arXiv subjects

Xu Yang

Publications and source records attributed to Xu Yang.

At least 37 records · Page 2Linked to original sources

arXiv2026

PINNForge: Execution-Grounded Evolutionary Design of Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) require coordinated choices across representation, architecture, sampling, constraints, loss construction, and optimization, yet effective configurations can vary substantially across partial differential equations (PDEs). Existing automated design methods search over these choices, but training outcomes are still used primarily to rank candidates rather than to inform subsequent designs. We develop PINNForge, an execution-grounded large language model (LLM)-driven evolutionary framework that treats observed PINN training behavior as a cross-generation design signal. PINNForge initializes diverse configurations from PDE-related prior knowledge, executes them, retains globally strong configurations, and feeds their execution evidence and accumulated run-level experience back to the LLM. The LLM then revises and recombines coupled design components or explores new combinations within a generate--execute--evaluate--evolve loop. Across 25 PDE benchmarks, PINNForge achieves the lowest mean MSE on 24 tasks among RoPINN, PINNsFormer, PINNsAgent, and PINNForge. Removing knowledge guidance, execution feedback, or evolutionary search increases the median task-wise MSE ratio to 3.74$\times$, 12.10$\times$, and 10.10$\times$, respectively, relative to full PINNForge.

cs.AI↗

arXiv2026

PULSE: Identifying Demonstration-Utility Features with Sparse Autoencoders

In-context learning is highly sensitive to demonstration choice, yet most methods select demonstrations using external query-demonstration similarity. Such criteria can miss model-specific signals: Similar demonstrations may activate different internal features and downstream behaviors. We introduce PULSE (Paired Utility Localization over Sparse Encodings), an SAE-based framework for identifying model-internal features associated with demonstration utility and using them for demonstration selection. Using a small labeled discovery set, PULSE samples candidate demonstration sets, measures their zero-shot-relative utility under the target model, and scores SAE features by how their activation differences align with utility differences. The top positive and negative coordinates form a sparse utility-localization vector. We use this vector in two complementary ways: as a signed score for controlled complete-set ranking, and as PULSE-Retriever, which converts its magnitude into a feature-relevance mask for scalable pool-scale retrieval. Across classification, generation, and reasoning benchmarks, PULSE-Retriever improves over the strongest baseline by 2-3 accuracy points, 0.6-0.9 BLEU-4, and 3.2 exact-match points, respectively, while controlled ranking validates the identified features encode a predictive set-level utility signal. Feature inspection and cross-dataset experiments suggest that the identified features capture task-relevant, dataset-conditioned patterns, yet retain utility signals that partially transfer across datasets. Our code is available at https://github.com/aohenuo/PULSE.

cs.AI↗

arXiv2026

Separating Diagnosis from Disease Representation: Dual-View EEG Learning with Neural-Dynamics-Guided Deformation

Electroencephalography (EEG)-based closed-loop neuromodulation calls for a subject-specific structured state, as opposed to a single disease probability, specifying which brain regions are deviant, at which frequencies, and at which lags. Sensor-space models keep the strongest diagnostic evidence without anatomy, source-space models give anatomy at a loss of predictive signal, and post-hoc attributions stay outside the prediction. We separate the two instead of forcing them into one representation, and propose DMD-EEG (Dual-view Multiscale Deformation for EEG), which keeps a fixed scalp spectral expert for diagnosis and models the source-space disease-related representation as a low-rank, sparse, iterative deformation of a healthy neural-dynamics prior in a $46$-region-of-interest (ROI) $\times$ $5$-frequency $\times$ $4$-lag (autocorrelation-timescale) space. The two experts meet only at a fixed decision level, so the source state is architecturally separate from the scalp expert. Across major depressive disorder (MDD), first-episode psychosis (FEP), and Parkinson's disease (PD), decision-level fusion matches the strongest single expert on MDD and FEP and exceeds the source branch on PD. On FEP the source expert is the strongest branch, the task where the deformation contributes most. The source state is an explicit ROI-frequency-lag attribution defined in a shared source coordinate system across montages, which we treat as an anatomically-coordinated predictive representation whose coordinates are directly readable and hypothesis-generating. The highest-saliency coordinates align with established disease circuitry (fronto-limbic-temporal regions in MDD, motor-cortex beta in PD), and the MDD state transfers by rank to an unseen cohort recorded with a different montage.

cs.AI↗

arXiv2026

PINNMorph: Evolving Online Adaptation Policies for Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) provide a learning-based framework for solving partial differential equations (PDEs), yet their training behavior can change substantially throughout optimization. Residual distributions, gradient interactions, regional learning difficulty, and model-capacity requirements may evolve over time, while the network architecture and major training mechanisms are typically determined before training. We propose PINNMorph, an online PINN adaptation framework based on large language model (LLM)-guided policy evolution. PINNMorph maintains a population of state-conditioned adaptation policies that map execution diagnostics to controlled interventions over topology modification, additive representation augmentation, objective balancing, gradient handling, adaptive sampling, and optimizer-phase control. At each intervention opportunity, candidate programs are instantiated from the current policy population, selected according to the observed training state, and applied directly to the PINN under training. The resulting model inherits its existing parameters and training state and continues optimization along the same trajectory. Execution outcomes are subsequently used to evaluate interventions and evolve the policy population. Unlike pre-training architecture search or fixed adaptation rules, PINNMorph jointly adapts the current PINN and the policies governing its interventions using feedback from actual training. Experiments on 13 PDE benchmarks show that PINNMorph achieves lower solution errors than SA-PINN, ConFIG, RoPINN, HARMONIC, and PINNsAgent across all evaluated problems. Ablation studies further examine the effects of online adaptation, state-conditioned intervention selection, and execution-feedback-driven policy evolution.

cs.AI↗

arXiv2026

Can Linguistic Reasoning Vectors Enhance Multimodal Reasoning Ability?

Most Vision-Language Models (VLMs) are built by extending pretrained Large Language Models (LLMs) with visual modules and multimodal alignment. However, this multimodal scaling often degrades the language-side reasoning ability originally encoded in the base LLM. While the base LLM retains usable reasoning after scaling, the aligned VLM itself cannot reliably access this ability. Therefore, recovering the degraded reasoning capability in VLMs would benefit more from seeking help from the base LLM than from the VLM alone. Motivated by this, we propose LIFT (Language-side reasonIng Facilitation and Transfer), a lightweight vector-intervention method that transfers reasoning capability from the base LLM to the VLM without retraining the backbone. LIFT defines Reasoning Vectors as answer-token hidden-state differences between a Reasoner path with an explicit reasoning trace and a Solver path without it, and injects these vectors into language-side activations of the target VLM. LIFT further supports learnable vector adaptation while keeping the VLM backbone frozen. We evaluate LIFT on two VLMs across six reasoning benchmarks, comparing Reasoning Vectors extracted from the base LLM and from the aligned VLM under matched protocols. Results show that LLM-derived vectors consistently outperform VLM-derived vectors, confirming that the base LLM is a more effective source for recovering reasoning. LIFT partially recovers degraded reasoning through lightweight language-side interventions. Further analyses show that Reasoning Vectors influence intermediate reasoning behavior rather than merely altering final answers. The source code will be released soon.

cs.AI↗

arXiv2026

Interaction induced flattening of optical transition quantum geometry

The Riemannian geometry of optical transition dipoles has become a useful picture for understanding linear and nonlinear optical response. Here we argue that the interacting quantum geometry of optical transitions possesses a rich structure and can be naturally delineated into two types: localized and delocalized particle-hole excitations. The former possess uniform quantum geometry with flat (vanishing) Hermitian curvature; the latter possess non-uniform quantum geometry with non-vanishing Hermitian curvature. As a striking example, we find that uniform quantum geometry can be produced by electron-hole interactions: even when composed from extended Bloch states in the particle and hole bands, we find excitons have uniform and flat quantum geometry. By developing a many-body length gauge formulation of nonlinear response, we find this uniform and flat excitonic quantum geometry zeros its third-order circular photoconductivity in non-magnetic materials in stark contrast to its non-interacting counterparts. Similarly, the zero Hermitian curvature of localized optical transitions locks their Hall response to that of the ground state. This demonstrates the rich landscape of many-body optical response controlled by an interacting quantum geometry.

cond-mat.mes-hall↗

arXiv2026

Strategic Inference in Stackelberg Games: Optimal Control for Revealing Adversary Intent

We study a continuous-time stochastic Stackelberg game in which a leader seeks to accomplish a primary objective while inferring a hidden parameter of a rational follower. The follower solves an entropy-regularized linear-quadratic tracking problem and responds to the leader's trajectory with a randomized policy. Anticipating this response, the leader designs informative controls to maximize the estimation efficiency for the follower's latent intent, through maximum likelihood estimation. Unlike prior work on discrete-time or finite-candidate inverse learning, our framework enables continuous parameter inference without prior assumptions and endogenizes the information source through the follower's strategic feedback. We derive semi-explicit solutions, prove well-posedness, and develop recurrent neural network algorithms to approximate the leader's path-dependent control. Numerical experiments demonstrate how the leader balances task performance and information gain, highlighting the practical value of our approach for adversarial strategic inference.

math.OC↗

arXiv2026

Behaviors near infinity for mutually enhancing continuous-state population dynamics

In this paper we consider a two-dimensional generalized continuous-state branching process with mutually enhancing two-way interactions characterized by two stochastic differential equations driven by Brownian motions and spectrally positive $α$-stable random measures. Such continuous-state population dynamics can also be identified as a stochastic Lotka-Volterra type population model. Infinite behavior such as explosion/nonexplosion for this population and staying infinite/coming down from infinity are derived under various conditions on the coefficients involved in the model and the conclusions are rather sharp

math.PR↗

arXiv2026

The Price of Covertness: Dual-Control Navigation in Uncertain Flows under Adversarial Sensing

Covert navigation in an uncertain flow couples motion planning with information acquisition. A vehicle must estimate the local flow to navigate, while estimation error induces corrective maneuvers that increase statistical distinguishability, or leakage, and hence the vehicle's detectability. We model this coupling by augmenting position with the estimation-error variance, whose evolution depends on the route through the local information rate. A small-error expansion shows that estimation uncertainty contributes a leading-order correction to the expected detectability rate. The resulting route-planning problem is characterized by a first-order Hamilton--Jacobi--Bellman (HJB) equation. The deadline-indexed value determines the covert-time frontier, while an exact sensitivity formula with respect to sensing quality shows that information acquired earlier along the route can reduce leakage incurred later, whereas information acquired later cannot reduce leakage already incurred. We then formulate a zero-sum sensing-allocation game, show that the leakage rate is convex in the sensing allocation, and prove that the defender may restrict attention to randomization over extreme single-site allocations. The resulting equilibrium is computed by column generation, with vehicle best responses obtained from the HJB equation. Numerical experiments illustrate the learning detour, the deadline--leakage tradeoff, the spatial-ordering effect, and the benefit of randomized sensing allocations.

math.OC↗

arXiv2026

Real-Validated UAV Audition Under Rotor Ego-Noise for Low-False-Alarm Human Detection

Detecting human acoustic cues from UAV-mounted microphones could support acoustic search and rescue, but rotor ego-noise often masks speech, cries, coughs, and other human sounds at extremely low SNRs. We study UAV human-audible-presence detection under this real operating constraint. Models are trained on a reproducible synthetic mixture pipeline built from public audio, but selected and evaluated on real DroneAudioSet recordings using a metadata-defined audibility filter, recording-grouped Dev/Test splits, and group-bootstrap confidence intervals. Our results show that synthetic accuracy is a weak and non-monotonic proxy for real UAV transfer: a from-scratch SE-ResNet appears competitive on synthetic mixtures but collapses on real ego-noise, while frozen audio foundation models and lightweight adapters transfer more reliably. We further evaluate a BEATs adapter family with rotor-aware conditioning and domain regularization. The Real-Dev-selected EgoRAP-DA configuration achieves the best locked-test low-false-alarm recall among the candidates, but its advantage over a vanilla adapter is not statistically significant under paired group bootstrap. The main contribution is therefore a real-validated benchmark and evaluation protocol showing that honest progress in UAV audition requires real, group-level validation rather than synthetic scores alone.

cs.CE↗

arXiv2026

CIPL: A Channel-Aware Framework for Recoverable Privacy Leakage in LLM Agents

Privacy leakage in LLM agents is commonly evaluated within individual components such as memory, retrieval, or tool-use pipelines, which makes it difficult to distinguish internal exposure from information that an external observer can actually recover. We present CIPL (Channel Inversion for Privacy Leakage), a channel-aware evaluation framework for black-box privacy leakage in LLM agents. CIPL represents a target through sensitive source, selection, assembly, execution, observation, and extraction stages and evaluates the transition from selected sensitive units to attacker-recoverable output under a shared protocol. Experiments across memory-based, retrieval-mediated, and tool-mediated targets, together with a BrowserUse live-agent case study, show that storage labels alone do not determine recoverability. Memory targets form a near-saturated reference case, retrieval-mediated leakage is frequently partial, and tool-mediated and live-agent leakage varies strongly with observation surface, prompt-to-channel alignment, retrieval depth, and provider behavior. A stratified semantic audit further identifies attacker-useful disclosures that canonical exact matching misses. CIPL therefore provides a common framework for comparing how internal sensitive dependence is realized as externally recoverable leakage across heterogeneous agent pipelines.

cs.CR↗

arXiv2026

Before the Arrest: Benchmarking LLMs on Criminal Profiling from Incomplete Evidence

Large Language Models (LLMs) are increasingly applied to legal and criminal justice tasks, yet existing work focuses almost exclusively on post-arrest scenarios where the suspect's identity is already known, leaving the critical pre-arrest challenge of inferring suspect characteristics from incomplete evidence largely unexplored. To fill this gap, we introduce the Profiling, Investigation, and Judgment (PIJ), comprising 2,500 real homicide cases from five countries. PIJ evaluates LLMs across three tasks that span the entire criminal investigation pipeline: criminal profiling, which requires abductive reasoning to infer suspect attributes from fragmentary scene evidence, crime process reconstruction, which tests structured information extraction, and sentence prediction, which demands legal deductive reasoning. We evaluate 9 powerful LLMs and find that performance degrades systematically as tasks shift from explicit fact extraction to implicit reasoning over unknown suspect profiles. Categories requiring inferential reasoning, such as motivation and victim-offender relationships, remain the primary bottlenecks. Further analysis reveals substantial gaps between LLMs and human experts, along with pervasive biases in gender, age, and motive attribution. Our findings indicate that pre-arrest inference from incomplete evidence remains an open challenge.

cs.CL↗

arXiv2026

Sharp Rainbow Path Covers in Dense and Complete Multipartite Graphs

A path in a properly edge-colored graph is rainbow if its edges have pairwise distinct colors. For a proper edge-coloring $c$ of a graph $G$, let $\operatorname{rpc}(G,c)$ be the minimum number of rainbow paths needed to cover $E(G)$, and let $\operatorname{rpc}(G)$ be the maximum of $\operatorname{rpc}(G,c)$ over all proper edge-colorings of $G$. We prove that, for every fixed $0<α<1$, every properly edge-colored $n$-vertex graph with minimum degree at least $αn$ satisfies $\operatorname{rpc}(G,c)\leq(1+o(1))n/2$, where the coefficient $1/2$ is best possible. We also determine $\operatorname{rpc}(G)$ asymptotically for every complete multipartite graph. If $G=K_{n_1,\ldots,n_r}$ has order $n$ and largest and smallest part sizes $M$ and $s$, respectively, then, uniformly over all choices of the number and sizes of the parts, $\operatorname{rpc}(G)=(1+o(1))\max\{\min\{\lfloor n/2\rfloor,n-M\},(n-s)/2\}$. The proof combines pseudorandom packings of globally rainbow linear forests with a decomposition into dense parts and prescribed avoidance for arbitrary dense graphs, and with reserved connectors and a direct dominant-part argument for complete multipartite graphs.

math.CO↗

arXiv2026

Extinction behaviour for mutually enhancing continuous-state population dynamics

In this paper, we study a two-dimensional process arising as the unique nonnegative solution to a system of two stochastic differential equations (SDEs) with mutually enhancing two-way interactions driven by independent Brownian motions and spectrally positive $α$-stable random measures. Such a SDE system can be identified as a continuous-state Lotka-Volterra type population model. Extinction properties of the populations are studied for different choices of the coefficients involved in the SDEs.

math.PR↗

arXiv2026

Strongly separating graph edges with $10n$ paths

A family of paths strongly separates the edges of a graph if every two distinct edges are separated in both directions by paths in the family. Bonamy, Botler, Dross, Naia, and Skokan proved that every $n$-vertex graph admits such a family of at most $19n$ paths. We improve this bound to $10n-o(n)$.

math.CO↗

arXiv2026

SimCRAFT: Distilling Remote Sensing Agents via Synthetic Trajectories and Contextual Retrieval-Augmented Fine-Tuning

The unprecedented surge in Earth observation data volume and diversity has exposed a critical bottleneck for traditional manual workflows, catalyzing the emergence of Remote Sensing (RS) Agents. However, the practical deployment of these advanced agents is severely hindered by their heavy reliance on large-scale general-purpose LLMs, which lack deep domain expertise and impose prohibitive infrastructure demands. To resolve this, we propose SimCRAFT, a model-agnostic framework that distills sophisticated RS orchestration capabilities into a compact 7B-scale model. Addressing data scarcity, we first pair a multiagent synthesis engine with a Mock Execution Engine that checks schema correctness, inter-tool dependencies, and sensor/tool compatibility, producing SimRS-14k, a large-scale, constraint-validated workflow planning corpus. Second, we propose Contextual Retrieval-Augmented Fine-Tuning (CRAFT) that finetunes the model to reason analogically by adapting retrieved Standard Operating Procedures to novel queries under a noise-robust objective, generalizing RAFT to multi-step RS workflow planning without mechanical copying. Extensive experiments demonstrate that SimCRAFT-7B significantly outperforms openweights LLMs and rivals advanced closedsource models and specialized RS agents, while reproducing across three 7B backbones. This work contributes a competitive open-weights baseline for lightweight RS intelligence, enabling efficient autonomous deployment under resource-constrained or resource-conserving conditions.

cs.AI↗

arXiv2026

Information Revelation and Alignment Faking in Stochastic Differential Games

In competitive games with private objectives, actions can reveal information about hidden parameters. However, quantifying such information revelation is nontrivial, since it depends not only on the opponent's hidden parameter but also on the opponent's model of the game. We study this problem via a two-player linear-quadratic stochastic differential game under partial information, in which each player knows its own coupling parameter and models the opponent's hidden parameter through a prior. Starting from the full-information game, we characterize the Nash equilibrium by coupled Riccati equations. We then define baseline implementable controls by averaging the equilibrium under each player's prior. Building on this baseline, we formulate an alignment faking control problem in which one player trades off fidelity to its implementable policy against information acquisition about the opponent's hidden parameter. The information incentive is constructed from a proxy Fisher information matrix based only on the player's available model. This leads to a tractable saddle-point formulation with semi-explicit control characterization through Riccati systems. Numerical illustrations show that alignment faking can substantially improve information gain over baseline play when the faker's model is accurate, but often at the cost of greater detectability. They also show that the proxy Fisher information can systematically differ from the true information under model misspecification.

math.OC↗

arXiv2026

Well-posedness of the relaxed Electron MHD equations with random diffusion

We investigate a three-dimensional active-vector relaxation of the electron magnetohydrodynamics (EMHD) equations without resistivity. The relaxation replaces the electron current by a generalized current defined through a fractional power of the Laplacian. We drive the system by multiplicative pseudo-differential noise and apply an exponential transformation that removes the stochastic differential term and produces an effective fractional damping. For deterministic divergence-free, mean-zero initial data, we prove local well-posedness up to a stopping time in suitable Gevrey spaces. Under a smallness condition, we further establish global well-posedness with high explicit probability.

math.AP↗