Searcharxiv⌕ Search

arXiv subjects

Yuxin Zhou

Publications and source records attributed to Yuxin Zhou.

At least 19 recordsLinked to original sources
arXiv2026

On the Gardner Transition in the Ising Pure $p$-Spin Glass II

In previous work, we identified, for every $p\geq3$, a unique first critical temperature $β_1^p$ and proved that the Parisi measure is replica symmetric (RS) for $0<β\leqβ_1^p$ and one-step replica symmetry breaking (1-RSB) on a nonempty interval immediately above $β_1^p$. In this sequel, we determine the rest of the phase diagram. We prove that there is a unique second critical temperature $β_2^p>β_1^p$ such that the Parisi measure is 1-RSB for $β_1^p<β\leqβ_2^p$, while for $β>β_2^p$, $\operatorname{supp}μ_β=\{0\}\cup[q_β,q'_β],$ for some $0<q_β<q'_β<1$, and is therefore full replica symmetry breaking (FRSB). Combined with our earlier results, this establishes the two transitions predicted by Gardner for the Ising pure $p$-spin glass.

math.PR↗
arXiv2026

Automatic Knowledge Graph Construction and Query for Earthquake Catalogs

In recent years, the number of events in earthquake catalogs has significantly increased due to the utilization of more effective deep learning based detectors and phase pickers but answering open ended questions such as what characterizes this sequence? remains constrained by rigid spatiotemporal windowing and subjective expert interpretation. We present the first systematic application of graph based retrieval augmented generation GraphRAG directly to raw, tabular catalog records across three independently featured catalogs, a reservoir adjacent swarm, the 2019 Ridgecrest tectonic sequence, and the 2021 Maduo Mw7.4 aftershock sequence. Without the need for manual data structuring, the pipeline builds structurally complete, queryable knowledge graphs for all three. Rigorous evaluation individually verified against catalog derived ground truth and a rule based reference graph exposes failure modes, and four seismology informed prompt fixes eliminate all targeted fabrications while sharply improving mechanism reasoning. A vector RAG baseline demonstrates the graph layers distinctive value, catalog wide summarization and temporal stage comparison. In addition, we have identified two main pitfalls that need attention. GraphRAG thus offers a practical, transferable, near zero cost query interface for earthquake catalogs, where careful prompting ensures the results are consistently accurate and trustworthy.

physics.geo-ph↗
arXiv2026

Spatiotemporal Analysis for Frequency-Magnitude Distribution of Earthquakes Across Mainland China: Comparing Classical and b-positive b-values Across Tectonic Regimes

The Gutenberg-Richter power-law relationship essentially governs the frequency-magnitude distribution (FMD) of seismic activity. The b-value, as the slope of this distribution, is widely recognized not only as a quantifier of the relative proportion of small to large events but also as a diagnostic precursor of crustal differential stress and potential big to huge earthquakes. This study investigates the spatiotemporal evolution of b-values across major tectonic regions in mainland China, for instance, the Sichuan-Yunnan, North China, and Northwest regions. Our analysis of over 50 years of seismic records systematically examines b-value variations across major seismogenic zones in mainland China to evaluate their utility as precursors for moderate-to-strong earthquakes. Moreover, our comparison suggests that the b-positive estimator is less affected by changes in detection capability and short-term aftershock incompleteness than the classical b-value estimator in the cases examined. While the Sichuan-Yunnan region displays a complex "mosaic" of high and low b-values reflecting fluid-driven processes and localized locking, the North China Plain and Northwest regions exhibit consistently lower b-values, suggesting high background stress and rigid crustal integrity. We also find that several of the rupture areas of moderate-to-large earthquakes in our catalog coincide with localized low-b patches identified in earlier long-term windows. This study provides a foundation for systematic, prospective evaluation of b-value-based precursor monitoring in mainland China.

physics.geo-ph↗
arXiv2026

Juxtaposition of Shallow Reservoir-Triggered Seismicity and Deep Tectonic Locking in the Qiaojia-Dongchuan Seismic Gap

Identifying the critical state of mature seismic gaps is challenging, especially when anthropogenic stress perturbations, such as reservoir impoundment, superimpose on tectonic loading. Here, utilizing a high-resolution dense array catalog from the Qiaojia-Dongchuan seismic gap (hosting the second-largest hydropower station in the world), we reveal a distinct vertical decoupling mechanism. The shallow activities exhibit high b-values (1.0), indicative of fluid-driven reservoir-triggered seismicity. Conversely, deep seismicity (20 km) outlines a 'locked asperity' characterized by low b-values (less than 0.8) and high Coulomb stress accumulation rate. We further identify a complex dipping structure, suggesting compound fault kinematics. Additionally, the calculated stress accumulation suggests this seismic gap is in a critical state with elevated rupture potential. Our findings indicate that shallow induced seismicity can mask the silent accumulation of deep tectonic strain. This decoupling model provides a new framework for assessing seismic risks in reservoir-fault systems globally.

physics.geo-ph↗
arXiv2026

LAP: Simple Command-line Tools for Teaching Logic, Algorithms, and Proof in Computer Science

The LAP toolset is a set of command line tools for teaching logic in computer science. It provides implementations of standard algorithms for propositional and first order logic, including conversions to various normal forms, propositional satisfiability algorithms such as DPLL, Tseytin's transformation, and equivalence checking. Significantly, LAP also supports a language for expressing a natural deduction derivation for propositional or first order logic. The tools can check the derivation, provide meaningful feedback if it is wrong, or display the derivation in a variety of formats. The toolset is written in Java and has no dependencies other than a Java Virtual Machine. The code has been designed to be easy to read and to illuminate the data definitions and algorithms.

cs.LO↗
arXiv2026

AutoVecCoder: Teaching LLMs to Generate Explicitly Vectorized Code

Vectorization via Single Instruction, Multiple Data (SIMD) architectures is a cornerstone of high-performance computing. To fully exploit hardware potential, developers often resort to explicit vectorization using intrinsics, as compiler-based auto-vectorization frequently yields suboptimal results due to conservative static analysis. While Large Language Models (LLMs) have demonstrated remarkable proficiency in general code generation, they struggle with explicit vectorization due to the scarcity of high-quality corpora and the strict semantic constraints of low-level hardware instructions. In this paper, we propose AutoVecCoder, a novel framework designed to empower LLMs with the capability of automated explicit vectorization. AutoVecCoder integrates two core components: VecPrompt, an automated data synthesis pipeline to inject domain-specific intrinsic knowledge; and VecRL, a reinforcement learning framework that aligns code generation with execution efficiency. AutoVecCoder-8B trained by this framework achieves state-of-the-art performance on the SSE and AVX subsets of SimdBench and, in some cases, generates implementations surpassing standard -O3 optimizations, effectively overcoming the inherent bottlenecks of traditional automated vectorization.

cs.CL↗
arXiv2026

BLASST: Dynamic BLocked Attention Sparsity via Softmax Thresholding

The growing demand for long-context inference capabilities in Large Language Models (LLMs) has intensified the computational and memory bottlenecks inherent to the self-attention mechanism. To address this challenge, we introduce BLASST, a drop-in, dynamic sparse attention mechanism that accelerates inference by using only a fixed scalar threshold to skip attention blocks. Our method targets practical inference deployment by removing the barriers to adoption present in existing works. As such, BLASST eliminates training requirements, avoids expensive pre-computation passes, accelerates both prefill and decode across all major attention variants (MHA, GQA, MQA, and MLA), provides optimized support for modern hardware, and easily integrates into existing frameworks. This is achieved by reusing online softmax statistics to identify negligible attention scores, skipping softmax, value block loads, and the subsequent matrix multiplication. We demonstrate the BLASST algorithm by delivering optimized kernels with negligible latency overhead. Our automated threshold calibration procedure reveals a simple inverse relationship between optimal threshold and context length, meaning we require only a single threshold each for prefill and decode per model. Preserving benchmark accuracy, we demonstrate a 1.52x speedup for prefill at 71.9% sparsity and a 1.48x speedup for decode at 73.2% sparsity on modern GPUs.

cs.CL↗
arXiv2026

True Bound States in the Continuum in Compact All-Dielectric Structures

Bound states in the continuum (BICs), known for their theoretically infinite quality (Q) factors and strong field localization, hold great promise for high-performance photonic devices. However, conventional true BICs typically rely on infinitely periodic structures, and their realization in finite-sized compact structures faces fundamental challenges, which severely limits device miniaturization and integration. In this work, a compact BIC design method based on optical conformal mapping is proposed, where a conventionally infinite periodic structure extended along one direction is mapped into a finite-sized annular structure. This symmetry transition, i.e., from translational to rotational, enables structural miniaturization while fully preserving the eigenvalues and BIC type of the original system. These transformations require only the adjustment of background permittivity and source distribution, without introducing extreme material parameters. As a concrete example, we show through theoretical and numerical analysis that a transformed compact all-dielectric structure, consisting of a double annular dielectric grating embedded in a gradient-index dielectric background, can support true BICs in a finite region. This work provides a simple and general strategy for achieving true BICs in compact all-dielectric structures, paving the way toward miniaturized high-Q photonic devices.

physics.optics↗
arXiv2026

A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training

We investigate the functional role of emergent outliers in large language models, specifically attention sinks (a few tokens that consistently receive large attention logits) and residual sinks (a few fixed dimensions with persistently large activations across most tokens). We hypothesize that these outliers, in conjunction with the corresponding normalizations (\textit{e.g.}, softmax attention and RMSNorm), effectively rescale other non-outlier components. We term this phenomenon \textit{outlier-driven rescaling} and validate this hypothesis across different model architectures and training token counts. This view unifies the origin and mitigation of both sink types. Our main conclusions and observations include: (1) Outliers function jointly with normalization: removing normalization eliminates the corresponding outliers but degrades training stability and performance; directly clipping outliers while retaining normalization leads to degradation, indicating that outlier-driven rescaling contributes to training stability. (2) Outliers serve more as rescale factors rather than contributors, as the final contributions of attention and residual sinks are significantly smaller than those of non-outliers. (3) Outliers can be absorbed into learnable parameters or mitigated via explicit gated rescaling, leading to improved training performance (average gain of 2 points) and enhanced quantization robustness (1.2 points degradation under W4A4 quantization).

cs.CL↗
arXiv2025

VLCache: Computing 2% Vision Tokens and Reusing 98% for Vision-Language Inference

This paper presents VLCache, a cache reuse framework that exploits both Key-Value (KV) cache and encoder cache from prior multimodal inputs to eliminate costly recomputation when the same multimodal inputs recur. Unlike previous heuristic approaches, we formally identify the cumulative reuse error effect and demonstrate how to minimize the non-prefix cache reuse error effectively. We further analyze the varying importance of model layers and propose a dynamic, layer-aware recomputation strategy to balance accuracy and efficiency. Experimental results show that VLCache achieves an accuracy on par with full recomputation, while requiring only 2-5% of the tokens to compute, yielding 1.2x-16x TTFT speedups. We develop an experimental implementation of the proposed VLCache pipeline based on SGLang, enabling significantly faster inference in practical deployments.

cs.CV↗
arXiv2025

SplatCo: Structure-View Collaborative Gaussian Splatting for Detail-Preserving Rendering of Large-Scale Unbounded Scenes

We present SplatCo, a structure-view collaborative Gaussian splatting framework for high-fidelity rendering of complex outdoor scenes. SplatCo builds upon three novel components: 1) a cross-structure collaboration module that combines global tri-plane representations, which capture coarse scene layouts, with local context grid features representing fine details. This fusion is achieved through a hierarchical compensation mechanism, ensuring both global spatial awareness and local detail preservation; 2) a cross-view pruning mechanism that removes overfitted or inaccurate Gaussians based on structural consistency, thereby improving storage efficiency and preventing rendering artifacts; 3) a structure view co-learning module that aggregates structural gradients with view gradients,thereby steering the optimization of Gaussian geometric and appearance attributes more robustly. By combining these key components, SplatCo effectively achieves high-fidelity rendering for large-scale scenes. Code and project page are available at https://splatco-tech.github.io.

cs.CV↗
arXiv2025

Qwen3Guard Technical Report

As large language models (LLMs) become more capable and widely used, ensuring the safety of their outputs is increasingly critical. Existing guardrail models, though useful in static evaluation settings, face two major limitations in real-world applications: (1) they typically output only binary "safe/unsafe" labels, which can be interpreted inconsistently across diverse safety policies, rendering them incapable of accommodating varying safety tolerances across domains; and (2) they require complete model outputs before performing safety checks, making them fundamentally incompatible with streaming LLM inference, thereby preventing timely intervention during generation and increasing exposure to harmful partial outputs. To address these challenges, we present Qwen3Guard, a series of multilingual safety guardrail models with two specialized variants: Generative Qwen3Guard, which casts safety classification as an instruction-following task to enable fine-grained tri-class judgments (safe, controversial, unsafe); and Stream Qwen3Guard, which introduces a token-level classification head for real-time safety monitoring during incremental text generation. Both variants are available in three sizes (0.6B, 4B, and 8B parameters) and support up to 119 languages and dialects, providing comprehensive, scalable, and low-latency safety moderation for global LLM deployments. Evaluated across English, Chinese, and multilingual benchmarks, Qwen3Guard achieves state-of-the-art performance in both prompt and response safety classification. All models are released under the Apache 2.0 license for public use.

cs.CL↗
arXiv2025

Large-scale dynamics in visual quorum sensing chiral suspensions

Motility induced phase separation is an efficient aggregation mechanism of active matter, yet biological systems exhibit richer organization through communication among constituents. We investigate suspensions of active particles that change chirality when neighbor density within their visual cone exceeds a threshold, a communication based non-reciprocal interaction akin to quorum sensing. Tuning the visual cone triggers programmable transitions: from disorder to phase separation to hyper-uniformity. Notably, phase separation triggers large-scale circulation, with robust edge currents persistently flowing around dense clusters, while particle distributions inside become effectively hyper-uniform. These are genuine non-reciprocal effects which occur even in the absence of steric interactions. Remarkably, in active-passive mixtures, only 5% quorum-sensing chiral particles suffice to induce collective circulation. Thus, simple perception-based rules can generate life-like order, offering design principles for programmable active materials and micro-robotic swarms.

cond-mat.soft↗
arXiv2025

Binary Mixtures in Linear Convection Arrays

We numerically investigated the dynamics of a mixture of finite-size active and passive particles in a linear array of convection rolls. The interplay of advection and steric interactions produces a number of interesting effects, like the stirring of a passive colloidal fluid by a small fraction of slow active particles, or the separation of the mixture active and passive colloidal fractions by increasing the motility of the active one, which eventually clusters in stagnation areas along the array walls. These mechanisms are quantitatively characterized by studying the dependence of the diffusion constants of the active and passive particles on the parameters of the active mixture fraction.

cond-mat.soft↗
arXiv2025

FloE: On-the-Fly MoE Inference on Memory-constrained GPU

With the widespread adoption of Mixture-of-Experts (MoE) models, there is a growing demand for efficient inference on memory-constrained devices. While offloading expert parameters to CPU memory and loading activated experts on demand has emerged as a potential solution, the large size of activated experts overburdens the limited PCIe bandwidth, hindering the effectiveness in latency-sensitive scenarios. To mitigate this, we propose FloE, an on-the-fly MoE inference system on memory-constrained GPUs. FloE is built on the insight that there exists substantial untapped redundancy within sparsely activated experts. It employs various compression techniques on the expert's internal parameter matrices to reduce the data movement load, combined with low-cost sparse prediction, achieving perceptible inference acceleration in wall-clock time on resource-constrained devices. Empirically, FloE achieves a 9.3x compression of parameters per expert in Mixtral-8x7B; enables deployment on a GPU with only 11GB VRAM, reducing the memory footprint by up to 8.5x; and delivers a 48.7x inference speedup compared to DeepSpeed-MII on a single GeForce RTX 3090 - all with only a 4.4$\%$ - 7.6$\%$ average performance degradation.

cs.LG↗
arXiv2025

Stealthy Jailbreak Attacks on Large Language Models via Benign Data Mirroring

Large language model (LLM) safety is a critical issue, with numerous studies employing red team testing to enhance model security. Among these, jailbreak methods explore potential vulnerabilities by crafting malicious prompts that induce model outputs contrary to safety alignments. Existing black-box jailbreak methods often rely on model feedback, repeatedly submitting queries with detectable malicious instructions during the attack search process. Although these approaches are effective, the attacks may be intercepted by content moderators during the search process. We propose an improved transfer attack method that guides malicious prompt construction by locally training a mirror model of the target black-box model through benign data distillation. This method offers enhanced stealth, as it does not involve submitting identifiable malicious instructions to the target model during the search phase. Our approach achieved a maximum attack success rate of 92%, or a balanced value of 80% with an average of 1.5 detectable jailbreak queries per sample against GPT-3.5 Turbo on a subset of AdvBench. These results underscore the need for more robust defense mechanisms.

cs.CL↗
arXiv2025

Diffusion Transients in Motility-Induced Phase Separation

We numerically investigate normal diffusion in a two-dimensional athermal suspension of active particles undergoing motility-induced phase separation. The particles are modeled as achiral Janus disks with fixed self-propulsion speed and weakly fluctuating orientation. When plotted versus the overall suspension packing fraction, the relevant diffusion constant traces a hysteresis loop with sharp jumps in correspondence with the binodal and spinodal of the gaseous phase. No hysteresis loop is observed between the spinodal and binodal of the dense phase, as they appear to overlap. Moreover, even under steady-state phase separation, the particle displacement distributions exhibit non-Gaussian normal diffusion with transient fat (thin) tails in the presence (absence) of phase separation.

cond-mat.soft↗