arXiv subjects

Xin Zhang

Publications and source records attributed to Xin Zhang.

At least 289 records · Page 16Linked to original sources

arXiv2026

PSGS: Text-driven Panorama Sliding Scene Generation via Gaussian Splatting

Generating realistic 3D scenes from text is crucial for immersive applications like VR, AR, and gaming. While text-driven approaches promise efficiency, existing methods suffer from limited 3D-text data and inconsistent multi-view stitching, resulting in overly simplistic scenes. To address this, we propose PSGS, a two-stage framework for high-fidelity panoramic scene generation. First, a novel two-layer optimization architecture generates semantically coherent panoramas: a layout reasoning layer parses text into structured spatial relationships, while a self-optimization layer refines visual details via iterative MLLM feedback. Second, our panorama sliding mechanism initializes globally consistent 3D Gaussian Splatting point clouds by strategically sampling overlapping perspectives. By incorporating depth and semantic coherence losses during training, we greatly improve the quality and detail fidelity of rendered scenes. Our experiments demonstrate that PSGS outperforms existing methods in panorama generation and produces more appealing 3D scenes, offering a robust solution for scalable immersive content creation.

cs.CV↗

arXiv2026

Spectra: Rethinking Optimizers for LLMs Under Spectral Anisotropy

Gradient signals in LLM training are highly anisotropic: recurrent linguistic structure concentrates energy into a small set of dominant spectral directions, while context specific information resides in a long tail. We show that this spike tail separation persists throughout training, with the spike occupying only about 1.5% of directions yet dominating optimizer statistics. This dominance suppresses tail learning by contracting tail updates through second moment normalization and tightening the globally stable learning rate bound. Motivated by this analysis, we propose Spectra, a spike aware optimizer that suppresses the dominant low rank spike subspace without amplifying the noise sensitive spectral tail. Spectra tracks the spike subspace via cached, warm started power iteration and applies low rank spectral shaping with negligible overhead and substantially reduced optimizer state memory. On LLaMA3 8B trained on 50B tokens, Spectra reaches the same target loss 30% faster than AdamW, reduces per step end to end overhead by 0.7%, cuts optimizer state memory by 49.25%, and improves average downstream accuracy by 1.62%. Compared to Muon, Spectra is 5.1x faster in optimizer processing time, achieves a lower final loss, and improves average accuracy by 0.66%.

cs.LG↗

arXiv2026

SemBind: Binding Diffusion Watermarks to Semantics Against Black-Box Forgery Attacks

Latent-based watermarks, integrated into the generation process of latent diffusion models (LDMs), simplify detection and attribution of generated images. However, recent black-box forgery attacks, where an attacker needs at least one watermarked image and black-box access to the provider's model, can embed the provider's watermark into images not produced by the provider, posing outsized risk to provenance and trust. We propose SemBind, the first defense framework for latent-based watermarks that resists black-box forgery by binding latent signals to image semantics via a learned semantic masker. Trained with contrastive learning, the masker yields near-invariant codes for the same prompt and near-orthogonal codes across prompts; these codes are reshaped and permuted to modulate the target latent before any standard latent-based watermark. SemBind is generally compatible with existing latent-based watermarking schemes and keeps image quality essentially unchanged, while a simple mask-ratio parameter offers a tunable trade-off between anti-forgery strength and robustness. Across four mainstream latent-based watermark methods, our SemBind-enabled anti-forgery variants markedly reduce false acceptance under black-box forgery while providing a controllable robustness-security balance.

cs.CR↗

arXiv2026

Speaking Clearly: A Simplified Whisper-Based Codec for Low-Bitrate Speech Coding

Speech codecs serve as bridges between continuous speech signals and large language models, yet face an inherent conflict between acoustic fidelity and semantic preservation. To mitigate this conflict, prevailing methods augment acoustic codecs with complex semantic supervision. We explore the opposite direction: a semantic-first approach that starts from a semantically-capable model and adapts it for high-fidelity acoustic reconstruction. Through empirical analysis, we discover that targeted architectural simplification can unlock the acoustic modeling potential of Whisper, a text-aligned Automatic Speech Recognition (ASR) model. Based on this finding, we propose SimWhisper-Codec, a novel codec that balances the semantic and acoustic preservation by leveraging a frozen, simplified Whisper encoder without requiring external supervision. Experimental results demonstrate that SimWhisper-Codec achieves superior performance in both semantic preservation and acoustic quality compared to semantically-supervised codecs such as Mimi Codec and SpeechTokenizer at similar bitrates, validating the effectiveness of our semantic-first approach. Code is available at https://github.com/ZhangXinWhut/SimWhisper-Codec.

cs.SD↗

arXiv2026

Circularly polarized metamaterial cage for homogeneous signal-to-noise ratio enhancement in magnetic resonance imaging

The signal-to-noise ratio (SNR) in magnetic resonance imaging (MRI) governs the quality of signal detection and directly impacts the clarity and reliability of the acquired images. Recent advances in metamaterials have enabled lightweight solutions with selective magnetic responses, offering a route to locally boost SNR in targeted anatomical regions but often with compromised field homogeneity. Here, a wireless metamaterial cage constructed from coaxial cables is engineered for homogeneous SNR enhancement at 3.0 T. With its cylindrical geometry and electromagnetic architecture, the device supports circularly polarized resonance through engineered phase-shifted currents, enabling selective and omnidirectional interaction with the rotating B_1^- field to achieve uniform magnetic field distribution. Integrated with the body coil, the device yields a 32-fold SNR enhancement while maintaining comparable homogeneity to the body coil alone, exhibiting only 12.07% variation within the region of interest (ROI). Benchmarking against a state-of-the-art 16-channel extremity coil further shows that the metacage achieves at least 1.94-fold and 2.24-fold higher SNR in axial and coronal planes, respectively, and exhibits substantially lower SNR variation (12.07% compared to 54.83% for the extremity coil). The results establish the metacage as a compelling platform for next-generation wireless MRI technologies.

physics.med-ph↗

arXiv2026

21 cm forest one-dimensional power spectrum as an indirect probe of dark matter particles and primordial black holes

Understanding the nature of dark matter (DM) particles remains a pivotal challenge in modern cosmology. Current cosmological research on these phenomena primarily utilizes cosmic microwave background (CMB) observations and other late-time probes, which predominantly focus on large scales. We introduce a novel probe, the 21 cm forest signal, which can be used to investigate DM properties on small scales during the epoch of reionization, thereby addressing the gap left by other cosmological probes. Annihilation and decay of DM particles, as well as Hawking radiation from PBHs, can heat the intergalactic medium (IGM). This heating suppresses the amplitude of the 21 cm forest 1D power spectrum. Therefore, the 1D power spectrum provides an effective method for constraining DM properties. However, astrophysical heating processes in the early universe can also affect the 21 cm forest 1D power spectrum. In this work, we assess the potential of using the SKA to observe the 21 cm forest 1D power spectrum for constraining DM properties, under the assumption that astrophysical heating can be constrained reliably by other independent probes. Under low astrophysical heating conditions, the 1D power spectrum could constrain the DM annihilation cross section and decay lifetime to $\langleσv\rangle \sim {10^{-31}}\,{\rm cm^{3}\,s^{-1}}$ and $τ\sim {10^{30}}\,{\rm s}$ for ${10}\,{\rm GeV}$ DM particles, and probe PBHs with masses $\sim {10^{15}}\,{\rm\,g}$ at abundances $f_{\mathrm{PBH}} \simeq 10^{-13}$. These constraints represent improvements of 5-6 orders of magnitude over current limits. Furthermore, the 21 cm forest 1D power spectrum has the potential to exceed existing bounds on sub-GeV DM and to probe PBHs with masses above $10^{18}\,{\rm g}$, which are otherwise inaccessible by conventional cosmological probes.

astro-ph.CO↗

arXiv2026

AgriPINN: A Process-Informed Neural Network for Interpretable and Scalable Crop Biomass Prediction Under Water Stress

Accurate prediction of crop above-ground biomass (AGB) under water stress is critical for monitoring crop productivity, guiding irrigation, and supporting climate-resilient agriculture. Data-driven models scale well but often lack interpretability and degrade under distribution shift, whereas process-based crop models (e.g. DSSAT, APSIM, LINTUL5) require extensive calibration and are difficult to deploy over large spatial domains. To address these limitations, we propose AgriPINN, a process-informed neural network that integrates a biophysical crop-growth differential equation as a differentiable constraint within a deep learning backbone. This design encourages physiologically consistent biomass dynamics under water-stress conditions while preserving model scalability for spatially distributed AGB prediction. AgriPINN recovers latent physiological variables, including leaf area index (LAI), absorbed photosynthetically active radiation (PAR), radiation use efficiency (RUE), and water-stress factors, without requiring direct supervision. We pretrain AgriPINN on 60 years of historical data across 397 regions in Germany and fine-tune it on three years of field experiments under controlled water treatments. Results show that AgriPINN consistently outperforms state-of-the-art deep-learning baselines (ConvLSTM-ViT, SLTF, CNN-Transformer) and the process-based LINTUL5 model in terms of accuracy (RMSE reductions up to $43\%$) and computational efficiency. By combining the scalability of deep learning with the biophysical rigor of process-based modeling, AgriPINN provides a robust and interpretable framework for spatio-temporal AGB prediction, offering practical value for planning of irrigation infrastructure, yield forecasting, and climate-adaptation planning.

cs.AI↗

arXiv2026

A Practical Framework for Estimating the Repetition Likelihood of Fast Radio Bursts from Spectral Morphology

The repeating behavior of fast radio bursts (FRBs) is regarded as a key clue to understanding their physical origin, yet reliably distinguishing repeaters from apparent non-repeaters with current observations remains challenging. Here we propose a physically interpretable and practically quantifiable classification framework based on spectral morphology. Using dimensionality reduction, clustering, and feature-importance analysis, we identify the spectral running $r$ and spectral index $γ$ as the most critical parameters for distinguishing repeaters from apparent non-repeaters in the CHIME/FRB sample. In the $γ$-$r$ space, repeaters preferentially occupy regions with steeper, narrower-band spectra, whereas non-repeaters cluster in flatter, broader-band regions, resulting in a clear density separation. We further construct an empirical probability map in the $γ$-$r$ space, showing a clear gradient of repetition likelihood, from $\sim 65\%$ in the high-repetition region to $\sim 5\%$ in the low-repetition region. Combining this with Gaussian Mixture Model posterior analysis, we identify several apparent non-repeaters with high inferred repetition probability, recommending them as priority targets for future monitoring. This framework provides a simple and generalizable tool for assessing repeatability in the CHIME/FRB sample and highlights the diagnostic power of spectral morphology in unveiling FRB origins.

astro-ph.HE↗

arXiv2026

EVOREFUSE: Evolutionary Prompt Optimization for Evaluation and Mitigation of LLM Over-Refusal to Pseudo-Malicious Instructions

Large language models (LLMs) frequently refuse to respond to pseudo-malicious instructions: semantically harmless input queries triggering unnecessary LLM refusals due to conservative safety alignment, significantly impairing user experience. Collecting such instructions is crucial for evaluating and mitigating over-refusals, but existing instruction curation methods, like manual creation or instruction rewriting, either lack scalability or fail to produce sufficiently diverse and effective refusal-inducing prompts. To address these limitations, we introduce EVOREFUSE, a prompt optimization approach that generates diverse pseudo-malicious instructions consistently eliciting confident refusals across LLMs. EVOREFUSE employs an evolutionary algorithm exploring the instruction space in more diverse directions than existing methods via mutation strategies and recombination, and iteratively evolves seed instructions to maximize evidence lower bound on LLM refusal probability. Using EVOREFUSE, we create two novel datasets: EVOREFUSE-TEST, a benchmark of 582 pseudo-malicious instructions that outperforms the next-best benchmark with 85.34% higher average refusal triggering rate across 9 LLMs without a safety-prior system prompt, 34.86% greater lexical diversity, and 40.03% improved LLM response confidence scores; and EVOREFUSE-ALIGN, which provides 3,000 pseudo-malicious instructions with responses for supervised and preference-based alignment training. With supervised fine-tuning on EVOREFUSE-ALIGN, LLAMA3.1-8B-INSTRUCT achieves up to 29.85% fewer over-refusals than models trained on the second-best alignment dataset, without compromising safety. Our analysis with EVOREFUSE-TEST reveals models trigger over-refusals by overly focusing on sensitive keywords while ignoring broader context. Our code and datasets are available at https://github.com/FishT0ucher/EVOREFUSE.

cs.AI↗

arXiv2026

LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion

The safety mechanisms of large language models (LLMs) exhibit notable fragility, as even fine-tuning on datasets without harmful content may still undermine their safety capabilities. Meanwhile, existing safety alignment methods predominantly rely on the fine-tuning process, which inadvertently leads to the increased complexity and computational resources required. To address these issues, we introduce LSSF, a novel safety re-alignment framework with \underline{L}ow-Rank \underline{S}afety \underline{S}ubspace \underline{F}usion. Our proposed method exploits the low-rank characteristics of safety information in LLMs by constructing a low-rank projection matrix to extract the principal components of safety vectors. Notably, this projection matrix represents the low-rank safety subspace of the LLMs, which we have observed to remain stable during fine-tuning process and is isolated from the model's general capabilities. These principal components are used to effectively restore safety alignment when combined with fine-tuned LLMs through linear arithmetic. Additionally, to account for the varying encoding densities of safety information across different layers of LLMs, we propose a novel metric called safety singular value entropy. This metric quantifies the encoding density and allows for the dynamic computation of the safety-critical rank for each safety vector. Extensive experiments demonstrate that our proposed post-hoc alignment method can effectively restore the safety alignment of fine-tuned models with minimal impact on their performance in downstream tasks.

cs.CY↗

arXiv2026

RCDN: Real-Centered Detection Network for Robust Face Forgery Identification

Image forgery has become a critical threat with the rapid proliferation of AI-based generation tools, which make it increasingly easy to synthesize realistic but fraudulent facial content. Existing detection methods achieve near-perfect performance when training and testing are conducted within the same domain, yet their effectiveness deteriorates substantially in crossdomain scenarios. This limitation is problematic, as new forgery techniques continuously emerge and detectors must remain reliable against unseen manipulations. To address this challenge, we propose the Real-Centered Detection Network (RCDN), a frequency spatial convolutional neural networks(CNN) framework with an Xception backbone that anchors its representation space around authentic facial images. Instead of modeling the diverse and evolving patterns of forgeries, RCDN emphasizes the consistency of real images, leveraging a dual-branch architecture and a real centered loss design to enhance robustness under distribution shifts. Extensive experiments on the DiFF dataset, focusing on three representative forgery types (FE, I2I, T2I), demonstrate that RCDN achieves both state-of-the-art in-domain accuracy and significantly stronger cross-domain generalization. Notably, RCDN reduces the generalization gap compared to leading baselines and achieves the highest cross/in-domain stability ratio, highlighting its potential as a practical solution for defending against evolving and unseen image forgery techniques.

cs.CV↗

arXiv2026

Multi-messenger standard-siren cosmology for third-generation gravitational-wave detectors: forecasts considering observations of gamma-ray bursts and kilonovae

In the third-generation (3G) gravitational-wave (GW) detector era, GW multi-messenger observations for binary neutron star merger events can exert great impacts on exploring the cosmic expansion history. Extending the previous work, we explore the potential of 3G GW standard siren observations in cosmological parameter estimation by considering their associated electromagnetic (EM) counterparts, including $γ$-ray burst (GRB) coincidence observations by the Gravitational wave high-energy Electromagnetic Counterpart All-sky Monitor and GW-triggered target-of-opportunity observations of kilonovae by different optical survey projects. During an assumed 10-year observation, we predict that the number of detectable GW-kilonova events is $\sim 4900$ with redshifts below $\sim 0.4$ under GW network and Large Synoptic Survey Telescope in the $i$ band, which is three times more than that of GW-GRB detections. For the cosmological analysis, we find that with the inclusion of GW-kilonova detections, the constraints on cosmological parameters from GW-EM detections are significantly improved compared to those from GW-GRB detections. In particular, GW-EM detections can tightly constrain the Hubble constant with a precision ranging from $0.076\%$ to $0.034\%$. Moreover, GW multi-messenger observations could effectively break the cosmological parameter degeneracies generated by the mainstream EM observations, CMB+BAO+SN (CBS). The combination of CBS and GW-EM can tightly constrain the equation of state parameters of dark energy $w$ in the $w$CDM model and $w_0$ in the $w_0w_a$CDM model with precisions of $0.72\%$ and $0.99\%$, respectively, meeting the standard of precision cosmology. In conclusion, GW multi-messenger observations could play a crucial role in helping solve the Hubble tension and probing the fundamental nature of dark energy.

astro-ph.CO↗

arXiv2026

Implicit Graph, Explicit Retrieval: Towards Efficient and Interpretable Long-horizon Memory for Large Language Models

Long-horizon applications increasingly require large language models (LLMs) to answer queries when relevant evidence is sparse and dispersed across very long contexts. Existing memory systems largely follow two paradigms: explicit structured memories offer interpretability but often become brittle under long-context overload, while latent memory mechanisms are efficient and stable yet difficult to inspect. We propose LatentGraphMem, a memory framework that combines implicit graph memory with explicit subgraph retrieval. LatentGraphMem stores a graph-structured memory in latent space for stability and efficiency, and exposes a task-specific subgraph retrieval interface that returns a compact symbolic subgraph under a fixed budget for downstream reasoning and human inspection. During training, an explicit graph view is materialized to interface with a frozen reasoner for question-answering supervision. At inference time, retrieval is performed in latent space and only the retrieved subgraph is externalized. Experiments on long-horizon benchmarks across multiple model scales show that LatentGraphMem consistently outperforms representative explicit-graph and latent-memory baselines, while enabling parameter-efficient adaptation and flexible scaling to larger reasoners without introducing large symbolic artifacts.

cs.CL↗

arXiv2026

Loupe: A Generalizable and Adaptive Framework for Image Forgery Detection

The proliferation of generative models has raised serious concerns about visual content forgery. Existing deepfake detection methods primarily target either image-level classification or pixel-wise localization. While some achieve high accuracy, they often suffer from limited generalization across manipulation types or rely on complex architectures. In this paper, we propose Loupe, a lightweight yet effective framework for joint deepfake detection and localization. Loupe integrates a patch-aware classifier and a segmentation module with conditional queries, allowing simultaneous global authenticity classification and fine-grained mask prediction. To enhance robustness against distribution shifts of test set, Loupe introduces a pseudo-label-guided test-time adaptation mechanism by leveraging patch-level predictions to supervise the segmentation head. Extensive experiments on the DDL dataset demonstrate that Loupe achieves state-of-the-art performance, securing the first place in the IJCAI 2025 Deepfake Detection and Localization Challenge with an overall score of 0.846. Our results validate the effectiveness of the proposed patch-level fusion and conditional query design in improving both classification accuracy and spatial localization under diverse forgery patterns. The code is available at https://github.com/Kamichanw/Loupe.

cs.CV↗

arXiv2026

On regional treatment effect assessment using robust MAP priors

Bayesian dynamic borrowing has become an increasingly important tool for evaluating the consistency of regional treatment effects which is a key requirement for local regulatory approval of a new drug. It helps increase the precision of regional treatment effect estimate when regional and global data are similar, while guarding against potential bias when they differ. In practice, the two-component mixture prior, of which one mixture component utilizes the power prior to incorporate external data, is widely used. It allows convenient prior specification, analytical posterior computation, and fast evaluation of operating characteristics. Though the robust meta-analytical-predictive (MAP) prior is broadly used with multiple external data sources, it remains underutilized for regional treatment effect assessment (typically only one external data source is available) due to its inherit complexity in prior specification and posterior computation. In this article, we illustrate the applicability of the robust MAP prior in the regional treatment effect assessment by developing a closed-form approximation for its posterior distribution while leveraging its relationship with the power prior. The proposed methodology substantially reduces the computational burden of identifying prior parameters for desired operating characteristics. Moreover, we have demonstrated that the MAP prior is an attractive choice to construct the informative component of the mixture prior compared to the power prior. The advantage can be explained through a Bayesian hypothesis testing perspective. Using a real-world example, we illustrate how our proposed method enables efficient and transparent development of a Bayesian dynamic borrowing design to show regional consistency.

stat.ME↗

arXiv2026

Gravitational wave standard sirens: A brief review of cosmological parameter estimation

Gravitational wave (GW) observations are expected to serve as a powerful and independent probe of the expansion history of the universe. By providing direct and calibration-free measurements of luminosity distances through waveform analysis, GWs provide a fundamentally different and potentially more robust approach to measuring cosmic-scale distances compared to traditional electromagnetic (EM) observations, which is known as the standard siren method. In this review, we present an overview of recent developments in GW standard siren cosmology, including up-to-date $H_0$ constraints, and prospects for constraining cosmological parameters using future GW detections. A central focus of this review is the unique ability of GW observations to break cosmological parameter degeneracies inherent in the EM observations. We also briefly highlight the impact of systematic uncertainties, such as detector calibration, weak lensing, peculiar velocities, and host-galaxy catalog completeness, and corresponding potential mitigation strategies, which currently limit the constraint precision of cosmological parameters. Looking forward, we highlight the importance of combining GW standard sirens with other emerging late-universe cosmological probes such as fast radio bursts, 21 cm intensity mapping, and strong gravitational lensing to forge a precise cosmological probe for exploring the late universe.

astro-ph.CO↗

arXiv2025

TeleWorld: Towards Dynamic Multimodal Synthesis with a 4D World Model

World models aim to endow AI systems with the ability to represent, generate, and interact with dynamic environments in a coherent and temporally consistent manner. While recent video generation models have demonstrated impressive visual quality, they remain limited in real-time interaction, long-horizon consistency, and persistent memory of dynamic scenes, hindering their evolution into practical world models. In this report, we present TeleWorld, a real-time multimodal 4D world modeling framework that unifies video generation, dynamic scene reconstruction, and long-term world memory within a closed-loop system. TeleWorld introduces a novel generation-reconstruction-guidance paradigm, where generated video streams are continuously reconstructed into a dynamic 4D spatio-temporal representation, which in turn guides subsequent generation to maintain spatial, temporal, and physical consistency. To support long-horizon generation with low latency, we employ an autoregressive diffusion-based video model enhanced with Macro-from-Micro Planning (MMPL)--a hierarchical planning method that reduces error accumulation from frame-level to segment-level-alongside efficient Distribution Matching Distillation (DMD), enabling real-time synthesis under practical computational budgets. Our approach achieves seamless integration of dynamic object modeling and static scene representation within a unified 4D framework, advancing world models toward practical, interactive, and computationally accessible systems. Extensive experiments demonstrate that TeleWorld achieves strong performance in both static and dynamic world understanding, long-term consistency, and real-time generation efficiency, positioning it as a practical step toward interactive, memory-enabled world models for multimodal generation and embodied intelligence.

cs.CV↗

arXiv2025

A Stepwise-Enhanced Reasoning Framework for Large Language Models Based on External Subgraph Generation

Large Language Models (LLMs) have achieved strong performance across a wide range of natural language processing tasks in recent years, including machine translation, text generation, and question answering. As their applications extend to increasingly complex scenarios, however, LLMs continue to face challenges in tasks that require deep reasoning and logical inference. In particular, models trained on large scale textual corpora may incorporate noisy or irrelevant information during generation, which can lead to incorrect predictions or outputs that are inconsistent with factual knowledge. To address this limitation, we propose a stepwise reasoning enhancement framework for LLMs based on external subgraph generation, termed SGR. The proposed framework dynamically constructs query relevant subgraphs from external knowledge bases and leverages their semantic structure to guide the reasoning process. By performing reasoning in a step by step manner over structured subgraphs, SGR reduces the influence of noisy information and improves reasoning accuracy. Specifically, the framework first generates an external subgraph tailored to the input query, then guides the model to conduct multi step reasoning grounded in the subgraph, and finally integrates multiple reasoning paths to produce the final answer. Experimental results on multiple benchmark datasets demonstrate that SGR consistently outperforms strong baselines, indicating its effectiveness in enhancing the reasoning capabilities of LLMs.

cs.CL↗