arXiv subjects

Lei Chen

Publications and source records attributed to Lei Chen.

At least 199 records · Page 11Linked to original sources

arXiv2025

Dichotomy of flat bands in the van der Waals ferromagnet Fe$_5$GeTe$_2$

Quantum materials with bands of narrow bandwidth near the Fermi level represent a promising platform for exploring a diverse range of fascinating physical phenomena, as the high density of states within the small energy window often enables the emergence of many-body physics. On one hand, flat bands can arise from strong Coulomb interactions that localize atomic orbitals. On the other hand, quantum destructive interference can quench the electronic kinetic energy. Although both have a narrow bandwidth, the two types of flat bands should exhibit very distinct spectral properties arising from their distinctive origins. So far, the two types of flat bands have only been realized in very different material settings and chemical environments, preventing a direct comparison. Here, we report the observation of the two types of flat bands within the same material system--an above-room-temperature van der Waals ferromagnet, Fe$_{5-x}$GeTe$_2$, distinguishable by a switchable iron site order. The contrasting nature of the flat bands is also identified by the remarkably distinctive temperature-evolution of the spectral features, indicating that one arises from electron correlations in the Fe(1) site-disordered phase, while the other geometrical frustration in the Fe(1) site-ordered phase. Our results therefore provide a direct juxtaposition of the distinct formation mechanism of flat bands in quantum materials, and an avenue for understanding the distinctive roles flat bands play in the presence of magnetism, topology, and lattice geometrical frustration, utilizing sublattice ordering as a key control parameter.

cond-mat.str-el↗

arXiv2025

Understanding the Embedding Models on Hyper-relational Knowledge Graph

Recently, Hyper-relational Knowledge Graphs (HKGs) have been proposed as an extension of traditional Knowledge Graphs (KGs) to better represent real-world facts with additional qualifiers. As a result, researchers have attempted to adapt classical Knowledge Graph Embedding (KGE) models for HKGs by designing extra qualifier processing modules. However, it remains unclear whether the superior performance of Hyper-relational KGE (HKGE) models arises from their base KGE model or the specially designed extension module. Hence, in this paper, we data-wise convert HKGs to KG format using three decomposition methods and then evaluate the performance of several classical KGE models on HKGs. Our results show that some KGE models achieve performance comparable to that of HKGE models. Upon further analysis, we find that the decomposition methods alter the original HKG topology and fail to fully preserve HKG information. Moreover, we observe that current HKGE models are either insufficient in capturing the graph's long-range dependency or struggle to integrate main-triple and qualifier information due to the information compression issue. To further justify our findings and offer a potential direction for future HKGE research, we propose the FormerGNN framework. This framework employs a qualifier integrator to preserve the original HKG topology, and a GNN-based graph encoder to capture the graph's long-range dependencies, followed by an improved approach for integrating main-triple and qualifier information to mitigate compression issues. Our experimental results demonstrate that FormerGNN outperforms existing HKGE models.

cs.LG↗

arXiv2025

DocTron-Formula: Generalized Formula Recognition in Complex and Structured Scenarios

Optical Character Recognition (OCR) for mathematical formula is essential for the intelligent analysis of scientific literature. However, both task-specific and general vision-language models often struggle to handle the structural diversity, complexity, and real-world variability inherent in mathematical content. In this work, we present DocTron-Formula, a unified framework built upon general vision-language models, thereby eliminating the need for specialized architectures. Furthermore, we introduce CSFormula, a large-scale and challenging dataset that encompasses multidisciplinary and structurally complex formulas at the line, paragraph, and page levels. Through straightforward supervised fine-tuning, our approach achieves state-of-the-art performance across a variety of styles, scientific domains, and complex layouts. Experimental results demonstrate that our method not only surpasses specialized models in terms of accuracy and robustness, but also establishes a new paradigm for the automated understanding of complex scientific documents.

cs.CV↗

arXiv2025

A Survey on Large Language Model Acceleration based on KV Cache Management

Large Language Models (LLMs) have revolutionized a wide range of domains such as natural language processing, computer vision, and multi-modal tasks due to their ability to comprehend context and perform logical reasoning. However, the computational and memory demands of LLMs, particularly during inference, pose significant challenges when scaling them to real-world, long-context, and real-time applications. Key-Value (KV) cache management has emerged as a critical optimization technique for accelerating LLM inference by reducing redundant computations and improving memory utilization. This survey provides a comprehensive overview of KV cache management strategies for LLM acceleration, categorizing them into token-level, model-level, and system-level optimizations. Token-level strategies include KV cache selection, budget allocation, merging, quantization, and low-rank decomposition, while model-level optimizations focus on architectural innovations and attention mechanisms to enhance KV reuse. System-level approaches address memory management, scheduling, and hardware-aware designs to improve efficiency across diverse computing environments. Additionally, the survey provides an overview of both text and multimodal datasets and benchmarks used to evaluate these strategies. By presenting detailed taxonomies and comparative analyses, this work aims to offer useful insights for researchers and practitioners to support the development of efficient and scalable KV cache management techniques, contributing to the practical deployment of LLMs in real-world applications. The curated paper list for KV cache management is in: \href{https://github.com/TreeAI-Lab/Awesome-KV-Cache-Management}{https://github.com/TreeAI-Lab/Awesome-KV-Cache-Management}.

cs.AI↗

arXiv2025

Multi-Prompt Style Interpolation for Fine-Grained Artistic Control

Text-driven image style transfer has seen remarkable progress with methods leveraging cross-modal embeddings for fast, high-quality stylization. However, most existing pipelines assume a \emph{single} textual style prompt, limiting the range of artistic control and expressiveness. In this paper, we propose a novel \emph{multi-prompt style interpolation} framework that extends the recently introduced \textbf{StyleMamba} approach. Our method supports blending or interpolating among multiple textual prompts (eg, ``cubism,'' ``impressionism,'' and ``cartoon''), allowing the creation of nuanced or hybrid artistic styles within a \emph{single} image. We introduce a \textit{Multi-Prompt Embedding Mixer} combined with \textit{Adaptive Blending Weights} to enable fine-grained control over the spatial and semantic influence of each style. Further, we propose a \emph{Hierarchical Masked Directional Loss} to refine region-specific style consistency. Experiments and user studies confirm our approach outperforms single-prompt baselines and naive linear combinations of styles, achieving superior style fidelity, text-image alignment, and artistic flexibility, all while maintaining the computational efficiency offered by the state-space formulation.

cs.GR↗

arXiv2025

GTA: Grouped-head latenT Attention

Attention mechanisms underpin the success of large language models (LLMs), yet their substantial computational and memory overhead poses challenges for optimizing efficiency and performance. A critical bottleneck arises as KV cache and attention computations scale rapidly with text length, challenging deployment on hardware with limited computational and memory resources. We observe that attention mechanisms exhibit substantial redundancy, since the KV cache can be significantly compressed and attention maps across heads display high similarity, revealing that much of the computation and storage is unnecessary. Leveraging these insights, we propose \textbf{G}rouped-Head Laten\textbf{T} \textbf{A}ttention (GTA), a novel attention mechanism that reduces memory usage and computational complexity while maintaining performance. GTA comprises two components: (1) a shared attention map mechanism that reuses attention scores across multiple heads, decreasing the key cache size; and (2) a nonlinear value decoder with learned projections that compresses the value cache into a latent space, further cutting memory needs. GTA cuts attention computation FLOPs by up to \emph{62.5\%} versus Grouped-Query Attention and shrink the KV cache by up to \emph{70\%}, all while avoiding the extra overhead of Multi-Head Latent Attention to improve LLM deployment efficiency. Consequently, GTA models achieve a \emph{2x} increase in end-to-end inference speed, with prefill benefiting from reduced computational cost and decoding benefiting from the smaller cache footprint.

cs.CL↗

arXiv2025

Towards Temporal Knowledge Graph Alignment in the Wild

Temporal Knowledge Graph Alignment (TKGA) seeks to identify equivalent entities across heterogeneous temporal knowledge graphs (TKGs) for fusion to improve their completeness. Although some approaches have been proposed to tackle this task, most assume unified temporal element standards and simplified temporal structures across different TKGs. They cannot deal with TKGA in the wild (TKGA-Wild), where multi-scale temporal element entanglement and cross-source temporal structural imbalances are common. To bridge this gap, we study the task of TKGA-Wild and propose HyDRA, a new and effective solution. HyDRA is the first to reformulate the task via multi-scale hypergraph retrieval-augmented generation to address the challenges of TKGA-Wild.In addition, we design a new scale-weave synergy mechanism for HyDRA, which incorporates intra-scale interactions and cross-scale conflict detection. This mechanism is designed to alleviate the fragmentation caused by multi-source temporal incompleteness and resolves inconsistencies arising from complex and uneven temporal event density distributions, thereby enhancing the model capacity to handle the intricacies of real-world temporal alignment. Finally, there is no standard benchmark that captures these challenges of TKGA-Wild and effectively evaluates existing methods. To this end, we formally propose to benchmark challenges for TKGA-Wild and validate the effectiveness of the method by establishing two new datasets(BETA and WildBETA). Extensive experiments on the new datasets and six representative benchmarks show that BETA and WildBETA better reflect real-world challenges. Meanwhile, HyDRA proposes a new paradigm for TKGA-Wild, consistently outperforming 24 competitive baselines, while maintaining strong efficiency and scalability.

cs.DB↗

arXiv2025

When Speed meets Accuracy: an Efficient and Effective Graph Model for Temporal Link Prediction

Temporal link prediction in dynamic graphs is a critical task with applications in diverse domains such as social networks, recommendation systems, and e-commerce platforms. While existing Temporal Graph Neural Networks (T-GNNs) have achieved notable success by leveraging complex architectures to model temporal and structural dependencies, they often suffer from scalability and efficiency challenges due to high computational overhead. In this paper, we propose EAGLE, a lightweight framework that integrates short-term temporal recency and long-term global structural patterns. EAGLE consists of a time-aware module that aggregates information from a node's most recent neighbors to reflect its immediate preferences, and a structure-aware module that leverages temporal personalized PageRank to capture the influence of globally important nodes. To balance these attributes, EAGLE employs an adaptive weighting mechanism to dynamically adjust their contributions based on data characteristics. Also, EAGLE eliminates the need for complex multi-hop message passing or memory-intensive mechanisms, enabling significant improvements in efficiency. Extensive experiments on seven real-world temporal graphs demonstrate that EAGLE consistently achieves superior performance against state-of-the-art T-GNNs in both effectiveness and efficiency, delivering more than a 50x speedup over effective transformer-based T-GNNs.

cs.AI↗

arXiv2025

ContextCache: Context-Aware Semantic Cache for Multi-Turn Queries in Large Language Models

Semantic caching significantly reduces computational costs and improves efficiency by storing and reusing large language model (LLM) responses. However, existing systems rely primarily on matching individual queries, lacking awareness of multi-turn dialogue contexts, which leads to incorrect cache hits when similar queries appear in different conversational settings. This demonstration introduces ContextCache, a context-aware semantic caching system for multi-turn dialogues. ContextCache employs a two-stage retrieval architecture that first executes vector-based retrieval on the current query to identify potential matches and then integrates current and historical dialogue representations through self-attention mechanisms for precise contextual matching. Evaluation of real-world conversations shows that ContextCache improves precision and recall compared to existing methods. Additionally, cached responses exhibit approximately 10 times lower latency than direct LLM invocation, enabling significant computational cost reductions for LLM conversational applications.

cs.CL↗

arXiv2025

Tactical Decision for Multi-UGV Confrontation with a Vision-Language Model-Based Commander

In multiple unmanned ground vehicle confrontations, autonomously evolving multi-agent tactical decisions from situational awareness remain a significant challenge. Traditional handcraft rule-based methods become vulnerable in the complicated and transient battlefield environment, and current reinforcement learning methods mainly focus on action manipulation instead of strategic decisions due to lack of interpretability. Here, we propose a vision-language model-based commander to address the issue of intelligent perception-to-decision reasoning in autonomous confrontations. Our method integrates a vision language model for scene understanding and a lightweight large language model for strategic reasoning, achieving unified perception and decision within a shared semantic space, with strong adaptability and interpretability. Unlike rule-based search and reinforcement learning methods, the combination of the two modules establishes a full-chain process, reflecting the cognitive process of human commanders. Simulation and ablation experiments validate that the proposed approach achieves a win rate of over 80% compared with baseline models.

cs.AI↗

arXiv2025

Effective Self-Attention-Based Deep Learning Model with Evolutionary Grid Search for Robust Wave Farm Energy Forecasting

Achieving carbon neutrality, a key focus of UN SDG #13, drives the exploration of wave energy, a renewable resource with the potential to generate 30,000 TWh of clean electricity annually, surpassing global demand. However, wave energy remains underdeveloped due to technical and economic challenges, particularly in forecasting wave farm power output, which is vital for grid stability and commercial viability. This study proposes a novel predictive framework to enhance wave energy integration into power grids. It introduces a hybrid sequential learning model combining Self-Attention-enhanced Convolutional Bi-LSTM with hyperparameter optimization. The model leverages spatial data from Wave Energy Converters (WECs) and is validated using datasets from wave farms in Adelaide, Sydney, Perth, and Tasmania, Australia. Benchmarked against ten machine learning algorithms, the model achieves superior accuracy, with R2 scores of 91.7% (Adelaide), 88.0% (Perth), 82.8% (Tasmania), and 91.0% (Sydney). It outperforms conventional ML and deep learning methods, offering robust and scalable predictions for wave energy output across diverse marine environments, supporting reliable integration into energy systems.

cs.NE↗

arXiv2025

Pushing the Limits of Safety: A Technical Report on the ATLAS Challenge 2025

Multimodal Large Language Models (MLLMs) have enabled transformative advancements across diverse applications but remain susceptible to safety threats, especially jailbreak attacks that induce harmful outputs. To systematically evaluate and improve their safety, we organized the Adversarial Testing & Large-model Alignment Safety Grand Challenge (ATLAS) 2025}. This technical report presents findings from the competition, which involved 86 teams testing MLLM vulnerabilities via adversarial image-text attacks in two phases: white-box and black-box evaluations. The competition results highlight ongoing challenges in securing MLLMs and provide valuable guidance for developing stronger defense mechanisms. The challenge establishes new benchmarks for MLLM safety evaluation and lays groundwork for advancing safer multimodal AI systems. The code and data for this challenge are openly available at https://github.com/NY1024/ATLAS_Challenge_2025.

cs.CR↗

arXiv2025

Interactive Text-to-SQL via Expected Information Gain for Disambiguation

Relational databases are foundational to numerous domains, including business intelligence, scientific research, and enterprise systems. However, accessing and analyzing structured data often requires proficiency in SQL, which is a skill that many end users lack. With the development of Natural Language Processing (NLP) technology, the Text-to-SQL systems attempt to bridge this gap by translating natural language questions into executable SQL queries via an automated algorithm. Yet, when operating on complex real-world databases, the Text-to-SQL systems often suffer from ambiguity due to natural ambiguity in natural language queries. These ambiguities pose a significant challenge for existing Text-to-SQL translation systems, which tend to commit early to a potentially incorrect interpretation. To address this, we propose an interactive Text-to-SQL framework that models SQL generation as a probabilistic reasoning process over multiple candidate queries. Rather than producing a single deterministic output, our system maintains a distribution over possible SQL outputs and seeks to resolve uncertainty through user interaction. At each interaction step, the system selects a branching decision and formulates a clarification question aimed at disambiguating that aspect of the query. Crucially, we adopt a principled decision criterion based on Expected Information Gain to identify the clarification that will, in expectation, most reduce the uncertainty in the SQL distribution.

cs.DB↗

arXiv2025

Topology-ferrimagnetism intertwining via weak interactions in Lieb lattices

A common wisdom about quantum many-body systems is that emergent phases typically fall into either the Landau-Ginzburg paradigm or topological classifications. Experimentally realizing the intertwined emergence of spontaneous symmetry breaking and topological order remains challenging. Here, we present an experimentally accessible platform for studying magnetic topological states in a spin-orbit-coupled Lieb lattice. Remarkably, we observe the coexistence of topological characteristics, quantified by the Chern number and Bott index, with spontaneous symmetry-breaking orders, such as ferrimagnetism, in the many-body ground states. Computational analyses combining dynamical mean-field theory and Hartree-Fock approximations reveal a pronounced parameter regime where magnetic topological insulators emerge even under weak interactions. This unconventional phenomenon originates from the Lieb lattice's unique band structure, which facilitates the synergy between interaction-driven symmetry breaking and spin-orbit coupling induced band inversion. Crucially, spin polarization and spin winding co-emerge as inherently coupled phenomena due to their shared origin in the same interacting, spinful atoms. We further propose a specific experimental implementation scheme for ultracold atoms, utilizing currently available Raman lattice techniques. Our findings pave the way for exploring the interplay between symmetry-broken states and topological order in strongly correlated systems.

cond-mat.quant-gas↗

arXiv2025

Learning Counterfactually Decoupled Attention for Open-World Model Attribution

In this paper, we propose a Counterfactually Decoupled Attention Learning (CDAL) method for open-world model attribution. Existing methods rely on handcrafted design of region partitioning or feature space, which could be confounded by the spurious statistical correlations and struggle with novel attacks in open-world scenarios. To address this, CDAL explicitly models the causal relationships between the attentional visual traces and source model attribution, and counterfactually decouples the discriminative model-specific artifacts from confounding source biases for comparison. In this way, the resulting causal effect provides a quantification on the quality of learned attention maps, thus encouraging the network to capture essential generation patterns that generalize to unseen source models by maximizing the effect. Extensive experiments on existing open-world model attribution benchmarks show that with minimal computational overhead, our method consistently improves state-of-the-art models by large margins, particularly for unseen novel attacks. Source code: https://github.com/yzheng97/CDAL.

cs.CV↗

arXiv2025

Piecewise Linear Approximation in Learned Index Structures: Theoretical and Empirical Analysis

A growing trend in the database and system communities is to augment conventional index structures, such as B+-trees, with machine learning (ML) models. Among these, error-bounded Piecewise Linear Approximation ($ε$-PLA) has emerged as a popular choice due to its simplicity and effectiveness. Despite its central role in many learned indexes, the design and analysis of $ε$-PLA fitting algorithms remain underexplored. In this paper, we revisit $ε$-PLA from both theoretical and empirical perspectives, with a focus on its application in learned index structures. We first establish a fundamentally improved lower bound of $Ω(κ\cdot ε^2)$ on the expected segment coverage for existing $ε$-PLA fitting algorithms, where $κ$ is a data-dependent constant. We then present a comprehensive benchmark of state-of-the-art $ε$-PLA algorithms when used in different learned data structures. Our results highlight key trade-offs among model accuracy, model size, and query performance, providing actionable guidelines for the principled design of future learned data structures.

cs.DB↗

arXiv2025

Nonlinear Enhancement of Measurement Precision via a Hybrid Quantum Switch

Quantum metrology promises measurement precision beyond the classical limit by using suitably tailored quantum states and detection strategies. However, scaling up this advantage is experimentally challenging, due to the difficulty of generating high-quality large-scale probes. Here, we build a photonic setup that achieves enhanced precision scaling by manipulating the probe's dynamics through operations performed in a coherently controlled order. Our setup applies an unknown rotation and a known orbital angular momentum increase in a coherently controlled order, in a way that reproduces a hybrid quantum SWITCH involving gates generated by both discrete and continuous variables. The unknown rotation angle $θ$ is measured with precision scaling as $1/4ml$ when a photon undergoes a rotation of $2mθ$ and an angular momentum shift of $2l \hbar$. With a practical enhancement factor as high as 2317, the ultimate precision in our experiment is $0.0105^{\prime \prime}$ when using $7.16\times10^7$ photons, corresponding to a normalized precision of $\approx 10^{-4}$rad per photon. No photon interaction occurs in our experiment, and the precision enhancement consumes only a linearly increasing amount of physical resources while achieving a nonlinear scaling of the precision. We further indicate that this nonlinear enhancement roots in an in-depth exploration of the Heisenberg uncertainty principle (HUP), and our findings not only deepen the understanding of the HUP but also pave a pathway for advancements in quantum metrology.

quant-ph↗

arXiv2025

Topological Kondo semimetals emulated in hetero-bilayer transition metal dichalcogenides

The moiré structure of AB-stacked $\rm{MoTe_2/WSe_2}$ represents a natural platform to realize Kondo lattice models, due to the discrepancy of the bandwidth between the individual layers. Here, we study this system at the commensurate filling of $ν_{\rm tot}=2$. Our focus is on the $1+1$ filling setting of $ν_{\rm Mo} =ν_{\rm W}=1$, which enables a Kondo lattice description. We find a Kondo semimetal due to the sizable intra-orbital hopping among the electrons in the $\rm{MoTe_2}$ layer. The Kondo-driven (emergent) flat band is naturally pinned to the Fermi energy. When combined with the inherent topology of the electronic structure, a topological Kondo semimetal phase ensues. We calculate the valley Hall response and, due to the breaking of inversion symmetry, also identify a spontaneous Hall effect. There is a Berry curvature dodecapole, which leads to a fourth-order spontaneous Hall effect in the perturbative regime of the electric field that is further amplified in the non-perturbative regime. As such, the system provides a tunable setting to simulate topological Kondo semimetals. Finally, we discuss the pathways that connect the physics realized here to the Weyl Kondo semimetals and their proximate phases that have been advanced in recent years in topological Kondo lattice models and materials.

cond-mat.str-el↗