with the linear size of such a three dimensional system, the nature of its ground state in the thermodynamic limit remains elusive after about 30 years' intense debate. Here we apply a recently developed powerful algorithm to perform large scale unrestricted variational optimization of the ground state of the spin-$\frac{1}{2}$ PAFH from the resonating valence bond(RVB) theory perspective. We find a highly competitive candidate ground state of the system. This novel state features a maximally resonating valence bond crystal(VBC) pattern with $2\vec{a}_{1}\times2\vec{a}_{2}\times2\vec{a}_{3}$ periodicity. There are at least four levels of hierarchical structure in such a VBC state, with the first and the second level of hierarchy related to the breaking of the inversion and the translational symmetry. Intriguingly, we find that within the RVB framework such a maximally resonating VBC state is almost degenerate with a recently proposed VBC state that is obtained from dressing hard hexagon covering of the pyrochlore lattice, although they have very different structures. We also find that further symmetry breaking will occur in the dressed hard hexagon VBC state under unrestricted optimization, which results in strong disparity in $\langle \hat{\mathbf{S}}^{2}_{u} \rangle$ for up and down tetrahedrons as we observe in the maximally resonating VBC state. We show that the maximally resonating VBC state found here will be favored by a tiny next-neighboring exchange coupling over the dressed hard hexagon covering VBC state.

cond-mat.str-el↗

arXiv2025

Accommodate Knowledge Conflicts in Retrieval-augmented LLMs: Towards Robust Response Generation in the Wild

The proliferation of large language models (LLMs) has significantly advanced intelligent systems. Unfortunately, LLMs often face knowledge conflicts between internal memory and retrieved external information, arising from misinformation, biases, or outdated knowledge. These conflicts undermine response reliability and introduce uncertainty in decision-making. In this work, we analyze how LLMs navigate knowledge conflicts from an information-theoretic perspective and reveal that when conflicting and supplementary information exhibit significant differences, LLMs confidently resolve their preferences and alleviate the uncertainty during their response generation. When this difference is ambiguous, LLMs experience considerable uncertainty about their generation. Based on this insight, we propose Swin-VIB, a novel framework that integrates a pipeline of variational information bottleneck models to adapt the retrieved information difference, facilitating robust response generation of LLMs even in conflicting contexts. Extensive experiments confirm our theoretical analysis and demonstrate the performance of Swin-VIB. Notably, Swin-VIB outperforms all competitive baselines in terms of the accuracy of the multiple-choice task, while improving the EM values in the open-ended QA task by at least 11.14%.

cs.CL↗

arXiv2025

Quantifying nonclassical correlations relative to local channels

Nonclassical correlations are significant physical resources with extensive applications in quantum information processing. We introduce the modified Wigner-Yanase-Dyson skew information of a quantum state relative to a quantum channel, and a quantitative measure of quantum correlations. Their basic properties are explored in detail. Through a specific example, we also compare our correlations measure with the existing one. Moreover, the correlations relative to various channels including the von Neumann measurements, the unitary channels and the twirling channels are analyzed.

quant-ph↗

arXiv2025

Tightening the entropic uncertainty relations with quantum memory in a multipartite scenario

The quantum uncertainty principle stands as a cornerstone and a distinctive feature of quantum mechanics, setting it apart from classical mechanics. We introduce a tripartite quantum-memory-assisted entropic uncertainty relation, and extend the relation to encompass multiple measurements conducted within multipartite systems. The related lower bounds are shown to be tighter than those formulated by Zhang et al. [Phys. Rev. A 108, 012211 (2023)]. Additionally, we present generalized quantum-memory-assisted entropic uncertainty relations (QMA-EURs) tailored for arbitrary positive-operator-valued measures (POVMs). Finally, we demonstrate the applications of our results to both the relative entropy of unilateral coherence and the quantum key distribution protocols.

quant-ph↗

arXiv2025

A further study on the high-precision optics for HIAF-BRing

The High Intensity heavy ion Accelerator Facility (HIAF) successfully accelerated the 18O6+ beam on October 27, 2025. This paper presents a further simulation study on the high-precision optics, namely sliced optics, of the Booster Ring (BRing) at HIAF based on measured magnetic fields, focusing on three aspects: (1) closed-orbit distortion (COD) and variations in optical parameters induced by errors; (2) closed-orbit correction; (3) dynamic aperture. Specifically, detailed investigations are conducted on COD and optical parameter variations caused by magnet alignment errors and dipole magnet field errors, alongside simulations of closed-orbit correction and detailed calculations of BRing's dynamic aperture. Results show the sliced optics outperforms the original optics in COD control. Without chromaticity correction, its dynamic aperture is superior to the original; after chromaticity correction, it remains comparable. This study provides valuable insights for accelerator tuning and optimization.

physics.acc-ph↗

arXiv2025

CBraMod: A Criss-Cross Brain Foundation Model for EEG Decoding

Electroencephalography (EEG) is a non-invasive technique to measure and record brain electrical activity, widely used in various BCI and healthcare applications. Early EEG decoding methods rely on supervised learning, limited by specific tasks and datasets, hindering model performance and generalizability. With the success of large language models, there is a growing body of studies focusing on EEG foundation models. However, these studies still leave challenges: Firstly, most of existing EEG foundation models employ full EEG modeling strategy. It models the spatial and temporal dependencies between all EEG patches together, but ignores that the spatial and temporal dependencies are heterogeneous due to the unique structural characteristics of EEG signals. Secondly, existing EEG foundation models have limited generalizability on a wide range of downstream BCI tasks due to varying formats of EEG data, making it challenging to adapt to. To address these challenges, we propose a novel foundation model called CBraMod. Specifically, we devise a criss-cross transformer as the backbone to thoroughly leverage the structural characteristics of EEG signals, which can model spatial and temporal dependencies separately through two parallel attention mechanisms. And we utilize an asymmetric conditional positional encoding scheme which can encode positional information of EEG patches and be easily adapted to the EEG with diverse formats. CBraMod is pre-trained on a very large corpus of EEG through patch-based masked EEG reconstruction. We evaluate CBraMod on up to 10 downstream BCI tasks (12 public datasets). CBraMod achieves the state-of-the-art performance across the wide range of tasks, proving its strong capability and generalizability. The source code is publicly available at https://github.com/wjq-learning/CBraMod.

eess.SP↗

arXiv2025

Deep Learning in Dental Image Analysis: A Systematic Review of Datasets, Methodologies, and Emerging Challenges

Efficient analysis and processing of dental images are crucial for dentists to achieve accurate diagnosis and optimal treatment planning. However, dental imaging inherently poses several challenges, such as low contrast, metallic artifacts, and variations in projection angles. Combined with the subjectivity arising from differences in clinicians' expertise, manual interpretation often proves time-consuming and prone to inconsistency. Artificial intelligence (AI)-based automated dental image analysis (DIA) offers a promising solution to these issues and has become an integral part of computer-aided dental diagnosis and treatment. Among various AI technologies, deep learning (DL) stands out as the most widely applied and influential approach due to its superior feature extraction and representation capabilities. To comprehensively summarize recent progress in this field, we focus on the two fundamental aspects of DL research-datasets and models. In this paper, we systematically review 260 studies on DL applications in DIA, including 49 papers on publicly available dental datasets and 211 papers on DL-based algorithms. We first introduce the basic concepts of dental imaging and summarize the characteristics and acquisition methods of existing datasets. Then, we present the foundational techniques of DL and categorize relevant models and algorithms according to different DIA tasks, analyzing their network architectures, optimization strategies, training methods, and performance. Furthermore, we summarize commonly used training and evaluation metrics in the DIA domain. Finally, we discuss the current challenges of existing research and outline potential future directions. We hope that this work provides a valuable and systematic reference for researchers in this field. All supplementary materials and detailed comparison tables will be made publicly available on GitHub.

cs.CV↗

arXiv2025

Grazing-sliding bifurcations in planar $\mathbb{Z}_2$-symmetric Filippov systems

This paper aims to explore the effect of $\mathbb{Z}_2$-symmetry on grazing-sliding bifurcations in planar Filippov systems. We consider the scenario where the unperturbed system is $\mathbb{Z}_2$-symmetric and its subsystem exhibits a hyperbolic limit cycle grazing the discontinuity boundary at a fold. Employing differential manifold theory, we reveal the intrinsic quantities of unfolding all bifurcations and rigorously demonstrate the emergence of a codimension-two bifurcation under generic $\mathbb{Z}_2$-symmetric perturbations within the Filippov framework. After deriving an explicit non-degenerate condition with respect to parameters, we systematically establish the complete bifurcation diagram with exact asymptotics for all bifurcation boundaries by displacement map method combined with asymptotic analysis.

math.DS↗

arXiv2025

Globally Stable Discrete Time PID Passivity-based Control of Power Converters: Simulation and Experimental Results

The key idea behind PID Passivity-based Control (PID-PBC) is to leverage the passivity property of PIDs (for all positive gains) and wrap the PID controller around a passive output to ensure global stability in closed-loop. However, the practical applicability of PID-PBC is stymied by two key facts: (i) the vast majority of practical implementations of PIDs is carried-out in discrete time -- discretizing the continuous time dynamical system of the PID; (ii) the well-known problem that passivity is not preserved upon discretization, even with small sampling times. Therefore, two aspects of the PID-PBC must be revisited for its safe practical application. First, we propose a discretization of the PID that ensures its passivity. Second, since the output that is identified as passive for the continuous time system is not necessarily passive for its discrete time version, we construct a new output that ensures the passivity property for the discretization of the system. In this paper, we provide a constructive answer to both issues for the case of power converter models. Instrumental to achieve this objective is the use of the implicit midpoint discretization method -- which is a symplectic integration technique that preserves system invariants. Since the reference value for the output to be regulated in power converters is non-zero, we are henceforth interested in the property of passivity of the incremental model -- currently known as shifted passivity. Therefore, we demonstrate that the resulting discrete-time PID-PBC defines a passive map for the incremental model and establish shifted passivity for the discretized power converter model. Combining these properties, we prove global stability for the feedback interconnection of the power converter with the discretized PID-PBC. The paper also presents simulations and experiments that demonstrate the performance of the proposed discretization.

eess.SY↗

arXiv2025

WebRouter: Query-specific Router via Variational Information Bottleneck for Cost-sensitive Web Agent

LLM-brained web agents offer powerful capabilities for web automation but face a critical cost-performance trade-off. The challenge is amplified by web agents' inherently complex prompts that include goals, action histories, and environmental states, leading to degraded LLM ensemble performance. To address this, we introduce WebRouter, a novel query-specific router trained from an information-theoretic perspective. Our core contribution is a cost-aware Variational Information Bottleneck (ca-VIB) objective, which learns a compressed representation of the input prompt while explicitly penalizing the expected operational cost. Experiments on five real-world websites from the WebVoyager benchmark show that WebRouter reduces operational costs by a striking 87.8\% compared to a GPT-4o baseline, while incurring only a 3.8\% accuracy drop.

cs.CL↗

arXiv2025

Towards Natural Machine Unlearning

Machine unlearning (MU) aims to eliminate information that has been learned from specific training data, namely forgetting data, from a pre-trained model. Currently, the mainstream of existing MU methods involves modifying the forgetting data with incorrect labels and subsequently fine-tuning the model. While learning such incorrect information can indeed remove knowledge, the process is quite unnatural as the unlearning process undesirably reinforces the incorrect information and leads to over-forgetting. Towards more \textit{natural} machine unlearning, we inject correct information from the remaining data to the forgetting samples when changing their labels. Through pairing these adjusted samples with their labels, the model will tend to use the injected correct information and naturally suppress the information meant to be forgotten. Albeit straightforward, such a first step towards natural machine unlearning can significantly outperform current state-of-the-art approaches. In particular, our method substantially reduces the over-forgetting and leads to strong robustness to hyperparameters, making it a promising candidate for practical machine unlearning.

cs.LG↗