arXiv subjects

Zheng Luo

Publications and source records attributed to Zheng Luo.

At least 19 recordsLinked to original sources

arXiv2026

TaxDistill: Improving Metagenomic Taxonomic Annotation via Distilled Genomic Foundation Models

Metagenomic taxonomic annotation aims to identify the microbial origins of DNA fragments in environmental samples. Traditional methods that rely on sequence similarity are often constrained by the high microbial diversity and the incompleteness of reference databases, which has motivated the development of learning approaches such as Taxometer that perform post hoc correction to learn more informative metagenomic sequence representations. However, these methods typically rely on labels derived from similarity search tools during training, which inevitably introduces noise that can impair representation learning and degrade classification performance. To address this issue, we propose TaxDistill, a knowledge distillation framework for metagenomic classification. We introduce GenomeOcean, a 500M parameter genomic foundation model, as the teacher network to extract deep semantic features and generate soft labels based on confidence. By distilling this soft label information into a lightweight student network, TaxDistill effectively reduces the label noise introduced by initial retrieval tools. Comprehensive experiments on seven diverse CAMI2 datasets demonstrate that TaxDistill outperforms existing baselines in most scenarios. For instance, on the Gastrointestinal dataset, it improves the F1 score of MMseqs2 from 0.763 to 0.941, outperforming the Taxometer baseline. Overall, TaxDistill provides a reliable method for label correction in complex metagenomic analysis.

cs.LG↗

arXiv2026

When Simulation Lies: A Sim-to-Real Benchmark and Domain-Randomized RL Recipe for Tool-Use Agents

Tool-use language agents are evaluated on benchmarks that assume clean inputs, unambiguous tool registries, and reliable APIs. Real deployments violate all these assumptions: user typos propagate into hallucinated tool names, a misconfigured request timeout can stall an agent indefinitely, and duplicate tool names across servers can freeze an SDK. We study these failures as a sim-to-real gap in the tool-use partially observable Markov decision process (POMDP), where deployment noise enters through the observation, action space, reward-relevant metadata, or transition dynamics. We introduce RobustBench-TC, a benchmark with 22 perturbation types organized by these four POMDP components, each grounded in a verified GitHub issue or documented tool-calling failure. Across 21 models from 1.5B to 32B parameters (including the closed-source o4-mini), the robustness profile is sharply uneven: observation perturbations reduce accuracy by less than 5%, while reward-relevant and transition perturbations reduce accuracy by roughly 40% and 30%, respectively; scale alone does not close these gaps. We then propose ToolRL-DR, a domain-randomization reinforcement learning (RL) recipe that trains a tool-use agent on perturbation-augmented trajectories spanning the three statically encodable POMDP components. On a 3B backbone, ToolRL-DR-Full retains roughly three-quarters of clean accuracy and reaches an aggregate perturbed accuracy comparable to open-source 14B function-calling baselines while substantially narrowing the gap to o4-mini. It closes approximately 27% of the Transition gap despite never seeing transition perturbations in training, suggesting that RL on adversarial static tool-use inputs induces a more persistent retry policy that transfers to unseen runtime failures. The dataset, code and benchmark leaderboard are publicly available.

cs.AI↗

arXiv2026

A scalable infrastructure for strontium optical clocks with integrated photonics

Optical atomic clocks provide exceptionally accurate and precise signals for timekeeping and precision measurements, but they require high-power, free-space laser configurations that limit scalability. We introduce and explore a scalable infrastructure for strontium (Sr) optical-lattice clocks that incorporates co-design of atomic-beam slowing and a magneto-optical trap (MOT) from an effusion source, generation of complex, three-dimensional free-space laser configurations with a photonic integrated circuit (PIC) and metasurface (MS) optics, and laser stabilization to a frequency-comb supercontinuum generated with integrated nonlinear photonics. With these elements, we realize MOTs of all stable strontium isotopes ($^{84}$Sr, $^{86}$Sr, $^{87}$Sr, $^{88}$Sr) with populations commensurate with natural abundances, demonstrating precise beam control and robustness. Access to laser-cooled alkaline-earth atoms with scalable integrated photonics enables system engineering for optical clocks, quantum sensing, and quantum information, and our experiments demonstrate extensible technologies that advance toward a Sr optical clock largely free of bulk optics.

physics.app-ph↗

arXiv2026

Data-knowledge dual-driven intelligent framework for full-chain, experiment-efficient synthesis of 2D dendrites

Exemplified by the chemical vapor deposition growth of two-dimensional dendrites, which has potential applications in catalysis and presents a parameter-intensive, data-scarce and reaction process-complex model problem, we devise a machine intelligence-empowered framework for the full chain support of material synthesis, encompassing rapid process optimization, accurate customized synthesis, and comprehensive mechanism deciphering.First, active learning is integrated into the experimental workflow, identifying an optimal recipe for the growth of highly-branched, electrocatalytically-active ReSe2 dendrites through 60 experiments (4 iterations), which account for less than 1.3% of the numerous possible parameter combinations.Then, a prediction accuracy-guided data augmentation strategy is developed combined with a tree-based machine learning (ML) algorithm, unveiling a non-linear correlation between 5 process variables and fractal dimension (DF) of ReSe2 dendrites with only 9 experiment additions, which guides the synthesis of various user-defined DF. Finally, we construct a data-knowledge dual-driven mechanism model by integration of cross-scale characterizations, interpretable ML models, and domain knowledge in thermodynamics and kinetics, unraveling synergistic contributions of multiple process parameters to the product morphology. This work demonstrates the ML potential to transform the research paradigm and is adaptable to broader material synthesis.

cond-mat.mtrl-sci↗

arXiv2026

HIFICL: High-Fidelity In-Context Learning for Multimodal Tasks

In-Context Learning (ICL) is a significant paradigm for Large Multimodal Models (LMMs), using a few in-context demonstrations (ICDs) for new task adaptation. However, its performance is sensitive to demonstration configurations and computationally expensive. Mathematically, the influence of these demonstrations can be decomposed into a dynamic mixture of the standard attention output and the context values. Current approximation methods simplify this process by learning a "shift vector". Inspired by the exact decomposition, we introduce High-Fidelity In-Context Learning (HIFICL) to more faithfully model the ICL mechanism. HIFICL consists of three key components: 1) a set of "virtual key-value pairs" to act as a learnable context, 2) a low-rank factorization for stable and regularized training, and 3) a simple end-to-end training objective. From another perspective, this mechanism constitutes a form of context-aware Parameter-Efficient Fine-Tuning (PEFT). Extensive experiments show that HiFICL consistently outperforms existing approximation methods on several multimodal benchmarks. The code is available at https://github.com/bbbandari/HiFICL.

cs.CV↗

arXiv2026

Fairness or Fluency? An Investigation into Language Bias of Pairwise LLM-as-a-Judge

Recent advances in Large Language Models (LLMs) have incentivized the development of LLM-as-a-judge, an application of LLMs where they are used as judges to decide the quality of a certain piece of text given a certain context. However, previous studies have demonstrated that LLM-as-a-judge can be biased towards different aspects of the judged texts, which often do not align with human preference. One of the identified biases is language bias, which indicates that the decision of LLM-as-a-judge can differ based on the language of the judged texts. In this paper, we study two types of language bias in pairwise LLM-as-a-judge: (1) performance disparity between languages when the judge is prompted to compare options from the same language, and (2) bias towards options written in major languages when the judge is prompted to compare options of two different languages. We find that for same-language judging, there exist significant performance disparities across language families, with European languages consistently outperforming African languages, and this bias is more pronounced in culturally-related subjects. For inter-language judging, we observe that most models favor English answers, and that this preference is influenced more by answer language than question language. Finally, we investigate whether language bias is in fact caused by low-perplexity bias, a previously identified bias of LLM-as-a-judge, and we find that while perplexity is slightly correlated with language bias, language bias cannot be fully explained by perplexity only.

cs.CL↗

arXiv2026

Data-driven active learning approaches for accelerating materials discovery

Materials discovery is a cornerstone of modern technological advancement, yet it remains constrained by traditional trial-and-error paradigms and the inherent bias of human intuition. Artificial intelligence (AI) has emerged as a transformative tool in materials science by effectively modeling structure-property relationships. Despite substantial efforts to enhance model expressiveness, data efficiency remains an equally critical challenge, given the limited availability of experimental and computational resources. Active learning (AL), as a data-driven machine learning paradigm, has shown great promise for discovering novel materials and enabling the efficient navigation of vast materials spaces. In this review, we follow the evolution of sampling strategy design techniques in AL, from Bayesian optimization to advanced deep learning-based strategies. We then highlight how AL enhances data efficiency across various data regimes, ranging from task-specific settings with limited data to the development of general-purpose datasets and large-scale models. We further provide a systematic overview of AL applications throughout the materials research pipeline, including computational simulation, composition and structural design, process optimization, and self-driving laboratory systems. Finally, we pinpoint key challenges and future perspectives of AL in materials discovery.

cond-mat.mtrl-sci↗

arXiv2026

Lost in Execution: On the Multilingual Robustness of Tool Calling in Large Language Models

Large Language Models (LLMs) are increasingly deployed as agents that invoke external tools through structured function calls. While recent work reports strong tool-calling performance under standard English-centric evaluations, the robustness of tool calling under multilingual user interactions remains underexplored. In this work, we introduce MLCL, a diagnostic benchmark, and conduct a systematic evaluation of multilingual tool calling across Chinese, Hindi, and the low-resource language Igbo. Through fine-grained error analysis, we show that many failures occur despite correct intent understanding and tool selection. We identify parameter value language mismatch as a dominant failure mode, where models generate semantically appropriate parameter values in the user's language, violating language-invariant execution conventions. We further evaluate several inference-time system strategies and find that while these strategies substantially reduce language-induced execution errors, none of them can fully recover English-level performance.

cs.CL↗

arXiv2025

ADORE: Autonomous Domain-Oriented Relevance Engine for E-commerce

Relevance modeling in e-commerce search remains challenged by semantic gaps in term-matching methods (e.g., BM25) and neural models' reliance on the scarcity of domain-specific hard samples. We propose ADORE, a self-sustaining framework that synergizes three innovations: (1) A Rule-aware Relevance Discrimination module, where a Chain-of-Thought LLM generates intent-aligned training data, refined via Kahneman-Tversky Optimization (KTO) to align with user behavior; (2) An Error-type-aware Data Synthesis module that auto-generates adversarial examples to harden robustness; and (3) A Key-attribute-enhanced Knowledge Distillation module that injects domain-specific attribute hierarchies into a deployable student model. ADORE automates annotation, adversarial generation, and distillation, overcoming data scarcity while enhancing reasoning. Large-scale experiments and online A/B testing verify the effectiveness of ADORE. The framework establishes a new paradigm for resource-efficient, cognitively aligned relevance modeling in industrial applications.

cs.CL↗

arXiv2025

Algorithms for Optimizing Acyclic Queries

Most research on query optimization has centered on binary join algorithms like hash join and sort-merge join. However, recent years have seen growing interest in theoretically optimal algorithms, notably Yannakakis' algorithm. These algorithms rely on join trees, which differ from the operator trees for binary joins and require new optimization techniques. We propose three approaches to constructing join trees for acyclic queries. First, we give an algorithm to enumerate all join trees of an alpha-acyclic query by edits with amortized constant delay, which forms the basis of a cost-based optimizer for acyclic joins. Second, we show that the Maximum Cardinality Search algorithm by Tarjan and Yannakakis constructs a unique shallowest join tree, rooted at any relation, for a Berge-acyclic query; this tree enables parallel execution of large join queries. Finally, we prove that any connected left-deep linear plan for a gamma-acyclic query can be converted into a join tree by a simple algorithm, allowing reuse of optimization infrastructure developed for binary joins.

cs.DB↗

arXiv2025

Multi-objective Aligned Bidword Generation Model for E-commerce Search Advertising

Retrieval systems primarily address the challenge of matching user queries with the most relevant advertisements, playing a crucial role in e-commerce search advertising. The diversity of user needs and expressions often produces massive long-tail queries that cannot be matched with merchant bidwords or product titles, which results in some advertisements not being recalled, ultimately harming user experience and search efficiency. Existing query rewriting research focuses on various methods such as query log mining, query-bidword vector matching, or generation-based rewriting. However, these methods often fail to simultaneously optimize the relevance and authenticity of the user's original query and rewrite and maximize the revenue potential of recalled ads. In this paper, we propose a Multi-objective aligned Bidword Generation Model (MoBGM), which is composed of a discriminator, generator, and preference alignment module, to address these challenges. To simultaneously improve the relevance and authenticity of the query and rewrite and maximize the platform revenue, we design a discriminator to optimize these key objectives. Using the feedback signal of the discriminator, we train a multi-objective aligned bidword generator that aims to maximize the combined effect of the three objectives. Extensive offline and online experiments show that our proposed algorithm significantly outperforms the state of the art. After deployment, the algorithm has created huge commercial value for the platform, further verifying its feasibility and robustness.

cs.CL↗

arXiv2024

Exploring structure diversity in atomic resolution microscopy with graph neural networks

The emergence of deep learning (DL) has provided great opportunities for the high-throughput analysis of atomic-resolution micrographs. However, the DL models trained by image patches in fixed size generally lack efficiency and flexibility when processing micrographs containing diversified atomic configurations. Herein, inspired by the similarity between the atomic structures and graphs, we describe a few-shot learning framework based on an equivariant graph neural network (EGNN) to analyze a library of atomic structures (e.g., vacancies, phases, grain boundaries, doping, etc.), showing significantly promoted robustness and three orders of magnitude reduced computing parameters compared to the image-driven DL models, which is especially evident for those aggregated vacancy lines with flexible lattice distortion. Besides, the intuitiveness of graphs enables quantitative and straightforward extraction of the atomic-scale structural features in batches, thus statistically unveiling the self-assembly dynamics of vacancy lines under electron beam irradiation. A versatile model toolkit is established by integrating EGNN sub-models for single structure recognition to process images involving varied configurations in the form of a task chain, leading to the discovery of novel doping configurations with superior electrocatalytic properties for hydrogen evolution reactions. This work provides a powerful tool to explore structure diversity in a fast, accurate, and intelligent manner.

cond-mat.mtrl-sci↗

arXiv2024

Auto-resolving the atomic structure at van der Waals interfaces using a generative model

The high-resolution visualization of atomic structures is significant for understanding the relationship between the microscopic configurations and macroscopic properties of materials. However, a rapid, accurate, and robust approach to automatically resolve complex patterns in atomic-resolution microscopy remains difficult to implement. Here, we present a Trident strategy-enhanced disentangled representation learning method (a generative model), which utilizes a few unlabelled experimental images with abundant low-cost simulated images to generate a large corpus of annotated simulation data that closely resembles experimental results, producing a high-quality large-volume training dataset. A structural inference model is then trained via a residual neural network which can directly deduce the interlayer slip and rotation of diversified and complicated stacking patterns at van der Waals interfaces with picometer-scale accuracy across various materials (e.g. MoS2, WS2, ReS2, ReSe2, and 1T'-MoTe2) with different layer numbers , demonstrating robustness to defects, imaging quality, and surface contaminations. The framework can also identify pattern transition interfaces, quantify subtle motif variations, and discriminate moire patterns that are difficult to distinguish in frequency domains. Finally, the high-throughput processing ability of our method provides insights into a van der Waals epitaxy mode where various thermodynamically favorable slip stackings can coexist.

cond-mat.mtrl-sci↗

arXiv2024

Tractable Conjunctive Queries over Static and Dynamic Relations

We investigate the evaluation of conjunctive queries over static and dynamic relations. While static relations are given as input and do not change, dynamic relations are subject to inserts and deletes. We characterise syntactically three classes of queries that admit constant update time and constant enumeration delay. We call such queries tractable. Depending on the class, the preprocessing time is linear, polynomial, or exponential (under data complexity, so the query size is constant). To decide whether a query is tractable, it does not suffice to analyse separately the sub-queries over the static relations and over the dynamic relations, respectively. Instead, we need to take the interaction between the static and the dynamic relations into account. Even when the sub-query over the dynamic relations is not tractable, the overall query can become tractable if the dynamic relations are sufficiently constrained by the static ones.

cs.DB↗

arXiv2024

Laser cooling $^{88}$Sr to microkelvin temperature with an integrated-photonics system

We report on experiments generating a magneto-optical trap (MOT) of 88-strontium ($^{88}$Sr) atoms at microkelvin temperature, using integrated-photonics devices. With metasurface optics integrated on a fused-silica substrate, we generate six-beam, circularly polarized, counter-propagating MOTs on the blue broad-line, 461 nm, and red narrow-line, 689 nm, Sr cooling transitions without bulk optics. By use of a diverging beam configuration, we create up to 10 mm diameter MOT beams at the trapping location. To frequency stabilize and linewidth narrow the cooling lasers, we use fiber-packaged, integrated nonlinear waveguides to spectrally broaden a frequency comb. The ultra-coherent supercontinuum of the waveguides covers 650 nm to 2500 nm, enabling phase locks of the cooling lasers to hertz level linewidth. Our work highlights the possibility to simplify the preparation of an ultracold 88Sr gas for an optical-lattice clock with photonic devices. By implementing a timing sequence for control of the MOT lasers and the quadrupole magnetic-field gradient, we collect atoms directly from a thermal beam into the blue MOT and continuously cool into a red MOT with dynamic detuning and intensity control. There, the red MOT temperature is as low as $2~μ$K and the overall transfer efficiency up to 16%. We characterize this sequence, including an intermediate red MOT with modulated detuning. Our experiments demonstrate an integrated photonics system capable of cooling alkaline-earth gases to microkelvin temperature with sufficient transfer efficiencies for adoption in scalable optical clocks and quantum sensors.

physics.atom-ph↗

arXiv2024

Three-dimensional, multi-wavelength beam formation with integrated metasurface optics for Sr laser cooling

We demonstrate the formation of a complex, multi-wavelength, three-dimensional laser beam configuration with integrated metasurface optics. Our experiments support the development of a compact Sr optical-lattice clock, which leverages magneto-optical trapping on atomic transitions at 461 nm and 689 nm without bulk free-space optics. We integrate six, mm-scale metasurface optics on a fused-silica substrate and illuminate them with light from optical fibers. The metasurface optics provide full control of beam pointing, divergence, and polarization to create the laser configuration for a magneto-optical trap. We report the efficiency and integration of the three-dimensional visible laser beam configuration, demonstrating the suitability of metasurface optics for atomic laser cooling.

physics.optics↗

arXiv2022

A Bott periodicity theorem for $\ell^p$-spaces and the coarse Novikov conjecture at infinity

We formulate and prove a Bott periodicity theorem for an $\ell^p$-space ($1\leq p<\infty$). For a proper metric space $X$ with bounded geometry, we introduce a version of $K$-homology at infinity, denoted by $K_*^{\infty}(X)$, and the Roe algebra at infinity, denoted by $C^*_{\infty}(X)$. Then the coarse assembly map descents to a map from $\lim_{d\to\infty}K_*^{\infty}(P_d(X))$ to $K_*(C^*_{\infty}(X))$, called the coarse assembly map at infinity. We show that to prove the coarse Novikov conjecture, it suffices to prove the coarse assembly map at infinity is an injection. As a result, we show that the coarse Novikov conjecture holds for any metric space with bounded geometry which admits a fibred coarse embedding into an $\ell^p$-space. These include all box spaces of a residually finite hyperbolic group and a large class of warped cones of a compact space with an action by a hyperbolic group.

math.KT↗

arXiv2021

K-theory of the maximal and reduced Roe algebras of metric spaces with A-by-CE coarse fibrations

Let $X$ be a discrete metric space with bounded geometry. We show that if $X$ admits an "A-by-CE coarse fibration", then the canonical quotient map $λ: C^*_{\max}(X)\to C^*(X)$ from the maximal Roe algebra to the Roe algebra of $X$, and the canonical quotient map $λ: C^*_{u, \max}(X)\to C^*_u(X)$ from the maximal uniform Roe algebra to the uniform Roe algebra of $X$, induce isomorphisms on $K$-theory. A typical example of such a space arises from a sequence of group extensions $\{1\to N_n\to G_n\to Q_n\to 1\}$ such that the sequence $\{N_n\}$ has Yu's property A, and the sequence $\{Q_n\}$ admits a coarse embedding into Hilbert space. This extends an early result of J. Špakula and R. Willett \cite{JR2013} to the case of metric spaces which may not admit a coarse embedding into Hilbert space. Moreover, it implies that the maximal coarse Baum-Connes conjecture holds for a large class of metric spaces which may not admit a fibred coarse embedding into Hilbert space.

math.OA↗