Searcharxiv⌕ Search

arXiv subjects

Zhiyu Zhang

Publications and source records attributed to Zhiyu Zhang.

At least 19 recordsLinked to original sources
arXiv2026

LG-VLN: A Zero-Shot Vision-and-Language Navigation Framework with LangGraph State Orchestration

Continuous-environment vision-and-language navigation (VLN-CE) requires interpreting natural-language instructions in unseen 3D environments and executing continuous low-level actions. Existing methods often depend on LiDAR, panoramic cameras, or extra sensors; separate geometric-mapping and semantic-navigation visual representations can cause long-trajectory spatial-semantic inconsistencies. We propose LG-VLN, a monocular zero-shot framework with shared visual features and LangGraph-based state orchestration. An online feed-forward 3D reconstruction network predicts depth, camera poses, and dense point clouds for agent-pose estimation and global map fusion. Geometry and navigation share dense CleanDIFT features: semantic consistency rejects incorrect inter-frame correspondences, while target-instance constraints define visual references whose similarity combines with local BLIP-2 image-text relevance to form a semantic value map. LangGraph represents instruction parsing, geometric perception, semantic value updates, path planning, action execution, and failure recovery as a directed state graph with conditional transitions, persistent state, and modular recovery mechanisms. On a fixed 550-episode subset of the R2R-CE val-unseen split, LG-VLN achieves 21.3% success and 12.1% success weighted by path length. Ablations show shared semantic features improve navigation, further boosted by combining visual similarity and image-text relevance. Results establish shared visual representations and explicit state orchestration as effective for zero-shot VLN-CE using monocular RGB alone. Code will be publicly released for reproducibility.

cs.CV↗
arXiv2026

Complete-Coverage Searches for Lorentz Violation in the Minimal Matter Sector

Over the past several decades, dozens of tests have sought the 132 Lorentz-violating degrees of freedom in the nonrelativistic limit of the minimal matter sector of the Standard-Model Extension, yet 43 remained unconstrained. In this Letter, we limit all previously unconstrained degrees of freedom and make improvements on 13 prior limits. The approach introduced here offers the potential of further improvements for 49 degrees of freedom in suitable future experiments, along with additional discovery potential offered by combining data from experiments performed in different locations.

hep-ph↗
arXiv2026

WeSCE: A Benchmark for Measuring Security Drift in LLM-Driven Code Editing

In this work, we introduce WeSCE, a benchmark for quantifying security drift in code editing under weak-security constraints, where tasks specify only functional objectives without explicit security requirements. WeSCE consists of 400 executable programs derived from real-world code, covering feature addition, feature removal, bug fixing, and refactoring. To quantify security drift, we propose a continuous risk representation that aggregates heterogeneous vulnerability signals through a unified formulation, and define drift measures capturing changes in overall risk, worst-case severity, and vulnerability distribution under code transformations, providing a multi-scale view of security spanning average-case behavior to worst-case emphasis.

cs.CR↗
arXiv2026

GST-Bench: Can VLMs Develop Global Spatial Awareness from Video?

Spatial intelligence is fundamental to embodied agents, yet existing benchmarks focus on local spatial perception from single or few viewpoints, overlooking global spatial awareness over continuous, long-horizon visual streams. To address this limitation, we introduce the Global-Spatial-Temporal Benchmark (GST-Bench), a VQA benchmark for global spatial intelligence in video understanding, comprising human-verified questions derived from 6,790 minutes of synthetically generated video. It requires models to perform accurate spatial inference from novel viewpoints unseen in the input video and to map egocentric observations onto global top-down images. A comprehensive evaluation of 22 state-of-the-art VLMs exposes a striking gap between models and humans: the strongest zero-shot model attains only 42.68, far below the human score of 79.08. To probe the cause of this gap, we construct GST-Bench-Local and find that models, despite strong local spatial understanding under the same task formulation, still fail to consolidate long-horizon observations into a globally consistent scene representation. We further provide GST-Train, a dataset for global spatial reasoning, as a complementary resource to facilitate future research on this challenge.

cs.CV↗
arXiv2026

SkillFab: An Agent-Native Skill Production Platform

SkillFab is an agent-native platform for turning missing capabilities into reviewed, reusable Agent Skills. At runtime, agents first search for reusable skills; when no adequate skill exists, the unmet capability becomes a demand-first issue before any repository or implementation branch needs to exist. Development then proceeds through a SkillFab-managed repository, Git-ingested commit evidence, maintainer review, and registry publication. The same lifecycle is exposed through web, REST, and MCP surfaces, so humans, scripts, and external agents operate on shared state rather than separate task logs. The current system uses scoped Git push URLs, native range commit ingestion, workflow-state reads, and workflow-event histories to make long-running agent work reviewable and recoverable. We document the platform model, architecture, implemented capabilities, and three case studies: an end-to-end OS-detect skill run, a Docker research package that converts operational practice into reusable skill knowledge, and an external optimization case showing how improved skill artifacts can enter SkillFab as reviewable, versioned submissions. Deployment: https://skillfab.ai.

cs.SE↗
arXiv2026

DMind Benchmark: Toward a Holistic Assessment of LLM Capabilities across the Web3 Domain

The Web3 ecosystem, underpinned by cryptographic primitives and decentralized consensus, represents a high-stakes environment where software vulnerabilities and incentive misalignments translate directly into financial loss. As Large Language Models (LLMs) are increasingly integrated into this domain for tasks ranging from smart contract auditing to decentralized finance analytics, ensuring their reliability is paramount. However, general-purpose benchmarks fail to capture the specialized reasoning required for these adversarial and protocol-driven settings. To bridge this gap, we introduce DMind Benchmark, a comprehensive evaluation suite designed to rigorously assess LLM proficiency across the Web3 stack. DMind Benchmark encompasses nine distinct subdomains (spanning infrastructure, smart contracts, token economics, etc.) and combines objective knowledge retrieval with complex open-ended reasoning tasks that emulate real-world operational challenges. We conduct an extensive evaluation of 31 leading proprietary and open-weights models, employing a contamination-aware pipeline and verifying the statistical robustness of our scoring protocol through rigorous cross-judge consistency checks. Our analysis reveals a critical dichotomy: while models demonstrate competence in foundational infrastructure concepts, they exhibit significant vulnerabilities in high-reasoning tasks such as security auditing. Furthermore, we provide a Pareto analysis to guide cost-effective deployment and demonstrate through adversarial experiments that high performance on DMind Benchmark necessitates genuine reasoning rather than superficial memorization. Since its open-source release in April 2025, DMind Benchmark achieved the #1 trending position on Hugging Face for nearly a week and accumulated over 13k downloads by June 2026, establishing itself as a standard for advancing secure and trustworthy AI in Web3.

cs.CR↗
arXiv2026

A pulsar escaping an ancient open cluster via tidal stripping

Open clusters are the primary birthplaces of stars in the Milky Way disk, yet their neutron star progeny are rarely found within them, presumably due to supernova-induced kicks that eject them at birth. Here we report the arcsec-level localization of the pulsar PSR J1921+3745 to the tidal tail of NGC 6791, one of the oldest and most massive open clusters. Our N-body simulation shows that more than 95% of neutron stars formed in such clusters have been ejected. This pulsar's location in the tidal tail indicates it was retained for billions of years before being stripped by Galactic tides. This long-term retention requires low natal kicks, consistent with formation via electron-capture supernova. Our findings capture a rare snapshot of a neutron star transitioning into the Galactic field, identifying tidal stripping of ancient clusters as a verified source of the Galactic neutron star population.

astro-ph.HE↗
arXiv2026

Content-Adaptive Rate-Quality Curve Prediction Model in Media Processing System

In streaming media services, video transcoding is a common practice to alleviate bandwidth demands. Unfortunately, traditional methods employing a uniform rate factor (RF) across all videos often result in significant inefficiencies. Content-adaptive encoding (CAE) techniques address this by dynamically adjusting encoding parameters based on video content characteristics. However, existing CAE methods are often tightly coupled with specific encoding strategies, leading to inflexibility. In this paper, we propose a model that predicts both RF-quality and RF-bitrate curves, which can be utilized to derive a comprehensive bitrate-quality curve. This approach facilitates flexible adjustments to the encoding strategy without necessitating model retraining. The model leverages codec features, content features, and anchor features to predict the bitrate-quality curve accurately. Additionally, we introduce an anchor suspension method to enhance prediction accuracy. Experiments confirm that the actual quality metric (VMAF) of the compressed video stays within 1 of the target, achieving an accuracy of 99.14%. By incorporating our quality improvement strategy with the rate-quality curve prediction model, we conducted online A/B tests, obtaining both +0.107% improvements in video views and video completions and +0.064% app duration time.

cs.MM↗
arXiv2026

MOVA: Towards Scalable and Synchronized Video-Audio Generation

Audio is indispensable for real-world video, yet generation models have largely overlooked audio components. Current approaches to producing audio-visual content often rely on cascaded pipelines, which increase cost, accumulate errors, and degrade overall quality. While systems such as Veo 3 and Sora 2 emphasize the value of simultaneous generation, joint multimodal modeling introduces unique challenges in architecture, data, and training. Moreover, the closed-source nature of existing systems limits progress in the field. In this work, we introduce MOVA (MOSS Video and Audio), an open-source model capable of generating high-quality, synchronized audio-visual content, including realistic lip-synced speech, environment-aware sound effects, and content-aligned music. MOVA employs a Mixture-of-Experts (MoE) architecture, with a total of 32B parameters, of which 18B are active during inference. It supports IT2VA (Image-Text to Video-Audio) generation task. By releasing the model weights and code, we aim to advance research and foster a vibrant community of creators. The released codebase features comprehensive support for efficient inference, LoRA fine-tuning, and prompt enhancement.

cs.CV↗
arXiv2026

Operationalizing Stein's Method for Online Linear Optimization: CLT-Based Optimal Tradeoffs

Adversarial online linear optimization (OLO) is essentially about making performance tradeoffs with respect to the unknown difficulty of the adversary. In the setting of one-dimensional fixed-time OLO on a bounded domain, it has been observed since Cover (1966) that achievable tradeoffs are governed by probabilistic inequalities, and these descriptive results can be converted into algorithms via dynamic programming, which, however, is not computationally efficient. We address this limitation by showing that Stein's method, a classical framework underlying the proofs of probabilistic limit theorems, can be operationalized as computationally efficient OLO algorithms. The associated regret and total loss upper bounds are "additively sharp", meaning that they surpass the conventional big-O optimality and match normal-approximation-based lower bounds by additive lower order terms. Our construction is inspired by the remarkably clean proof of a Wasserstein martingale central limit theorem (CLT) due to Röllin (2018). Several concrete benefits can be obtained from this general technique. First, with the same computational complexity, the proposed algorithm improves upon the total loss upper bounds of online gradient descent (OGD) and multiplicative weight update (MWU). Second, our algorithm can realize a continuum of optimal two-point tradeoffs between the total loss and the maximum regret over comparators, improving upon prior works in parameter-free online learning. Third, by allowing the adversary to randomize on an unbounded support, we achieve sharp in-expectation performance guarantees for OLO with noisy feedback.

stat.ML↗
arXiv2026

Coherent vibrational dynamics in molecular bond breaking: methyl radical umbrella mode probed by femtosecond x-ray spectroscopy

We report on the observation of coherent molecular vibrations launched by the breaking of a molecular bond. The methyl radical, which is produced by $267\,\mathrm{nm}$ photodissociation of methyl iodide, is excited to high levels in its $ν_2$ ``umbrella" vibrational mode by the dissociation. The ensuing coherent vibrational dynamics are observed by measuring ultrafast time-dependent changes in the x-ray transition energy from the C$1s$ to the singly-occupied valence orbital. Due to symmetry, the real space vibrational motion appears predominantly in the x-ray energy shift at the difference frequencies of the $ν_2$ progression, although the fundamental frequencies of the $ν_2$ mode are also observed. By constructing a fully quantum-mechanical model of the dynamics the coherent superposition is rigorously characterized and the real-space motion of the radicals is reconstructed. The retrieved trajectories are dominated by pronounced quantum beating governed by the high degree of coherent excitation and the strong negative anharmonicity of the $ν_2$ mode.

physics.atom-ph↗
arXiv2026

The Initial Mass Function as the Equilibrium State of a Variational Process: why the IMF cannot be sampled stochastically

The stellar initial mass function (sIMF) is often treated as a stochastic probability distribution, yet such an interpretation implies Poisson noise that is inconsistent with growing observational evidence. In particular, the observed relation between the mass of the most massive star formed in an embedded cluster and the cluster's total stellar mass supports a deterministic sampling process, known as optimal sampling. However, the physical origin of optimal sampling has not been formally established in the literature. In this work, we show that the stellar mass distribution implied by optimal sampling emerges from applying the Maximum Entropy principle to the fragmentation of star-forming clumps, whose structure is set by density-dependent cooling in the optically thin regime. Here, the maximum entropy leads to unbiased distributions. By applying calculus of variations to minimize the entropy functional obtained assuming fragmentation, we recover the power-law form of the sIMF, and we show that any distribution deviating from the sIMF violates the Maximum Entropy principle. This work provides a first-principles foundation for the deterministic nature of star formation. Thus, the sIMF is the distribution resulting from a maximally unbiased system.

astro-ph.GA↗
arXiv2026

MELA-TTS: Joint transformer-diffusion model with representation alignment for speech synthesis

This work introduces MELA-TTS, a novel joint transformer-diffusion framework for end-to-end text-to-speech synthesis. By autoregressively generating continuous mel-spectrogram frames from linguistic and speaker conditions, our architecture eliminates the need for speech tokenization and multi-stage processing pipelines. To address the inherent difficulties of modeling continuous features, we propose a representation alignment module that aligns output representations of the transformer decoder with semantic embeddings from a pretrained ASR encoder during training. This mechanism not only speeds up training convergence, but also enhances cross-modal coherence between the textual and acoustic domains. Comprehensive experiments demonstrate that MELA-TTS achieves state-of-the-art performance across multiple evaluation metrics while maintaining robust zero-shot voice cloning capabilities, in both offline and streaming synthesis modes. Our results establish a new benchmark for continuous feature generation approaches in TTS, offering a compelling alternative to discrete-token-based paradigms.

eess.AS↗
arXiv2026

CloneMem: Benchmarking Long-Term Memory for AI Clones

AI Clones aim to simulate an individual's thoughts and behaviors to enable long-term, personalized interaction, placing stringent demands on memory systems to model experiences, emotions, and opinions over time. Existing memory benchmarks primarily rely on user-agent conversational histories, which are temporally fragmented and insufficient for capturing continuous life trajectories. We introduce CloneMem, a benchmark for evaluating longterm memory in AI Clone scenarios grounded in non-conversational digital traces, including diaries, social media posts, and emails, spanning one to three years. CloneMem adopts a hierarchical data construction framework to ensure longitudinal coherence and defines tasks that assess an agent's ability to track evolving personal states. Experiments show that current memory mechanisms struggle in this setting, highlighting open challenges for life-grounded personalized AI. Code and dataset are available at https://github.com/AvatarMemory/CloneMemBench

cs.AI↗
arXiv2026

Achieving Full Coverage of the SME Minimal Matter Sector

Existing experiments and data sets can be leveraged to obtain additional sensitivities to Lorentz violation, beyond those originally sought, through a more precise consideration of the boost of the experiment through the background. In fact, access to the full coefficient space of the flat-spacetime single-fermion limit of the minimal matter sector of the Standard-Model Extension can be obtained. In this work we present this coverage for a sample particle in the context of a simplified model of Earth's motion.

hep-ph↗
arXiv2025

Central values of Asai L-functions and twisted Gan--Gross--Prasad conjecture

We study certain new relative trace formulas on (non-reductive) period integrals involving Weil representations, in the context of the relative Langlands program. We study normal representatives using Galois theory, and establish geometric decompositions of relative trace formulas using normal representatives for good test functions. By comparing global representatives, local distributions and orbital integrals, we prove the twisted Gan--Gross--Prasad (GGP) conjecture on Asai L-functions, in any dimension under some local assumptions, allowing ramifications of number fields.

math.NT↗
arXiv2025

Holographic Transformers for Complex-Valued Signal Processing: Integrating Phase Interference into Self-Attention

Complex-valued signals encode both amplitude and phase, yet most deep models treat attention as real-valued correlation, overlooking interference effects. We introduce the Holographic Transformer, a physics-inspired architecture that incorporates wave interference principles into self-attention. Holographic attention modulates interactions by relative phase and coherently superimposes values, ensuring consistency between amplitude and phase. A dual-headed decoder simultaneously reconstructs the input and predicts task outputs, preventing phase collapse when losses prioritize magnitude over phase. We demonstrate that holographic attention implements a discrete interference operator and maintains phase consistency under linear mixing. Experiments on PolSAR image classification and wireless channel prediction show strong performance, achieving high classification accuracy and F1 scores, low regression error, and increased robustness to phase perturbations. These results highlight that enforcing physical consistency in attention leads to generalizable improvements in complex-valued learning and provides a unified, physics-based framework for coherent signal modeling. The code is available at https://github.com/EonHao/Holographic-Transformers.

eess.SP↗