arXiv subjects

Shijie Liu

Publications and source records attributed to Shijie Liu.

At least 19 recordsLinked to original sources

arXiv2026

Fox in the Henhouse: Supply-Chain Backdoor Attacks Against Reinforcement Learning

The current state-of-the-art backdoor attacks against Reinforcement Learning (RL) rely upon unrealistically permissive access models, that assume the attacker can read (or even write) the victim's policy parameters, observations, or rewards. In this work, we question whether such a strong assumption is required to launch backdoor attacks against RL. To answer this question, we propose the \underline{S}upply-\underline{C}h\underline{a}in \underline{B}ackdoor (SCAB) attack, which targets a common RL workflow: training agents using external agents that are provided separately or embedded within the environment. In contrast to prior works, our attack only relies on legitimate interactions of the RL agent with the supplied agents. Despite this limited access model, by poisoning a mere $3\%$ of training experiences, our attack can successfully activate over $90\%$ of triggered actions, reducing the average episodic return by $80\%$ for the victim. Our novel attack demonstrates that RL attacks are likely to become a reality under untrusted RL training supply-chains.

cs.LG↗

arXiv2026

SA-LIVO: Efficient LiDAR-Inertial-Visual Odometry with Subspace-Aware Degeneracy Handling

Tightly coupled LiDAR-inertial-visual odometry (LIVO) fuses geometric depth with visual measurements, but its exteroceptive sensors fail independently: LiDAR when scan geometry is under-constrained, vision under poor illumination or texture absence. Existing countermeasures (binary degeneracy detection, covariance inflation, scene-level quality gating) act at the modality level, so a single isotropic gain sends visual residuals into directions LiDAR already constrains well and cannot concentrate them where constraints are deficient. We propose Subspace-Aware LiDAR-inertial-visual odometry (SA-LIVO), whose Subspace-Aware Information Fusion (SAIF) eigendecomposes the joint LiDAR-visual information matrix and gates each eigendirection by a single-threshold linear clamp, attenuating low-amplitude directions while passing well-observed ones at full strength; robust per-residual gating and a scene-level quality factor screen corrupted measurements. LiDAR and visual residuals share one invariant extended Kalman filter (InEKF) loop and linearization point, letting photometric Jacobians be assembled once and reused across iterations. On 29 public-benchmark sequences (HILTI'22, Newer College Dataset (NCD), Oxford Spires), plus additional concurrent-degradation scenarios, SA-LIVO matches the strongest baselines in accuracy and stays bounded where competing systems diverge. On the HILTI'22 subset that every baseline completes, it averages 12.3 ms per frame on a laptop CPU and 26.8 ms on an embedded ARM board without GPU, at 3.6-6.3x lower peak memory.

cs.RO↗

arXiv2026

Fortifying Time Series: DTW-Certified Robust Anomaly Detection

Time-series anomaly detection is critical for ensuring safety in high-stakes applications, where robustness is a fundamental requirement rather than a mere performance metric. Addressing the vulnerability of these systems to adversarial manipulation is therefore essential. Existing defenses are largely heuristic or provide certified robustness only under $\ell_p$-norm constraints, which are incompatible with time-series data. In particular, $\ell_p$-norm fails to capture the intrinsic temporal structure in time series, causing small temporal distortions to significantly alter the $\ell_p$-norm measures. Instead, the similarity metric \emph{Dynamic Time Warping} (DTW) is more suitable and widely adopted in the time-series domain, as DTW accounts for temporal alignment and remains robust to temporal variations. To date, however, there has been no certifiable robustness result in this metric that provides guarantees. In this work, we introduce the first \emph{DTW-certified robust defense} in time-series anomaly detection by adapting the randomized smoothing paradigm. We develop this certificate by bridging the $\ell_p$-norm to DTW distance through a lower-bound transformation. Extensive experiments across various datasets and models validate the effectiveness and practicality of our theoretical approach. Results demonstrate significantly improved performance, e.g., up to 18.7\% in F1-score under DTW-based adversarial attacks compared to traditional certified models.

cs.LG↗

arXiv2026

MTServe: Efficient Serving for Generative Recommendation Models with Hierarchical Caches

Generative recommendation (GR) offers superior modeling capabilities but suffers from prohibitive inference costs due to the repeated encoding of long user histories. While cross-request Key-Value (KV) cache reuse presents a significant optimization opportunity, the massive scale of individual user states creates a storage explosion that far exceeds physical GPU limits. We propose MTServe, a hierarchical cache management system that virtualizes GPU memory by leveraging host RAM as a scalable backup store. To bridge the I/O gap between tiers, MTServe introduces a suite of system-level optimizations, including a hybrid storage layout, an asynchronous data transfer pipeline, and a locality-driven replacement policy. On both public and production datasets, MTServe delivers up to 3.1* speedup while maintaining near-perfect hit ratios (>98.5%).

cs.LG↗

arXiv2026

PhysInOne: Visual Physics Learning and Reasoning in One Suite

We present PhysInOne, a large-scale synthetic dataset addressing the critical scarcity of physically-grounded training data for AI systems. Unlike existing datasets limited to merely hundreds or thousands of examples, PhysInOne provides 2 million videos across 153,810 dynamic 3D scenes, covering 71 basic physical phenomena in mechanics, optics, fluid dynamics, and magnetism. Distinct from previous works, our scenes feature multiobject interactions against complex backgrounds, with comprehensive ground-truth annotations including 3D geometry, semantics, dynamic motion, physical properties, and text descriptions. We demonstrate PhysInOne's efficacy across four emerging applications: physics-aware video generation, long-/short-term future frame prediction, physical property estimation, and motion transfer. Experiments show that fine-tuning foundation models on PhysInOne significantly enhances physical plausibility, while also exposing critical gaps in modeling complex physical dynamics and estimating intrinsic properties. As the largest dataset of its kind, orders of magnitude beyond prior works, PhysInOne establishes a new benchmark for advancing physics-grounded world models in generation, simulation, and embodied AI.

cs.CV↗

arXiv2026

Omni-LIVO: Robust RGB-Colored Multi-Camera Visual-Inertial-LiDAR Odometry via Photometric Migration and ESIKF Fusion

Wide field-of-view (FoV) LiDAR sensors provide dense geometry across large environments, but existing LiDAR-inertial-visual odometry (LIVO) systems generally rely on a single camera, limiting their ability to fully exploit LiDAR-derived depth for photometric alignment and scene colorization. We present Omni-LIVO, a tightly coupled multi-camera LIVO system that leverages multi-view observations to comprehensively utilize LiDAR geometric information across extended spatial regions. Omni-LIVO introduces a Cross-View direct alignment strategy that maintains photometric consistency across non-overlapping views, and extends the Error-State Iterated Kalman Filter (ESIKF) with multi-view updates and adaptive covariance. The system is evaluated on public benchmarks and our custom dataset, showing improved accuracy and robustness over state-of-the-art LIVO, LIO, and visual-inertial SLAM baselines. Code and dataset will be released upon publication.

cs.RO↗

arXiv2026

HierarchicalKV: A GPU Hash Table with Cache Semantics for Continuous Online Embedding Storage

Traditional GPU hash tables preserve every inserted key -- a dictionary assumption that wastes scarce High Bandwidth Memory (HBM) when embedding tables routinely exceed single-GPU capacity. We challenge this assumption with cache semantics, where policy-driven eviction is a first-class operation. We introduce HierarchicalKV (HKV), the first general-purpose GPU hash table library whose normal full-capacity operating contract is cache-semantic: each full-bucket upsert (update-or-insert) is resolved in place by eviction or admission rejection rather than by rehashing or capacity-induced failure. HKV co-designs four core mechanisms -- cache-line-aligned buckets, in-line score-driven upsert, score-based dynamic dual-bucket selection, and triple-group concurrency -- and uses tiered key-value separation as a scaling enabler beyond HBM. On an NVIDIA H100 NVL GPU, HKV achieves up to 3.9 billion key-value pairs per second (B-KV/s) find throughput, stable across load factors 0.50-1.00 (<5% variation), and delivers 1.4x higher find throughput than WarpCore (the strongest dictionary-semantic GPU baseline at lambda=0.50) and up to 2.6-9.4x over indirection-based GPU baselines. Since its open-source release in October 2022, HKV has been integrated into multiple open-source recommendation frameworks.

cs.DB↗

arXiv2026

Predicting oscillations in complex networks with delayed feedback

Oscillatory dynamics are common features of complex networks, often playing essential roles in regulating function. Across scales from gene regulatory networks to ecosystems, delayed feedback mechanisms are key drivers of system-scale oscillations. The analysis and prediction of such dynamics are highly challenging, however, due to the combination of high-dimensionality, non-linearity and delay. Here, we systematically investigate how structural complexity and delayed feedback jointly induce oscillatory dynamics in complex systems, and introduce an analytic framework comprising theoretical dimension reduction and data-driven prediction. We reveal that oscillations emerge from the interplay of structural complexity and delay, with reduced models uncovering their critical thresholds and showing that greater connectivity lowers the delay required for their onset. Our theory is empirically tested in an experiment on a programmable electronic circuit, where oscillations are observed once structural complexity and feedback delay exceeded the critical thresholds predicted by our theory. Finally, we deploy a reservoir computing pipeline to accurately predict the onset of oscillations directly from timeseries data. Our findings deepen understanding of oscillatory regulation and offer new avenues for predicting dynamics in complex networks.

cond-mat.dis-nn↗

arXiv2025

Multi-level Certified Defense Against Poisoning Attacks in Offline Reinforcement Learning

Similar to other machine learning frameworks, Offline Reinforcement Learning (RL) is shown to be vulnerable to poisoning attacks, due to its reliance on externally sourced datasets, a vulnerability that is exacerbated by its sequential nature. To mitigate the risks posed by RL poisoning, we extend certified defenses to provide larger guarantees against adversarial manipulation, ensuring robustness for both per-state actions, and the overall expected cumulative reward. Our approach leverages properties of Differential Privacy, in a manner that allows this work to span both continuous and discrete spaces, as well as stochastic and deterministic environments -- significantly expanding the scope and applicability of achievable guarantees. Empirical evaluations demonstrate that our approach ensures the performance drops to no more than $50\%$ with up to $7\%$ of the training data poisoned, significantly improving over the $0.008\%$ in prior work~\citep{wu_copa_2022}, while producing certified radii that is $5$ times larger as well. This highlights the potential of our framework to enhance safety and reliability in offline RL.

cs.LG↗

arXiv2025

Limiting on ALP-photon Coupling through GRB221009A

This study investigates the constraints on ALPs parameters through the photon-ALP oscillation model, based on the high-energy photon propagation characteristics of the gamma-ray burst GRB 221009A. We briefly describe the Primakoff process and use it to derive the oscillation probability. Numerical simulations incorporate the GMF, IGMF, and EBL, utilizing the open-source code gammaALPs to simulate photon propagation from the source to Earth. By employing the chi2 statistical method and segmented spectral fitting of low-energy and high-energy observational data from HXMT-GECAM, the dependence of photon survival probability on ALP mass (ma) and coupling strength (ga) is analyzed. Results show that for the ALP mass range 10^-7 < ma < 10^2 neV, the upper limit on the coupling strength is ga < 0.27 * 10^-11 GeV^-1, improving constraints by one order of magnitude compared to the CAST experiment (6.6 * 10^-11 GeV^-1). Notably, high-energy data exhibit significantly stronger constraining power. This work provides a novel theoretical framework and observational basis for indirectly probing ALPs through extreme astrophysical phenomena.

astro-ph.HE↗

arXiv2025

Coherency Improved Explainable Recommendation via Large Language Model

Explainable recommender systems are designed to elucidate the explanation behind each recommendation, enabling users to comprehend the underlying logic. Previous works perform rating prediction and explanation generation in a multi-task manner. However, these works suffer from incoherence between predicted ratings and explanations. To address the issue, we propose a novel framework that employs a large language model (LLM) to generate a rating, transforms it into a rating vector, and finally generates an explanation based on the rating vector and user-item information. Moreover, we propose utilizing publicly available LLMs and pre-trained sentiment analysis models to automatically evaluate the coherence without human annotations. Extensive experimental results on three datasets of explainable recommendation show that the proposed framework is effective, outperforming state-of-the-art baselines with improvements of 7.3\% in explainability and 4.4\% in text quality.

cs.IR↗

arXiv2025

Spin separation and filtering assisted by topological corner states in the Kekulé lattice

Higher-order topological corner states have been realized in two-dimensional Kekulé lattice, which can be further coupled with spin polarization through the implementation of local magnetization. In this work, we numerically investigate the spin-dependent transport properties assisted by topological corner states in the Kekulé lattice. By applying local magnetization and electric potential, the topological corner states are spin polarized with opposite spins localized at different corners, thereby demonstrating a spin-corner state locking mechanism. Transport characteristics, including transmission, local density of states, and local current density, are calculated for a two-terminal setup consisting of a diamond-shaped Kekulé lattice connected to two leads. When opposite local magnetization is applied to the corners, spin-up and spin-down electrons are perfectly separated, forming two spin-polarized conducting channels and leading to spin spatial separation. In the presence of identical local magnetization on both corners and an electric potential at one corner, the spin-polarized corner states can facilitate selective filtering of different spins and generate spin-polarized currents by tuning the energy. Furthermore, spin-resolved transmission diagrams as functions of both the Fermi energy and electric potential are presented, illustrating the global distribution of spin filtering through topological corner states.

cond-mat.mes-hall↗

arXiv2024

High-performance thin-film lithium niobate Mach-Zehnder modulator on thick silica buffering layer

High-speed photonic integrated circuits leveraging the thin-film lithium niobate (TFLN) platform present a promising approach to address the burgeoning global data traffic demands. As a pivotal component, TFLN-based electro-optic (EO) Mach-Zehnder modulators (MZMs) should exhibit low driving voltage, broad operation bandwidth, high extinction ration, and low insertion loss. However, the pursuit of both maximal EO overlap integral and minimal microwave loss necessitates a fundamental compromise between driving voltage and operational bandwidth. Here, we demonstrate high-performance TFLN EO MZMs constructed on a 12-μm-thick silica buried layer using periodic capacitively loaded traveling-wave electrodes. In contrast to their counterparts utilizing undercut etched silicon substrates or quartz substrates, our devices exhibit streamlined fabrication processes and enhanced modulation efficiency. Notably, the fabricated MZMs attains a high modulation efficiency of 1.25 Vcm in the telecom C-band, while maintaining a low EO roll-off of 1.3 dB at 67 GHz. Our demonstration offers a pathway to achieving perfect group velocity matching and break the voltage-bandwidth limit in a simplified configuration suitable for volume fabrication, thereby laying foundational groundwork for the advancement of high-performance TFLN MZMs and benefiting the next-generation PICs in optical telecommunication, signal processing and other applications.

physics.optics↗

arXiv2024

Construction of various time-dependent Hamiltonians on a single photonic chip

Integrated photonics provides an important platform for simulating physical models with high-performance chip-scale devices, where the lattice size and the time-dependence of a model are key ingredients for further enriching the functionality of a photonic chip. Here, we propose and demonstrate the construction of various time-dependent Hamiltonian models using a single microresonator on thin-film lithium niobate chip. Such an integrated microresonator holds high quality factor to 10^6, and supports the construction of the synthetic frequency lattice with effective lattice sites up to 152 under the electro-optic modulation. By further applying a bichromatic modulation composed of two radio-frequency signals oppositely detuned from the resonant frequency in the microresonator, we build different time-dependent Hamiltonians with the time-varying nearest-neighbor coupling strength in synthetic frequency lattice. We measure the temporal features from capturing the dynamic band structures of the lattice and demonstrate a variety of time-dependent synthetic lattice models by engineering the driven pattern of the modulation, highlighting great flexibility of the microresonator. Our work shows a photonic chip for simulating versatile time-dependent Hamiltonians, which pushes forward quantum simulations in integrated photonics with great experimental tunability and reconfigurability.

physics.optics↗

arXiv2024

Enhanced Second-Harmonic Generation in Thin-Film Lithium Niobate Circular Bragg Nanocavity

Second-order nonlinearity gives rise to many distinctive physical phenomena, e.g., second-harmonic generation, which plays an important role in fundamental science and various applications. Lithium niobate, one of the most widely used nonlinear crystals, exhibits strong second-order nonlinear effects and electro-optic properties. However, its moderate refractive index and etching sidewall angle limit its capability in confining light into nanoscales, restricting its application in nanophotonics. Here, we exploit nanocavities formed by second-order circular Bragg gratings, which support resonant anapole modes to achieve highly enhanced SHG in thin film lithium niobate. The CBG nanocavity exhibits a record-high normalized conversion efficiency of $1.21\times10^{-2}\mathrm{cm^2/GW}$ under the pump intensity of $1.9$ $\mathrm{MW/cm^2}$. An SHG enhancement of $42,000$ is realized compared to TFLN. Besides, we also show s- and p-polarization independent SHG in elliptical Bragg nanocavities. This work could inspire studying nonlinear optics at the nanoscale on TFLN as well as other novel photonic platforms.

physics.optics↗

arXiv2024

Et Tu Certifications: Robustness Certificates Yield Better Adversarial Examples

In guaranteeing the absence of adversarial examples in an instance's neighbourhood, certification mechanisms play an important role in demonstrating neural net robustness. In this paper, we ask if these certifications can compromise the very models they help to protect? Our new \emph{Certification Aware Attack} exploits certifications to produce computationally efficient norm-minimising adversarial examples $74 \%$ more often than comparable attacks, while reducing the median perturbation norm by more than $10\%$. While these attacks can be used to assess the tightness of certification bounds, they also highlight that releasing certifications can paradoxically reduce security.

cs.LG↗

arXiv2024

Infrared Image Super-Resolution via Lightweight Information Split Network

Single image super-resolution (SR) is an established pixel-level vision task aimed at reconstructing a high-resolution image from its degraded low-resolution counterpart. Despite the notable advancements achieved by leveraging deep neural networks for SR, most existing deep learning architectures feature an extensive number of layers, leading to high computational complexity and substantial memory demands. These issues become particularly pronounced in the context of infrared image SR, where infrared devices often have stringent storage and computational constraints. To mitigate these challenges, we introduce a novel, efficient, and precise single infrared image SR model, termed the Lightweight Information Split Network (LISN). The LISN comprises four main components: shallow feature extraction, deep feature extraction, dense feature fusion, and high-resolution infrared image reconstruction. A key innovation within this model is the introduction of the Lightweight Information Split Block (LISB) for deep feature extraction. The LISB employs a sequential process to extract hierarchical features, which are then aggregated based on the relevance of the features under consideration. By integrating channel splitting and shift operations, the LISB successfully strikes an optimal balance between enhanced SR performance and a lightweight framework. Comprehensive experimental evaluations reveal that the proposed LISN achieves superior performance over contemporary state-of-the-art methods in terms of both SR quality and model complexity, affirming its efficacy for practical deployment in resource-constrained infrared imaging applications.

eess.IV↗

arXiv2024

Enhancing the Antidote: Improved Pointwise Certifications against Poisoning Attacks

Poisoning attacks can disproportionately influence model behaviour by making small changes to the training corpus. While defences against specific poisoning attacks do exist, they in general do not provide any guarantees, leaving them potentially countered by novel attacks. In contrast, by examining worst-case behaviours Certified Defences make it possible to provide guarantees of the robustness of a sample against adversarial attacks modifying a finite number of training samples, known as pointwise certification. We achieve this by exploiting both Differential Privacy and the Sampled Gaussian Mechanism to ensure the invariance of prediction for each testing instance against finite numbers of poisoned examples. In doing so, our model provides guarantees of adversarial robustness that are more than twice as large as those provided by prior certifications.

cs.LG↗