Searcharxiv⌕ Search

arXiv subjects

Rui Zhu

Publications and source records attributed to Rui Zhu.

At least 37 records · Page 2Linked to original sources
arXiv2026

Can Agents Generalize to the Open World? Unveiling the Fragility of Static Training in Tool Use

While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.

cs.AI↗
arXiv2026

Optical library of Ga2O3 polymorphs

Gallium oxide is an emerging material of interest due to its unique combination of functional properties and the existence of multiple polymorphs - α, β, γ, δ, and κ - each exhibiting distinct characteristics arising from their different lattice symmetries. Optical properties are particularly important, as they determine potential device applications and enable phase identification. However, direct comparison of optical signatures, including key parameters such as bandgaps, is hindered by inconsistent, sparse, or even missing data in the literature. To address this issue, in the present work we systematically cross-correlate optical emission and absorption features of α, β, γ, δ, and κ thin films, as well as differently oriented β-phase bulk crystals and γ/β double polymorph structures. We demonstrate that optical bandgaps and emission features scale consistently across the polymorphs when methodological uncertainties are minimized by applying identical experimental conditions and unified analysis procedures to a structurally similar set of thin film samples. In addition, we extend conventional far field optical phase identification to the nanoscale by reporting near field optical signatures of Ga2O3 polymorphs via nano FTIR. Overall, the present dataset provides a comprehensive reference of near- and far-field optical polymorph signatures to support ongoing multidisciplinary research on Ga2O3.

cond-mat.mtrl-sci↗
arXiv2026

A Physics-Informed Hierarchical Neural Network for Microwave Scattering Analysis of 3D PEC Targets

Accurate modeling of scattering from three-dimensional (3D) perfectly electrically conducting (PEC) targets at microwave frequencies constitutes a fundamental objective in computational electromagnetics, particularly for radar cross section (RCS) prediction and microwave scattering analysis. Classical solvers, such as the method of moments and the Multilevel Fast Multipole Algorithm (MLFMA), although provide high physical fidelity, they become costly under scenarios of repeated queries involving many incidence configurations or frequencies, whereas purely data-driven surrogates often lack accuracy on geometrically complex targets. This paper proposes a U-shaped physics-informed artificial neural network (U-PINet) for 3D microwave scattering analysis. Inspired by the near-far field decomposition of MLFMA, U-PINet combines a near-field graph encoder, parameterized by learnable univariate basis functions, with a hierarchical multi-scale fusion module organized on an octree partition. The proposed network is trained against a discretized residual of the electric-field integral equation at surface collocation points, without requiring reference current labels. Experiments on canonical and geometrically complex 3D PEC targets, conducted under multiple frequency and polarization configurations and assessed through bistatic RCS reconstruction, showcase that U-PINet outperforms representative physics-informed baselines, and yields substantial runtime savings over the classical MLFMA solver under repeated-query scenarios.

cs.LG↗
arXiv2026

Size Scaling Law for Radiation Losses of Modes in Photonic Crystal Surface Emitting Devices

Photonic-crystal surface-emitting lasers (PCSELs) have garnered significant attention due to their ability to generate laser beams with ultra high power and low divergence. This is because they support high power single mode lasing with volumes orders of magnitude larger than those of conventional semiconductor lasers. The finite lateral size in a PCSEL is a primary factor limiting its lasing mode volume and consequently, its output power. We demonstrate that the scaling relation between the total cavity loss $α=α_\perp + α_\parallel$ and the device size $L$ is such that the surface radiation loss scales as $α_{\perp} \sim O(L^{-2})$, while the edge radiation loss $α_{\parallel} \sim O(L^{-3})$. Both scaling relations can be explained by the second order expansions of the complex frequency $ω(k)$ of the band diagram. Our results constitute an explicit guideline for PCSEL designs to optimize various optical properties.

physics.optics↗
arXiv2026

Bi-CoG: Bi-Consistency-Guided Self-Training for Vision-Language Models

Exploiting unlabeled data through semi-supervised learning (SSL) or leveraging pre-trained models via fine-tuning are two prevailing paradigms for addressing label-scarce scenarios. Recently, growing attention has been given to combining fine-tuning of pre-trained vision-language models (VLMs) with SSL, forming the emerging paradigm of semi-supervised fine-tuning. However, existing methods often suffer from model bias and hyperparameter sensitivity, due to reliance on prediction consistency or pre-defined confidence thresholds. To address these limitations, we propose a simple yet effective plug-and-play methodology named $\underline{\textbf{Bi-Co}}$nsistency-$\underline{\textbf{G}}$uided Self-Training (Bi-CoG), which assigns high-quality and low-bias pseudo-labels, by simultaneously exploiting inter-model and intra-model consistency, along with an error-aware dynamic pseudo-label assignment strategy. Both theoretical analysis and extensive experiments over 14 datasets demonstrate the effectiveness of Bi-CoG, which consistently and significantly improves the performance of existing methods.

cs.LG↗
arXiv2026

Continuous Latent Diffusion Language Model

Large language models have achieved remarkable success under the autoregressive paradigm, yet high-quality text generation need not be tied to a fixed left-to-right order. Existing alternatives still struggle to jointly achieve generation efficiency, scalable representation learning, and effective global semantic modeling. We propose Cola DLM, a hierarchical latent diffusion language model that frames text generation through hierarchical information decomposition. Cola DLM first learns a stable text-to-latent mapping with a Text VAE, then models a global semantic prior in continuous latent space with a block-causal DiT, and finally generates text through conditional decoding. From a unified Markov-path perspective, its diffusion process performs latent prior transport rather than token-level observation recovery, thereby separating global semantic organization from local textual realization. This design yields a more flexible non-autoregressive inductive bias, supports semantic compression and prior fitting in continuous space, and naturally extends to other continuous modalities. Through experiments spanning 4 research questions, 8 benchmarks, strictly matched ~2B-parameter autoregressive and LLaDA baselines, and scaling curves up to about 2000 EFLOPs, we identify an effective overall configuration of Cola DLM and verify its strong scaling behavior for text generation. Taken together, the results establish hierarchical continuous latent prior modeling as a principled alternative to strictly token-level language modeling, where generation quality and scaling behavior may better reflect model capability than likelihood, while also suggesting a concrete path toward unified modeling across discrete text and continuous modalities.

cs.CL↗
arXiv2026

Can LLMs Predict Polymer Physics Just by Reading Synthesis and Processing Prose?

Can large language models predict physical and mechanical polymer properties simply by reading unstructured scientific prose? Polymer performance is rarely determined by chemical structure alone; identical nominal polymers can exhibit drastically different behaviors depending on their synthesis route, processing history, morphology, and testing conditions. Yet, state-of-the-art polymer property models typically rely on structure-only representations -- such as SMILES or molecular graphs -- which strip away this vital experimental context. In this work, we introduce \textbf{PolyLM}, a natural-language-only, process- and condition-aware framework that predicts materials performance directly from full-text literature. By circumventing structural inputs entirely, PolyLM preserves the nuanced, unstructured descriptions of synthesis and processing reported by domain scientists. To train this framework, we curated an unprecedented, literature-scale dataset encompassing 185,000 scientific papers and over 276,400 unique polymer samples across 22 physical, mechanical, and thermal properties. We fine-tuned a massive 9-billion-parameter language model (Qwen3.5-9B) using Low-Rank Adaptation (LoRA) and task-level uncertainty weighting. Evaluated on 68,283 held-out observations, the model achieves remarkably high predictive accuracy, establishing new state-of-the-art benchmarks for complex properties. Across the 22 diverse targets, the model achieves a median $R^2$ of 0.74, with predictions for key thermal, mechanical, and physicochemical properties frequently surpassing an $R^2$ of 0.80. These results unequivocally demonstrate that natural language is a powerful, highly scalable interface for realistic materials performance prediction.

cs.LG↗
arXiv2026

Unveiling the Atomistic Mechanisms of Shear-Induced LDA$\leftrightarrow$HDA Transformations and Shear Banding in Amorphous Silicon under High Pressures

Large-scale molecular dynamics simulations of shear deformation under constant pressures of amorphous silicon, PT from low-density-amorphous (LDA) to high-density-amorphous (HDA) Si, and formation of shear bands (SBs) are performed using the state-of-the-art Gaussian Approximation Potential. The simulations reveal that LDA$\leftrightarrow$HDA shear-induced PTs occur simultaneously until reaching steady state. The developed mechanism-based analytical model well describes shear-strain-governed kinetics and steady states at all pressures, independent of shear stresses. Shear reduces the pressure for initiation and completion of LDA$\rightarrow$HDA PT by $4.36$ and $5.10$ GPa, respectively. Without PT at low pressure, shear-banding occurs, which is partially suppressed by PT at higher pressure with uniform deformation-PT at $9.8$ GPa. Despite the much larger shear and expected fraction of HDA, surprising sharp drop in the HDA atomic fraction within the SB was discovered. In bulk, Si deforms by atomic rearrangement in localized shear transformation zones with high nonaffine displacements, which trigger nucleation of HDA clusters within LDA and, concurrently, of LDA clusters within HDA, without growth and coalescence. In SB, a turbulent-like flow with swirls is revealed, which promotes reverse PT from HDA$\rightarrow$LDA more effectively. Transformation-induced plasticity in amorphous Si is revealed. The findings open up basic research into the mechanisms and kinetics of plastic strain-induced PTs in amorphous materials under high pressure, with numerous important applications.

cond-mat.mtrl-sci↗
arXiv2026

Hey, That's My Data! Token-Only Dataset Inference in Large Language Models

Large Language Models (LLMs) rely on massive training datasets, often including proprietary data, which raises concerns about unauthorized usage and copyright infringement. Existing dataset inference methods typically require access to log probabilities or other internal signals, but many modern LLMs restrict such access, motivating token-only inference approaches. We propose CatShift, a token-only dataset inference framework based on catastrophic forgetting, where models overwrite prior knowledge when trained on new data. Fine-tuning an LLM on a subset of its training data induces larger output shifts than fine-tuning on unseen data. CatShift compares these shifts against those from a known non-member validation set to infer whether a dataset was included in training. Experiments on both open-source and API-based LLMs show that CatShift remains effective without logit access, enabling practical protection of proprietary datasets.

cs.CL↗
arXiv2026

Geographic Bias and Diversity in AI Evaluation

Among the many challenges hindering the responsible development and deployment of AI, arguably none has faced more intense scrutiny than bias in its various forms. This underscores the widespread concerns across AI researchers that model outputs, e.g., from generative AI, may encode structural distributional imbalances (stemming from training data or model design) that may amplify social inequality or introduce systemic distortions across application domains ranging from biodiversity to disaster mitigation. Yet, relatively little work has investigated the geographical nature of bias or developed measurable benchmarks for what it means for (generative) AI to be unbiased. In this chapter, we investigate this issue through a literature review. As foundation models are reshaping the landscape of bias research, we examine work spanning both the pre-generative AI and generative AI periods. First, we identify a range of geographic biases. These biases span from representation bias in the training data and regional disparities in the factual recall of language models to the tendency of generative AI to over-proportionally favor prototypical places (called defaults). Then, we showcase how recent studies address the latter bias by evaluating geographic diversity in the outputs of generative AI across various cognitive levels, parameter settings, and output modalities.

cs.CY↗
arXiv2026

The Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey: Quasar Properties from Data Release 10 to 12

We present the quasar catalog from Data Releases 10 to 12 of the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) Quasar Survey, comprising quasars observed between September 2021 and June 2024. We robustly identified $11,346$ quasars, of which $5,386$ are newly discovered objects not present in the Million Quasars catalog. This release brings the total number of quasars identified by the 12-year LAMOST survey to $67,521$, of which $29,513$ are newly discovered. While the absolute flux calibration for LAMOST quasar spectra from Data Releases 6 to 9 was previously performed using the SDSS/PanSTARRS1 multi-band photometric data, the inherent variability of quasars can affect the flux accuracy. To address this limitation, we recalibrated the LAMOST spectra using (quasi-)simultaneous photometric data from Zwicky Transient Facility (ZTF), which has conducted high-cadence sky monitoring since March 2018. Based on the recalibrated single-epoch spectra, we estimated the emission line fluxes, continuum fluxes, and virial black hole masses. These improved spectra facilitate direct comparison with the spectra of common quasars from the Sloan Digital Sky Survey (SDSS), enabling searches for rare quasars, such as changing-look quasars exhibiting the appearance or disappearance of broad emission lines and broad absorption line quasars. The combined dataset of photometry and multi-epoch spectra will enhance the detections of AGN-related transients, such as Bowen fluorescence flares and extreme variability quasars, thereby improving our understanding of quasar variability.

astro-ph.GA↗
arXiv2026

On buoyancy in disperse two-phase flow and its impact on well-posedness of two-fluid models

The Maxey-Riley-Gatignol equation for the flow around a sphere at low particle Reynolds number tells us that the fluid-particle interaction force decomposes into a contribution from the local flow disturbance caused by the particle's boundary -- consisting of the drag, virtual-mass, and history forces, and their Faxén corrections -- and another contribution from the stress of the background flow, termed generalized-buoyancy force. There is also a consensus that, for general disperse two-phase flow, the interfacial force density, coupling the average fluid phase and dispersed-phase momentum balances, decomposes in a likewise manner. However, there has been a long-standing controversy about the physical closure separating the generalized-buoyancy from the interfacial force density, especially whether or not pseudo-stresses, such as the Reynolds stress, should be attributed to the background flow. Furthermore, most existing propositions for this closure involve small-particle approximations. Here, we show that all existing buoyancy closures are inconsistent with particle-resolving numerical simulations and/or at least one of two simple thought experiments designed to determine the roles of pseudo-stresses and small-particle approximations. We then derive the unique consistent closure. It requires no approximation and implies that all stresses and pseudo-stresses in the average fluid phase momentum balance, except the Reynolds stress, fully contribute to the background flow responsible for buoyancy. Remarkably, it exhibits a low-pass filter property, attenuating buoyancy at short wavelengths, that prevents it from causing Hadamard instabilities, constituting a first-principle-based solution to the long-standing ill-posedness problem of two-fluid models. When employing the derived closure, even simplistic two-fluid models are linearly well-posed.

physics.flu-dyn↗
arXiv2026

ResPrune: Text-Conditioned Subspace Reconstruction for Visual Token Pruning in Large Vision-Language Models

Large Vision-Language Models (LVLMs) rely on dense visual tokens to capture fine-grained visual information, but processing all these tokens incurs substantial computational and memory overhead during inference. To address this issue, we propose ResPrune, a training-free visual token pruning framework that enables efficient LVLM inference by selecting a compact yet informative subset of visual tokens. ResPrune formulates visual token pruning as a subspace reconstruction problem and employs a greedy subspace expansion strategy guided by residual energy, allowing it to preserve the geometric structure of the original visual token space. To further incorporate cross modal alignment, the selection process is conditioned on textual relevance, encouraging the retention of tokens that are both informative and instruction-relevant. The proposed method is lightweight and model-agnostic, and can be seamlessly integrated into existing LVLM pipelines without retraining or architectural modifications. Extensive experiments on multiple LVLM backbones, including LLaVA-1.5, LLaVA-NeXT, and Qwen2.5-VL, demonstrate that ResPrune consistently outperforms existing pruning approaches across a wide range of benchmarks, while achieving effective reductions in computation, memory consumption, and inference latency.

cs.LG↗
arXiv2026

Geography According to ChatGPT -- How Generative AI Represents and Reasons about Geography

Understanding how AI will represent and reason about geography should be a key concern for all of us, as the broader public increasingly interacts with spaces and places through these systems. Similarly, in line with the nature of foundation models, our own research often relies on pre-trained models. Hence, understanding what world AI systems construct is as important as evaluating their accuracy, including factual recall. To motivate the need for such studies, we provide three illustrative vignettes, i.e., exploratory probes, in the hope that they will spark lively discussions and follow-up work: (1) Do models form strong defaults, and how brittle are model outputs to minute syntactic variations? (2) Can distributional shifts resurface from the composition of individually benign tasks, e.g., when using AI systems to create personas? (3) Do we overlook deeper questions of understanding when solely focusing on the ability of systems to recall facts such as geographic principles?

cs.AI↗
arXiv2026

Giant intrinsic dichroism in \b{eta}-Ga2O3 enables filter-free, high-fidelity polarization division multiplexing

Conventional polarization detection relies on external filters, which incur significant efficiency loss and polarization crosstalk, especially in the deep ultraviolet band where subwavelength nanofabrication is challenging. Here, we report that monoclinic \b{eta}-Ga2O3 exhibits intrinsic giant polarization dichroism, allowing near-ideal polarization photodetection without external optical elements and coherent polarization-division multiplexing (PDM) capability. The giant dichroism originates from the crystallographic symmetry-driven selectivity of optical transitions, which, combined with a large valence band splitting, results in vastly distinct absorption for orthogonal polarizations. A theoretical analysis of the transition selection rules in \b{eta}-Ga2O3 reveals only the E//c-polarized vb1-to-conduction band transition is activated, within the 245-258 nm spectral window. An admirable polarization ratio surpassing 500 and a polarization crosstalk ratio below 0.2% is hence achieved. The polarization-sensitive photodetector exhibits a high responsivity of 73 A/W and fast response (20 ms). Furthermore, we showcase its practical utility in PDM free-space communication, successfully decoding encoded optical signals, and demonstrate its capability for high-fidelity Stokes vector retrieval. The intrinsic anisotropy of \b{eta}-Ga2O3, dictated by its crystal symmetry, lays the groundwork for filter-free, high-fidelity polarization polarimetry. This work further paves the way for a general design principle in next-generation optoelectronics that harness polarization transition selection rules.

physics.app-ph↗
arXiv2026

On the biogenic hydrodynamic transport of upward and downward cruising copepods

Mesozooplankton aggregations undergoing vertical migrations in the upper ocean have been hypothesized to have an important role in the redistribution of carbon, nutrients, and oxygen via biogenic hydrodynamic transport (BHT). While laboratory studies have demonstrated how swarm-induced hydrodynamic instabilities can drive large-scale transport in strongly stratified environments, measurements are usually performed with model organisms that differ in morphology and swimming mode from ecologically relevant marine species. To bridge this gap, we conducted experiments with copepods and analyzed upward and downward trajectories to identify differences in flow fields, force distribution, and BHT for these two scenarios. Using two-dimensional bright-field Particle Image Velocimetry (PIV), we quantified the near-body velocity field and found that the average downward swimming speed significantly exceeds the average upward swimming speed, with the flow fields exhibiting direction-dependent characteristics. We incorporated these findings into a continuum squirmer model to estimate the swimmer-induced drift volume and mixing efficiency, focusing on the effects of the reduced gravity of the swimmers and the density stratification of the surrounding fluid. Our simulations reveal that both the excess weight of the organisms and the fluid stratification strongly constrain the net BHT. This study provides a critical step toward integrating lab-based models of marine mesozooplankton with remote sensing data to incorporate vertical migrations into global ocean models with realistic biogeochemistry and assess their ecological significance in actively sustaining local ecosystems.

physics.flu-dyn↗
arXiv2026

Trojans in Artificial Intelligence (TrojAI) Final Report

The Intelligence Advanced Research Projects Activity (IARPA) launched the TrojAI program to confront an emerging vulnerability in modern artificial intelligence: the threat of AI Trojans. These AI trojans are malicious, hidden backdoors intentionally embedded within an AI model that can cause a system to fail in unexpected ways, or allow a malicious actor to hijack the AI model at will. This multi-year initiative helped to map out the complex nature of the threat, pioneered foundational detection methods, and identified unsolved challenges that require ongoing attention by the burgeoning AI security field. This report synthesizes the program's key findings, including methodologies for detection through weight analysis and trigger inversion, as well as approaches for mitigating Trojan risks in deployed models. Comprehensive test and evaluation results highlight detector performance, sensitivity, and the prevalence of "natural" Trojans. The report concludes with lessons learned and recommendations for advancing AI security research.

cs.CR↗
arXiv2026

Exploring Multiple High-Scoring Subspaces in Generative Flow Networks

As a probabilistic sampling framework, Generative Flow Networks (GFlowNets) show strong potential for constructing complex combinatorial objects through the sequential composition of elementary components. However, existing GFlowNets often suffer from excessive exploration over vast state spaces, leading to over-sampling of low-reward regions and convergence to suboptimal distributions. Effectively biasing GFlowNets toward high-reward solutions remains a non-trivial challenge. In this paper, we propose CMAB-GFN, which integrates a combinatorial multi-armed bandit (CMAB) framework with GFlowNet policies. The CMAB component prunes low-quality actions, yielding compact high-scoring subspaces for exploration. Restricting GFNs to these compact high-scoring subspaces accelerates the discovery of high-value candidates, while the exploration of different subspaces ensures that diversity is not sacrificed. Experimental results on multiple tasks demonstrate that CMAB-GFN generates higher-reward candidates than existing approaches.

cs.LG↗