Searcharxiv⌕ Search

arXiv subjects

Gang Wang

Publications and source records attributed to Gang Wang.

At least 73 records · Page 4Linked to original sources
arXiv2026

Photonic state engineering via energy-level crossing by giant atoms in topological waveguide QED setup

Photonic state engineering in waveguide QED is typically based on local light-matter interactions. This limits its control over the spatial structure of bound photonic states. Here, we demonstrate a distinct mechanism arising from the interplay between nonlocal giant-atom coupling and topological band structure. Specifically, we consider giant atoms coupled to a Su-Schrieffer-Heeger waveguide and show that this configuration enables a controllable energy-level crossing protected by the topological gap. Adiabatically sweeping the atomic detuning across the crossing leads to a controlled exchange between distinct photonic bound states. In a two-giant-atom configuration, this mechanism achieves high-fidelity conversion of a spatially splitting state into a combining state. Extending this scheme to three-giant atoms, we further realize robust, shape-preserving photon transfer mediated by sequential in-gap crossings. Our results demonstrate how topology and nonlocal light-matter coupling can be combined to achieve programmable control of bound photonic states in waveguide QED platforms.

quant-ph↗
arXiv2026

Plug-and-Play Logit Fusion for Heterogeneous Pathology Foundation Models

Pathology foundation models (FMs) have become central to computational histopathology, offering strong transfer performance across a wide range of diagnostic and prognostic tasks. The rapid proliferation of pathology foundation models creates a model-selection bottleneck: no single model is uniformly best, yet exhaustively adapting and validating many candidates for each downstream endpoint is prohibitively expensive. We address this challenge with a lightweight and novel model fusion strategy, LogitProd, which treats independently trained FM-based predictors as fixed experts and learns sample-adaptive fusion weights over their slide-level outputs. The fusion operates purely on logits, requiring no encoder retraining and no feature-space alignment across heterogeneous backbones. We further provide a theoretical analysis showing that the optimal weighted product fusion is guaranteed to perform at least as well as the best individual expert under the training objective. We systematically evaluate LogitProd on \textbf{22} benchmarks spanning WSI-level classification, tile-level classification, gene mutation prediction, and discrete-time survival modeling. LogitProd ranks first on 20/22 tasks and improves the average performance across all tasks by ~3% over the strongest single expert. LogitProd enables practitioners to upgrade heterogeneous FM-based pipelines in a plug-and-play manner, achieving multi-expert gains with $\sim$12$\times$ lower training cost than feature-fusion alternatives.

cs.CV↗
arXiv2026

Efficient Learned Data Compression via Dual-Stream Feature Decoupling

While Learned Data Compression (LDC) has achieved superior compression ratios, balancing precise probability modeling with system efficiency remains challenging. Crucially, uniform single-stream architectures struggle to simultaneously capture micro-syntactic and macro-semantic features, necessitating deep serial stacking that exacerbates latency. Compounding this, heterogeneous systems are constrained by device speed mismatches, where throughput is capped by Amdahl's Law due to serial processing. To this end, we propose a Dual-Stream Multi-Scale Decoupler that disentangles local and global contexts to replace deep serial processing with shallow parallel streams, and incorporate a Hierarchical Gated Refiner for adaptive feature refinement and precise probability modeling. Furthermore, we design a Concurrent Stream-Parallel Pipeline, which overcomes systemic bottlenecks to achieve full-pipeline parallelism. Extensive experiments demonstrate that our method achieves state-of-the-art performance in both compression ratio and throughput, while maintaining the lowest latency and memory usage. The code is available at https://github.com/huidong-ma/FADE.

cs.CL↗
arXiv2026

Dual Contrastive Network for Few-Shot Remote Sensing Image Scene Classification

Few-shot remote sensing image scene classification (FS-RSISC) aims at classifying remote sensing images with only a few labeled samples. The main challenges lie in small inter-class variances and large intra-class variances, which are the inherent property of remote sensing images. To address these challenges, we propose a transfer-based Dual Contrastive Network (DCN), which incorporates two auxiliary supervised contrastive learning branches during the training process. Specifically, one is a Context-guided Contrastive Learning (CCL) branch and the other is a Detail-guided Contrastive Learning (DCL) branch, which focus on inter-class discriminability and intra-class invariance, respectively. In the CCL branch, we first devise a Condenser Network to capture context features, and then leverage a supervised contrastive learning on top of the obtained context features to facilitate the model to learn more discriminative features. In the DCL branch, a Smelter Network is designed to highlight the significant local detail information. And then we construct a supervised contrastive learning based on the detail feature maps to fully exploit the spatial information in each map, enabling the model to concentrate on invariant detail features. Extensive experiments on four public benchmark remote sensing datasets demonstrate the competitive performance of our proposed DCN.

cs.CV↗
arXiv2026

Object-Centric World Models from Few-Shot Annotations for Sample-Efficient Reinforcement Learning

While deep reinforcement learning (RL) from pixels has achieved remarkable success, its sample inefficiency remains a critical limitation for real-world applications. Model-based RL (MBRL) addresses this by learning a world model to generate simulated experience, but standard approaches that rely on pixel-level reconstruction losses often fail to capture small, task-critical objects in complex, dynamic scenes. We posit that an object-centric (OC) representation can direct model capacity toward semantically meaningful entities, improving dynamics prediction and sample efficiency. In this work, we introduce OC-STORM, an object-centric MBRL framework that enhances a learned world model with object representations extracted by a pretrained segmentation network. By conditioning on a minimal number of annotated frames, OC-STORM learns to track decision-relevant object dynamics and inter-object interactions without extensive labeling or access to privileged information. Empirical results demonstrate that OC-STORM significantly outperforms the STORM baseline on the Atari 100k benchmark and achieves state-of-the-art sample efficiency on challenging boss fights in the visually complex game Hollow Knight. Our findings underscore the potential of integrating OC priors into MBRL for complex visual domains. Project page: https://oc-storm.weipuzhang.com

cs.LG↗
arXiv2026

On the Stability of Spatially Distributed Cavity Laser and Boundary of Resonant Beam SLIPT

Spatially distributed cavity (SDC) lasers are a promising technology for simultaneous light information and power transfer (SLIPT), offering benefits such as increased mobility and intrinsic safety, which are advantageous for various Internet of Things (IoT) devices. \mll However, achieving beam transmission over meter-level long working distances presents significant challenges from cavity stability constraints, manufacturing/assembly tolerances, and diffraction losses\mrr. This paper conducts a theoretical investigation of the fundamental restrictions limiting long-range resonant beam generation. We investigate cavity stability and beam characteristics, and propose a binary-search-based Monte Carlo simulation algorithm as well as a linear approximation algorithm to quantify the maximum acceptable tolerances for stable operation. \mll Numerical results indicate that the stable region contracts sharply as distance increases. For fixed-component systems, an acceptable tolerance of 0.01 mm restricts the achievable transmission distance to less than 2 m. \mrr To address this limitation, we also prove the feasibility of long-range beam formation using precision adjustable elements, paving the way for advanced engineering applications. \mll Experimental results verified this assumption, demonstrating that by tuning the stable region during assembly, the transmission distance could be extended to 2.8 m. \mrr This work provides essential theoretical insights and practical design guidelines for realizing stable, long-range SDC systems.

eess.SY↗
arXiv2026

Accelerating TTL noise post-processing via combined coefficients and alternative TDI configuration

Tilt-to-length (TTL) noise induced by angular jitter of spacecraft and test masses can affect the sensitivity of space-based gravitational-wave detectors such as LISA, Taiji, and TianQin. Such angular jitter can be measured using the differential wavefront sensing technique, enabling the modeling and subtraction of TTL noise from the data. However, owing to the multiple degrees of freedom of the detector constellation, a linear TTL model requires at least 24 parameters, while a higher-fidelity quadratic model involves up to 60 coefficients, rendering parameter estimation computationally expensive. To accelerate parameter determination, we propose a modified parameter set obtained via a linear transformation of the original angular coupling coefficients, which effectively reduces correlations among TTL noise components. In addition, we perform parameter fitting using an alternative second-generation time-delay interferometry configuration, PD4L, rather than the fiducial Michelson configuration. These two improvements enhance the convergence speed of the fitting procedure by a factor of approximately 10 for the linear model and approximately 18 for the quadratic model. The proposed approach can therefore substantially improve the efficiency of TTL noise calibration in space-based gravitational-wave detectors.

gr-qc↗
arXiv2026

HySparse: A Hybrid Sparse Attention Architecture with Oracle Token Selection and KV Cache Sharing

This work introduces Hybrid Sparse Attention (HySparse), a new architecture that interleaves each full attention layer with several sparse attention layers. While conceptually simple, HySparse strategically derives each sparse layer's token selection and KV caches directly from the preceding full attention layer. This architecture resolves two fundamental limitations of prior sparse attention methods. First, conventional approaches typically rely on additional proxies to predict token importance, introducing extra complexity and potentially suboptimal performance. In contrast, HySparse uses the full attention layer as a precise oracle to identify important tokens. Second, existing sparse attention designs often reduce computation without saving KV cache. HySparse enables sparse attention layers to reuse the full attention KV cache, thereby reducing both computation and memory. We evaluate HySparse on both 7B dense and 80B MoE models. Across all settings, HySparse consistently outperforms both full attention and hybrid SWA baselines. Notably, in the 80B MoE model with 49 total layers, only 5 layers employ full attention, yet HySparse achieves substantial performance gains while reducing KV cache storage by nearly 10x.

cs.CL↗
arXiv2026

Mixture-of-World Models: Scaling Multi-Task Reinforcement Learning with Modular Latent Dynamics

A fundamental challenge in multi-task reinforcement learning (MTRL) is achieving sample efficiency in visual domains where tasks exhibit substantial heterogeneity in both observations and dynamics. Model-based reinforcement learning offers a promising path to improved sample efficiency through world models, but standard monolithic architectures struggle to capture diverse task dynamics, resulting in poor reconstruction and prediction accuracy. We introduce Mixture-of-World Models (MoW), a scalable architecture that combines modular variational autoencoders for task-adaptive visual compression, a hybrid Transformer-based dynamics model with task-conditioned experts and a shared backbone, and a gradient-based task clustering strategy for efficient parameter allocation. On the Atari 100k benchmark, a single MoW agent trained once on 26 Atari games achieves a mean human-normalized score of 110.4%, competitive with the score of 114.2% achieved by STORM, an ensemble of 26 task-specific models, while using 50% fewer parameters. On Meta-World, MoW achieves a 74.5% average success rate within 300 thousand environment steps, establishing a new state of the art. These results demonstrate that MoW provides a scalable and parameter-efficient foundation for generalist world models.

cs.LG↗
arXiv2026

Talos: Optimizing Top-$K$ Accuracy in Recommender Systems

Recommender systems (RS) aim to retrieve a small set of items that best match individual user preferences. Naturally, RS place primary emphasis on the quality of the Top-$K$ results rather than performance across the entire item set. However, estimating Top-$K$ accuracy (e.g., Precision@$K$, Recall@$K$) requires determining the ranking positions of items, which imposes substantial computational overhead and poses significant challenges for optimization. In addition, RS often suffer from distribution shifts due to evolving user preferences or data biases, further complicating the task. To address these issues, we propose Talos, a loss function that is specifically designed to optimize the Talos recommendation accuracy. Talos leverages a quantile technique that replaces the complex ranking-dependent operations into simpler comparisons between predicted scores and learned score thresholds. We further develop a sampling-based regression algorithm for efficient and accurate threshold estimation, and introduce a constraint term to maintain optimization stability by preventing score inflation. Additionally, we incorporate a tailored surrogate function to address discontinuity and enhance robustness against distribution shifts. Comprehensive theoretical analyzes and empirical experiments are conducted to demonstrate the effectiveness, efficiency, convergence, and distributional robustness of Talos. The code is available at https://github.com/cynthia-shengjia/WWW-2026-Talos.

cs.IR↗
arXiv2026

ChatAD: Reasoning-Enhanced Time-Series Anomaly Detection with Multi-Turn Instruction Evolution

LLM-driven Anomaly Detection (AD) helps enhance the understanding and explanatory abilities of anomalous behaviors in Time Series (TS). Existing methods face challenges of inadequate reasoning ability, deficient multi-turn dialogue capability, and narrow generalization. To this end, we 1) propose a multi-agent-based TS Evolution algorithm named TSEvol. On top of it, we 2) introduce the AD reasoning and multi-turn dialogue Dataset TSEData-20K and contribute the Chatbot family for AD, including ChatAD-Llama3-8B, Qwen2.5-7B, and Mistral-7B. Furthermore, 3) we propose the TS Kahneman-Tversky Optimization (TKTO) to enhance ChatAD's cross-task generalization capability. Lastly, 4) we propose a LLM-driven Learning-based AD Benchmark LLADBench to evaluate the performance of ChatAD and nine baselines across seven datasets and tasks. Our three ChatAD models achieve substantial gains, up to 34.50% in accuracy, 34.71% in F1, and a 37.42% reduction in false positives. Besides, via KTKO, our optimized ChatAD achieves competitive performance in reasoning and cross-task generalization on classification, forecasting, and imputation.

cs.AI↗
arXiv2026

AgentGC: Evolutionary Learning-based Lossless Compression for Genomics Data with LLM-driven Multiple Agent

Lossless compression has made significant advancements in Genomics Data (GD) storage, sharing and management. Current learning-based methods are non-evolvable with problems of low-level compression modeling, limited adaptability, and user-unfriendly interface. To this end, we propose AgentGC, the first evolutionary Agent-based GD Compressor, consisting of 3 layers with multi-agent named Leader and Worker. Specifically, the 1) User layer provides a user-friendly interface via Leader combined with LLM; 2) Cognitive layer, driven by the Leader, integrates LLM to consider joint optimization of algorithm-dataset-system, addressing the issues of low-level modeling and limited adaptability; and 3) Compression layer, headed by Worker, performs compression & decompression via a automated multi-knowledge learning-based compression framework. On top of AgentGC, we design 3 modes to support diverse scenarios: CP for compression-ratio priority, TP for throughput priority, and BM for balanced mode. Compared with 14 baselines on 9 datasets, the average compression ratios gains are 16.66%, 16.11%, and 16.33%, the throughput gains are 4.73x, 9.23x, and 9.15x, respectively.

cs.AI↗
arXiv2026

Unusual antiferromagnetic order and fluctuations in RbMn$_{6}$Bi$_{5}$

Quasi-one-dimensional RbMn$_{6}$Bi$_{5}$, the first pressure-induced ternary Mn-based superconductor, exhibits a phase diagram analogous to those of cuprate and iron-based superconductors, with superconductivity neighboring antiferromagnetic order. Here, we use $^{55}$Mn and $^{87}$Rb nuclear magnetic resonance (NMR) to unravel its magnetic structure and fluctuations. Above the Néel temperature ($T_{\rm N}$), strong antiferromagnetic fluctuations dominate, characteristic of a paramagnetic state with pronounced spin-lattice relaxation rate enhancement. Below $T_{\rm N}$, a first-order phase transition establishes a commensurate antiferromagnetic order, where Mn atoms at the pentagon corners exhibit distinct magnetic moments with different orientations, while the central Mn atom carries no magnetic moment. The complex magnetic architecture, revealed by zero-field and high-magnetic-field NMR spectra, contrasts with earlier neutron diffraction models proposing uniform spin density waves, instead supporting localized moments ordering with charge rearrangement. The proximity of robust antiferromagnetic fluctuations to the high-pressure superconducting phase suggests a potential role for magnetic excitations in mediating unconventional Cooper pairing, akin to paradigmatic high-$T_c$ systems. These findings provide critical insights into the interplay between geometric frustration, magnetic order, and superconductivity in manganese-based materials.

cond-mat.supr-con↗
arXiv2026

Time delay interferometry with minimal null frequencies and shortened time span

Time-delay interferometry (TDI) is essential for suppressing laser frequency noise in space-based gravitational wave (GW) observatories such as LISA. However, current second-generation TDI schemes often exhibit undesirable null frequencies and require long delay spans, which can impair data analysis performance. In this work, we introduce an alternative TDI configuration PD4L designed to minimize null frequencies and operate with a shorter effective time span. Constructed by synthesizing two distinct first-generation TDI schemes, PD4L achieves a delay span of 4$L$ (where $L$ is the arm length), half that of the standard Michelson and hybrid Relay configurations. We assess PD4L's performance by evaluating the spectral stability of instrumental noise via arm-length derivatives, simulating chirping GW signals from coalescing massive black hole binaries, and comparing waveform responses. Parameter estimation is performed in the frequency domain, and noise characterization is examined under realistic orbital dynamics. As demonstrated by the comparisons, the compact structure of PD4L offers several advantages: (1) reduced data margins at segment boundaries, (2) mitigated aliasing effects in the high-frequency regime, and (3) shortened signal tails arising from extended delay spans. Additionally, PD4L's null channels exhibit the same minimal null frequencies as its science channels, while maintaining greater spectral stability than other null streams. Overall, PD4L improves parameter estimation accuracy at high frequencies and supports reliable noise characterization over observation periods of up to four months. These results highlight PD4L as a compact and effective alternative for future TDI implementations, especially in high-frequency GW data analysis for LISA-like missions.

gr-qc↗
arXiv2026

Majorana Zero Modes and Topological Nature in Bi2Ta3S6-family Superconductors

In this work, we report that Bi2Ta3S6-family superconductors exhibit nontrivial band topology. They possess a natural quantum-well structure consisting of alternating stacks of TaS2 and honeycomb Bi layers, which contribute superconducting and topological properties, respectively. Symmetry-based indicators $(\mathbb{Z}_4;\mathbb{Z}_{2}\mathbb{Z}_{2}\mathbb{Z}_{2})=(2;000)$ reveal that the topological nature arises entirely from the Bi layers, which belong to a quantum spin Hall phase characterized by a $p_x-p_y$ model on a honeycomb lattice. The topological zigzag (ZZ) and armchair (AC) edge states are obtained. Using VASP2KP, the in-plane $g$ factors of these topological edge states are computed from the ab initio calculations: $g_{x/y}^{\mathrm{ZZ}}=2.07/1.60$ and $g_{x/y}^{\mathrm{AC}}=0.50/0.06$. The strong anisotropy of the edge-state $g$ factors allows us to explore Majorana zero modes in the Bi monolayer on a superconductor, which can be obtained by exfoliation or molecular beam epitaxy. The relaxed structures of the Bi2Ta3Se6, Bi2Nb3S6 and Bi2Nb3Se6 are obtained. Their superconducting transition temperature $T_c$ are estimated based on the electron-phonon coupling and the McMillan formula. Furthermore, using the experimental superconducting gap $Δ$ and the computed $g$ factors, we obtain the phase diagram, which shows that the in-plane field $B_y>2.62\mathrm{ T}$ can generate corner Majorana zero modes in the Bi monolayer of the superconductor Bi2Ta3S6. A similar paradigm also applies to the Bi2Ta3S6 bulk with the emergence of Majorana hinge states. These natural quantum-well superconductors therefore offer ideal platforms for exploring topological superconductivity and Majorana zero modes.

cond-mat.supr-con↗
arXiv2026

Beyond BeautifulSoup: Benchmarking LLM-Powered Web Scraping for Everyday Users

Web scraping has historically required technical expertise in HTML parsing, session management, and authentication circumvention, which limited large-scale data extraction to skilled developers. We argue that large language models (LLMs) have democratized web scraping, enabling low-skill users to execute sophisticated operations through simple natural language prompts. While extensive benchmarks evaluate these tools under optimal expert conditions, we show that without extensive manual effort, current LLM-based workflows allow novice users to scrape complex websites that would otherwise be inaccessible. We systematically benchmark what everyday users can do with off-the-shelf LLM tools across 35 sites spanning five security tiers, including authentication, anti-bot, and CAPTCHA controls. We devise and evaluate two distinct workflows: (a) LLM-assisted scripting, where users prompt LLMs to generate traditional scraping code but maintain manual execution control, and (b) end-to-end LLM agents, which autonomously navigate and extract data through integrated tool use. Our results demonstrate that end-to-end agents have made complex scraping accessible - requiring as little as a single prompt with minimal refinement (less than 5 changes) to complete workflows. We also highlight scenarios where LLM-assisted scripting may be simpler and faster for static sites. In light of these findings, we provide simple procedures for novices to use these workflows and gauge what adversaries could achieve using these.

cs.CR↗
arXiv2026

MiMo-V2-Flash Technical Report

We present MiMo-V2-Flash, a Mixture-of-Experts (MoE) model with 309B total parameters and 15B active parameters, designed for fast, strong reasoning and agentic capabilities. MiMo-V2-Flash adopts a hybrid attention architecture that interleaves Sliding Window Attention (SWA) with global attention, with a 128-token sliding window under a 5:1 hybrid ratio. The model is pre-trained on 27 trillion tokens with Multi-Token Prediction (MTP), employing a native 32k context length and subsequently extended to 256k. To efficiently scale post-training compute, MiMo-V2-Flash introduces a novel Multi-Teacher On-Policy Distillation (MOPD) paradigm. In this framework, domain-specialized teachers (e.g., trained via large-scale reinforcement learning) provide dense and token-level reward, enabling the student model to perfectly master teacher expertise. MiMo-V2-Flash rivals top-tier open-weight models such as DeepSeek-V3.2 and Kimi-K2, despite using only 1/2 and 1/3 of their total parameters, respectively. During inference, by repurposing MTP as a draft model for speculative decoding, MiMo-V2-Flash achieves up to 3.6 acceptance length and 2.6x decoding speedup with three MTP layers. We open-source both the model weights and the three-layer MTP weights to foster open research and community collaboration.

cs.CL↗
arXiv2025

MiMo-Audio: Audio Language Models are Few-Shot Learners

Existing audio language models typically rely on task-specific fine-tuning to accomplish particular audio tasks. In contrast, humans are able to generalize to new audio tasks with only a few examples or simple instructions. GPT-3 has shown that scaling next-token prediction pretraining enables strong generalization capabilities in text, and we believe this paradigm is equally applicable to the audio domain. By scaling MiMo-Audio's pretraining data to over one hundred million of hours, we observe the emergence of few-shot learning capabilities across a diverse set of audio tasks. We develop a systematic evaluation of these capabilities and find that MiMo-Audio-7B-Base achieves SOTA performance on both speech intelligence and audio understanding benchmarks among open-source models. Beyond standard metrics, MiMo-Audio-7B-Base generalizes to tasks absent from its training data, such as voice conversion, style transfer, and speech editing. MiMo-Audio-7B-Base also demonstrates powerful speech continuation capabilities, capable of generating highly realistic talk shows, recitations, livestreaming and debates. At the post-training stage, we curate a diverse instruction-tuning corpus and introduce thinking mechanisms into both audio understanding and generation. MiMo-Audio-7B-Instruct achieves open-source SOTA on audio understanding benchmarks (MMSU, MMAU, MMAR, MMAU-Pro), spoken dialogue benchmarks (Big Bench Audio, MultiChallenge Audio) and instruct-TTS evaluations, approaching or surpassing closed-source models. Model checkpoints and full evaluation suite are available at https://github.com/XiaomiMiMo/MiMo-Audio.

cs.CL↗