arXiv subjects

Tao Li

Publications and source records attributed to Tao Li.

At least 19 recordsLinked to original sources

arXiv2026

Scale-free wave localization at topological multicriticality

Spatial wave localization at boundaries is conventionally restricted to exponential decay. Here, we observe a scale-free boundary mode exhibiting anomalous polynomial decay, highlighted by strictly linear spatial envelopes, driven by topological multicriticality in acoustic metamaterials. We establish a universal geometric bulk-boundary correspondence rooted in singularity theory by mapping n-th order topological multicriticality onto Arnold An catastrophe. We prove that defectiveness in the spatial transfer matrix enforces a (n-1)-th power spatial decay of boundary modes. At a topological tri-critical point, the singularity of winding function manifests as an A2 fold catastrophe. This dictates a strictly linear spatial decay, marking a sharp departure from the conventional exponential scaling paradigm. Using acoustic metamaterial platforms tuned to a tricritical point, we experimentally demonstrate this linear decay envelope in 1D, and further generalize it to two-dimensional systems to realize scale-free corner modes with simultaneous linear decay along orthogonal directions. While existing critical topology frameworks rely on entanglement spectra to diagnose mode existence in gapless regimes, our work bridges catastrophe theory with topological matter at multicriticalities, advancing the field from abstract mode identification to deterministic real-space wavefunction engineering.

physics.optics↗

arXiv2026

Circuit complexity of coupled four-mode squeezed states in two-field inflation

We study the circuit complexity of coupled curvature and entropy perturbations in two-field inflation. We start from the quadratic action and derive the full Hamiltonian. The Hamiltonian includes all diagonal and off-diagonal quadratic terms. We then obtain three coupled equations for the pair amplitudes of the four-mode squeezed state. From these amplitudes, we calculate the two principal squeezing strengths. We use them to determine the Nielsen circuit complexity. For an equal-weight Gaussian circuit, the complexity takes the form $\mathcal{C}_{2,k}=\sqrt{2(r_{1k}^{2}+r_{2k}^{2})}$. It depends on the squeezing strengths but not explicitly on the squeezing angles. Our numerical results show that the complexity grows after horizon crossing in both coupled and uncoupled models. Interfield coupling changes the pair amplitudes and the four-mode squeezed state. However, its effect on the circuit complexity is small for the parameters considered. This result shows that different four-mode pair-amplitude matrices can yield similar values of the chosen equal-weight circuit complexity. The present calculation provides a reference for studying Gaussian circuit complexity in more general two-field inflationary models.

hep-th↗

arXiv2026

EAST: Embodied AI Safety Testing via Systematic Fault Injection in Digital Twins

Task success is not a sufficient measure of safety for embodied artificial intelligence (AI), because robots may complete objectives while violating safety requirements. Existing evaluations often emphasize performance and robustness, but lack a systematic route from documented safety hazards to executable, verifiable tests. We present EAST (Embodied AI Safety Testing), a digital-twin framework that converts literature-grounded safety evidence into fault-injection tests across perception, instructions, environments, actions, and timing. EAST combines an expert-reviewed safety-fault taxonomy and eight executable operators with applicability and validity constraints. Its three-stage workflow generates test specifications, verifies fault realization during simulated execution, and independently assesses safety challenges, violations, task outcomes, and behavioral signatures using simulator-state evidence. The framework preserves execution traces and replayable records, supporting attribution and reproducibility across experiments. We evaluate EAST on six RoboTwin manipulation tasks and two policy architectures, ACT and RDT, against alternative testing strategies under matched budgets. Across five comparative campaigns, EAST targets all 74 admissible task-policy-operator combinations and obtains valid evidence for 63, while achieving safety-challenge exposure rates close to Random-Legal and Rule-Based. Moreover, 82 valid EAST executions complete their tasks despite safety violations, revealing risks concealed by success-oriented evaluation. These findings support systematic, validity-aware, reproducible safety testing beyond task-completion metrics.

cs.SE↗

arXiv2026

SymLift: Neural Semantic Lifting for Compile-Free Vulnerability Detection in Context-Incomplete Code Fragments

Missing project context can prevent build-dependent analyzers from constructing program relations required for vulnerability detection, while neural predictions alone do not establish auditable derivations. We present SymLift, a compile-free framework centered on Vulnerability-Oriented Neural Semantic Lifting for context-incomplete code fragments. A fine-tuned large language model reconstructs vulnerability-relevant control-flow, data-flow, memory, and resource relations in a constrained LLVM-like intermediate representation (IR), without reconstructing the full project context or directly predicting vulnerability labels. A deterministic parser validates the IR and maps its supported operations and relations into normalized evidence facts. Explicit Datalog rules derive findings with supporting facts and derivation traces, ensuring deterministic reasoning for fixed facts and rules. We evaluate SymLift on FormAI, Juliet, and MegaVul. Reducing complete Juliet programs to function-level fragments lowers Clang compilation success from 100% to 12.00%, whereas SymLift's analysis completion rate decreases by only 1.17 percentage points. After task-specific fine-tuning, the lifter achieves a control-flow intersection over union (IoU) of 96.85% on FormAI. SymLift-Qwen-7B achieves a Target-F1 of 94.41% on FormAI and, across six vulnerability families on MegaVul, a micro-averaged precision of 98.24% and a micro-averaged Target-F1 of 91.66%. These results support accurate and auditable detection in the evaluated context-incomplete settings, even with partial program relations.

cs.SE↗

arXiv2026

Normal form method of center-focus problem in piecewise-smooth systems and algorithm design

We investigate the normal form method in piecewise-smooth monodromic systems to distinguish center from focus and determine the order of focus as well as degenerate Hopf bifurcations. For the normal form of piecewise-smooth monodromic systems given in previous publications in the sense of topologically equivalence, we prove the algebraic equivalence between the Lyapunov constants of the original piecewise-smooth system and the coefficient series of its normal form. With this algebraic equivalence, we prefer to compute the normal form coefficient series to avoid cumbersome integrals of trigonometric functions in the computation of Lyapunov constants, and find the relations of foci of the piecewise-smooth system and its subsystems: among the orders and among the stabilities. Moreover, we design algorithms for computing the normal form coefficient series and combine them with the obtained stability relation to construct stabilizing switching signals for state-dependent switched nonlinear systems, one basic problem in the control theory.

math.DS↗

arXiv2026

Krylov complexity and oscillatory quantum dynamics in two-field inflation

We investigate Krylov-space dynamics of primordial perturbations in two-field inflation. Starting from the quadratic action for coupled curvature and entropy perturbations, we solve the mode equations with Bunch-Davies initial conditions and construct the corresponding quadratic Hamiltonian. The Gaussian quantum state is described by a four-mode squeezed state with two squeezing channels. We derive the evolution of the squeezing strength and squeezing angle and use them to evaluate the Krylov complexity, Krylov entropy, and channel-resolved Lanczos coefficients. Our numerical results show that the two channels remain weakly squeezed inside the horizon but develop different squeezing histories after horizon exit. The second channel develops strong squeezing earlier and dominates the intermediate evolution, whereas the first channel shows a delayed enhancement. The corresponding Krylov complexities exhibit distinct oscillatory structures across horizon crossing and pronounced fluctuations in the strongly squeezed superhorizon regime. These features arise from the phase-sensitive squeezing dynamics and reflect the different responses of the two channels to curvature-entropy coupling.

hep-th↗

arXiv2026

EGiME: Entropy-Guided Model Evaluation in Few-Shot Learning

The feature-extraction quality of pre-trained models sets the ceiling of downstream transfer learning, yet few-shot model selection remains difficult: supervised evaluation on scarce labels is unstable and yields fluctuating rankings. We propose Entropy-Guided Model Evaluation (EGiME), a semi-supervised criterion that assesses a candidate model from the clustering structure of its feature space without any classifier training. EGiME performs $K$-means clustering over all labeled and unlabeled samples, then combines two entropy terms: an inter-cluster entropy measuring global separability, and an intra-cluster entropy measuring the compactness of labeled anchors, fused through an inverse-frequency weight that emphasizes small clusters. On five few-shot benchmarks and nine pre-trained models, EGiME achieves a correlation score of 0.966 with the full-pool ground-truth accuracy obtained by averaging the accuracies of K-means, KNN, and SVM. At labeling ratios of 4\%--20\%, it produces markedly more stable model rankings than cross-validated supervised baselines, establishing it as a practical alternative to conventional supervised evaluation for pre-trained model selection under data scarcity.

cs.LG↗

arXiv2026

A REST-based Monte Carlo Simulation and Data Analysis Framework for the N$ν$DEx Experiment

The N$ν$DEx experiment is designed to search for neutrinoless double-beta decay ($0νββ$) of $\mathrm{^{82}Se}$ and is planned to be deployed at the China Jinping Underground Laboratory (CJPL). It employs a high-pressure time projection chamber (TPC) filled with $\mathrm{^{82}SeF_6}$ gas and a Topmetal pixelated charge readout system. The high $Q_{ββ}$ value of $\mathrm{^{82}Se}$, together with the track reconstruction capability of a gaseous TPC, provides strong potential for background suppression. In this work, we develop a full simulation and reconstruction framework for the N$ν$DEx experiment based on the REST (Rare Event Searches Toolkit) software framework. To meet the detector-specific requirements of N$ν$DEx, customized simulation processes are implemented to describe the detector geometry, electron drift and diffusion, the Topmetal readout plane, signal formation, and electronics response. These developments enable a complete software description of the detector response from energy deposition to digitized waveforms. A waveform-fitting-based reconstruction method is further applied to the simulated signals to extract the deposited charge and reconstruct the event energy. For simulated signal events, the reconstructed energy resolution reaches a full width at half maximum (FWHM) of less than 1\%, satisfying the design requirement of the experiment. The resulting simulation and reconstruction chain provides a basis for future studies of detector response, background rejection, event classification, and optimization of the N$ν$DEx detector design.

hep-ex↗

arXiv2026

Imaging-anchored multiomics in cardiovascular disease: integrating cardiac imaging, bulk, single-cell, and spatial transcriptomics

Cardiovascular disease arises from interactions between inherited risk, molecular programmes, and tissue-scale remodelling that are observed clinically through imaging. Cardiac MRI (CMR), computed tomography (CT), and echocardiography are integral to routine cardiovascular care, while bulk RNA sequencing, single-cell RNA sequencing, and spatial transcriptomics are providing increasingly detailed molecular characterization of cardiac tissue. Yet, these imaging and molecular data are still analysed in largely separate pipelines. This review examines joint representations that link cardiac imaging phenotypes to transcriptomic and spatially resolved molecular states. An imaging-anchored perspective is adopted in which echocardiography, CMR, and CT define a spatial phenotype of the heart, and bulk, single-cell and spatial transcriptomics provide cell-type- and location-specific molecular context. We define the representation requirements of each modality, compare multimodal fusion strategies, and synthesize integrative pipelines for radiogenomics, spatial alignment, and image-based gene-expression prediction, together with their validation requirements, limitations, and failure modes. Spatial multiomic maps of human myocardium and atherosclerotic plaque, together with single-cell, spatial, and multimodal medical foundation models, are advancing imaging-anchored multiomics; however, cost, scalability, and tissue availability remain substantial barriers to large-scale cardiovascular translation.

q-bio.QM↗

arXiv2026

Spectral window engineering for temporal leakage of synthetic gain waveform

Arbitrary waveform generation enables precise engineering of optical fields. As an alternative, the synthetic wave approach has been proposed to generate virtual-gain waveforms for loss compensation. However, the achievable synthetic gain is severely limited by the finite spectral bandwidth available in realistic experiments. This imposes a rectangular spectral window, causing abrupt spectral truncation and inducing severe temporal leakage in the time domain synthetic response. Here, we introduce spectral window engineering to fundamentally resolve this limitation. By employing smooth-edged spectral filters, such as the Hann and Blackman windows, temporal leakage is pushed deeply into the late-time regime, yielding a significantly extended effective lifetime and reduced residual damping. Using a coupled plasmonic resonator platform, we experimentally validate that our windowed approach suppresses residual damping by more than a factor of two compared to the standard rectangular baseline. Our findings establish a practical and robust spectral filtering strategy to realize high efficiency synthetic gain, with promising prospects for nanodetection and advanced spectroscopy.

physics.optics↗

arXiv2026

SceneFactory-3D: Lifting 2D Traffic Scenes into 3D Physical Counterfactuals for Scalable Physically Grounded Safety Evaluation

Scalable driving simulators typically execute vehicle commands using prescribed behavioral or kinematic rules, overlooking the physics of tire-road interfaces, thereby limiting their ability to capture how adverse road and environmental conditions alter vehicle execution and propagate through traffic. To address this limitation, we present SceneFactory-3D, a GPU-batched, physics-grounded multi-agent driving simulator. Vehicles execute acceleration and steering commands via suspension- and friction-limited forces evaluated at each wheel-contact point. Spatially varying friction, per-world 3D heightfields, gravity, and rigid contact consistently govern wheel motion and chassis collisions. Per-world terrain isolation and GPU batching enable SceneFactory-3D to run matched physical counterfactuals in parallel: traffic scenario setup and vehicle controllers remain fixed while only the road condition changes, enabling the resulting closed-loop effects to be evaluated across parallel worlds. To demonstrate the advantage of the SceneFactory-3D-enabled counterfactual evaluation, we conduct an empirical study on vehicle controllers' sensitivity to road conditions. We study three learned-policy families on 1,024 matched 12-vehicle worlds per condition, and two classical planners on a shared 32-world subset, across 21 friction and grade conditions. When friction drops from 1.0 to 0.18, the share of vehicles that clear the work zone safely falls by 6 to 90 percentage points across learned policies (18-19 for classical planners), and near-collision situations become more frequent for every learned policy. Code: https://github.com/SmallWorldLab/SceneFactory_3D

cs.RO↗

arXiv2026

TerrainForge: Physics-Grounded Road Geometry Editing for Counterfactual Autonomous Driving

Road geometry (e.g., crests, sags, and speed humps) and surface conditions (e.g., wet or icy pavement) affect how vehicles move, what drivers and onboard cameras observe, and how much clearance remains between vehicles. Editing these properties in a driving scene therefore requires corresponding changes in vehicle motion. Capturing these differences in a driving video requires a road edit to propagate to vehicle motion, camera viewpoint, and the clearance between vehicles. We present TerrainForge, a framework for generating road geometry-focused counterfactuals from reconstructed multi-vehicle driving episodes. A unified road model connects scene deformation with four-wheel vehicle dynamics, allowing crests, sags, speed humps, and friction changes to propagate through vehicle motion, camera viewpoint, and inter-vehicle clearance. Vehicle dynamics are evaluated against CarSim, and prescribed road geometry is verified in reconstructed Waymo scenes. Across 18 episodes, leaving surrounding vehicles on their recorded trajectories instead of recomputing their responses produces median peak differences in predicted ego-lead distance of 1.52 m for crests and 1.41 m for sags. We further simulate the ego response to 15,758 road edits across 983 braking episodes, pairing each edit with its safety outcomes relative to an unedited replay. These pairs train a first-stage screening surrogate that takes the original driving context and candidate road-edit parameters as input and predicts the resulting change in the ego's terminal gap. On held-out scenes, this prediction achieves 22-40% lower mean absolute error than predicting no change, so candidates can be screened cheaply before the full multi-vehicle rollout.

cs.CV↗

arXiv2026

Graphon Particle Systems, Part I: Spatio-Temporal Approximation and Law of Large Numbers

We study a class of graphon particle systems with time-varying random coefficients. In a graphon particle system, the interactions among particles are characterized by the coupled mean field terms through an underlying graphon and the randomness of the coefficients comes from exogenous stochastic processes. By constructing two-level approximated sequences converging in 2-Wasserstein distance, we prove the existence and uniqueness of the solution to the system. Besides, by constructing two-level approximated functions converging to the graphon mean field terms, we establish the law of large numbers, which reveals that if the number of particles tends to infinity and the discretization step tends to zero, then the discrete-time interacting particle system over a large-scale network converges to the graphon particle system. As a byproduct, we discover that the graphon particle system can describe the limiting dynamics of the distributed stochastic gradient descent algorithm over the large-scale network and prove that if the gradients of the local cost functions are Lipschitz continuous, then the graphon particle system can be regarded as the spatio-temporal approximation of the discrete-time distributed stochastic gradient descent algorithm as the number of network nodes tends to infinity and the algorithm step size tends to zero.

eess.SY↗

arXiv2026

Ask the Expert: LLM-Guided Reinforcement Learning for Autonomous Cyber Defense

Policy-based reinforcement learning (RL) approaches have produced promising results for autonomous cyber defense; however, they are sample-inefficient in settings where defenders must respond under delayed, partial observations with actions from large action spaces. While large language models (LLMs) may reason semantically about security state space, high latency and trust assumptions prevent attractive in-line deployment models. We introduce Ask the Expert, a training-time guidance framework which first summarizes hard cyber-defense states, then intermittently queries an LLM for host-level defensive recommendations via a constrained action interface, and finally transforms those recommendations into tiered reward shaping for use with PPO. Because the LLM is discarded after training, deployment is a pure RL policy. Across TTCP CAGE CC1 and CC2 and both attacker types, this asymmetric design improves sample efficiency over PPO and outperforms the evaluated potential-based reward shaping (PBRS) baselines, while retaining the strongest terminal mean and requiring no LLM dependency at deployment time.

cs.CR↗

arXiv2026

A GPU-Parallel Framework for Heterogeneous Multi-Task Reinforcement Learning

GPU-parallel simulation provides abundant robot interaction, but existing benchmarks rarely combine this scale with heterogeneous manipulation tasks and standardized multi-task RL evaluation. We introduce Hebero (Heterogeneous Benchmark for Robot Learning), a GPU-parallel Isaac Lab benchmark that enables efficient joint training and evaluation of a single policy across all 40 heterogeneous tasks. Scaling experiments show that increasing parallel replicas per task improves success under a fixed wall-clock budget. To support learning with sparse rewards and limited demonstrations, we propose Demonstration-Guided Policy Optimization (DGPO), which reuses demonstrations for dense tracking rewards and asymmetric value learning. Its shared stack supports controlled comparisons of learner-specific demonstration interfaces within PPO. Within DGPO framework, we introduce IW-ABC, which uses a lightweight per-task learning progress signal to coordinate adaptive behavior cloning (ABC), relaxing demonstration guidance with task progress, and importance weighting (IW), emphasizing lagging tasks in PPO updates. With 50 demonstrations per task, IW-ABC achieves 90.1% state-input mean success, outperforming the strongest baseline FAMO-ABC by 7.8 percentage points. Its visual counterpart reaches 93.5% mean success. Real-world experiments further demonstrate that a single multi-task policy trained in simulation can successfully perform four tasks on a physical Piper robot. The project page is available at https://hebero-rl.github.io/.

cs.RO↗

arXiv2026

Navigating Route Latent Space for Synthesizable Molecular Design

Goal-directed molecular design has advanced rapidly, yet a substantial proportion of designed molecules remain difficult to synthesize in practice, limiting their real-world utility. Prior synthesizability-aware methods either project generated molecules back to synthesizable analogs that deviate from the intended target, or optimize directly in discrete synthesis spaces that lack a continuous landscape for efficient search. We argue that this limitation mainly comes from the search space rather than the optimizer. To address this, we propose RouteFlow, a framework that reformulates synthesizable molecular design as a search over a continuous route latent space, where each latent maps back to a complete synthesis route and synthesizability is inherently preserved. To navigate this space, we adopt reward-guided flow matching as an efficient sampler that steers toward high-property regions. Since reward optimization may push latents off the manifold of real synthesis routes, where decoding becomes unreliable, we further introduce a cycle-consistency mechanism to stabilize fine-tuning. Across 16 optimization tasks from Therapeutic Data Commons, RouteFlow achieves the best sample efficiency among synthesizability-aware baselines, with the best synthetic accessibility and the highest retrosynthesis success rate. Our results also confirm that the proposed cycle-consistency reliably keeps optimization on-manifold while improving target properties, supporting effective synthesizable molecular discovery.

cs.AI↗

arXiv2026

Independent Multi-Agent Reinforcement Learning with Counterfactual Semantic-Social World Models

Fully decentralized multi-agent reinforcement learning (MARL), also referred to as independent learning, requires each agent to learn and act using only its local information and experience, without a centralized critic or inter-agent communication. Such a stringent information structure renders the conventional reward signal ambiguous. A poor return may result from an ineffective ego action, an incompatible teammate response, or an effective opponent response, yet scalar rewards alone do not reveal which explanation is responsible. We argue that agents can learn more effectively by prospectively comparing the consequences of candidate actions rather than diagnosing failures only from realized returns. We introduce CASTLE (Counterfactual Action-conditioned Semantic Tokens for Local Execution in Decentralized MARL), an offline-training, online-in-context guidance framework with two complementary world models. A Local Dynamics World Model, offline pre-trained over agents' local trajectories, summarizes the agent's local trajectory dynamics and partial observability, while a Semantic-Social World Model predicts compact short-horizon task and social consequences for each candidate ego action. The latter is trained from counterfactual simulator rollouts that expose plausible teammate and opponent responses to alternative actions taken from the same logged rollout state. During online learning and execution, both world models remain frozen and are queried by agents using only locally available information. Their prediction logits provide in-context guidance to an independent PPO policy. Across 30 matched seeds on Tag, Spread, and Adversary in the benchmark multi-particle environments, our proposed CASTLE achieves the highest mean final score among the evaluated methods, exceeding the strongest baseline on each task by 10.67, 6.46, and 0.33 normalized points, respectively.

cs.MA↗

arXiv2026

Multi-Probe Zero Collision Hash (MPZCH): Mitigating Embedding Collisions and Enhancing Model Freshness in Large-Scale Recommenders

Embedding tables are critical components of large-scale recommendation systems, facilitating the efficient mapping of high-cardinality categorical features into dense vector representations. However, as the volume of unique IDs expands, traditional hash-based indexing methods suffer from collisions that degrade model performance and personalization quality. We present Multi-Probe Zero Collision Hash (MPZCH), a novel indexing mechanism based on linear probing that effectively mitigates embedding collisions. With reasonable table sizing, it often eliminates these collisions entirely while maintaining production-scale efficiency. MPZCH utilizes auxiliary tensors and high-performance CUDA kernels to implement configurable probing and active eviction policies. By retiring obsolete IDs and resetting reassigned slots, MPZCH prevents the stale embedding inheritance typical of hash-based methods, ensuring new features learn effectively from scratch. Despite its collision-mitigation overhead, the system maintains training QPS and inference latency comparable to existing methods. Rigorous online experiments demonstrate that MPZCH achieves zero collisions for user embeddings and significantly improves item embedding freshness and quality. The solution has been released within the open-source TorchRec library for the broader community.

cs.LG↗