arXiv subjects

Yijun Wang

Publications and source records attributed to Yijun Wang.

At least 55 records · Page 3Linked to original sources

arXiv2025

Object-level Correlation for Few-Shot Segmentation

Few-shot semantic segmentation (FSS) aims to segment objects of novel categories in the query images given only a few annotated support samples. Existing methods primarily build the image-level correlation between the support target object and the entire query image. However, this correlation contains the hard pixel noise, \textit{i.e.}, irrelevant background objects, that is intractable to trace and suppress, leading to the overfitting of the background. To address the limitation of this correlation, we imitate the biological vision process to identify novel objects in the object-level information. Target identification in the general objects is more valid than in the entire image, especially in the low-data regime. Inspired by this, we design an Object-level Correlation Network (OCNet) by establishing the object-level correlation between the support target object and query general objects, which is mainly composed of the General Object Mining Module (GOMM) and Correlation Construction Module (CCM). Specifically, GOMM constructs the query general object feature by learning saliency and high-level similarity cues, where the general objects include the irrelevant background objects and the target foreground object. Then, CCM establishes the object-level correlation by allocating the target prototypes to match the general object feature. The generated object-level correlation can mine the query target feature and suppress the hard pixel noise for the final prediction. Extensive experiments on PASCAL-${5}^{i}$ and COCO-${20}^{i}$ show that our model achieves the state-of-the-art performance.

cs.CV↗

arXiv2025

Surpassing the PLOB bound in continuous-variable quantum secret sharing using a state-discrimination detector

Continuous-variable quantum secret sharing (CVQSS) is a promising approach to ensuring multi-party information security. While CVQSS offers practical ease of implementation, its present performance remains limited. In this paper, we propose a novel CVQSS protocol integrated with a state-discrimination detector (SDD), dubbed SDD-CVQSS. In particular, we first develop the detailed procedure of SDD-CVQSS, which replaces the traditional coherent detector with an SDD and eliminates the long-standing necessary step of establishing multiple point-to-point quantum key distribution links between all users and the dealer. We then elaborate on the principle of the specifically designed SDD, which can efficiently discriminate mixed states with a much lower error probability. Finally, we construct a security model for SDD-CVQSS and derive its security bound against beam-splitting collective attacks. Numerical simulations show that SDD-CVQSS outperforms conventional CVQSS in both maximum transmission distance and secret key rate, even surpassing the PLOB bound. Additionally, we find that the performance degradation of SDD-CVQSS in long-distance transmission scenarios can be effectively compensated for using a post-selection scheme, providing a feasible way to achieve high-performance CVQSS.

quant-ph↗

arXiv2025

Dome-DETR: DETR with Density-Oriented Feature-Query Manipulation for Efficient Tiny Object Detection

Tiny object detection plays a vital role in drone surveillance, remote sensing, and autonomous systems, enabling the identification of small targets across vast landscapes. However, existing methods suffer from inefficient feature leverage and high computational costs due to redundant feature processing and rigid query allocation. To address these challenges, we propose Dome-DETR, a novel framework with Density-Oriented Feature-Query Manipulation for Efficient Tiny Object Detection. To reduce feature redundancies, we introduce a lightweight Density-Focal Extractor (DeFE) to produce clustered compact foreground masks. Leveraging these masks, we incorporate Masked Window Attention Sparsification (MWAS) to focus computational resources on the most informative regions via sparse attention. Besides, we propose Progressive Adaptive Query Initialization (PAQI), which adaptively modulates query density across spatial areas for better query allocation. Extensive experiments demonstrate that Dome-DETR achieves state-of-the-art performance (+3.3 AP on AI-TOD-V2 and +2.5 AP on VisDrone) while maintaining low computational complexity and a compact model size. Code is available at https://github.com/RicePasteM/Dome-DETR.

cs.CV↗

arXiv2025

Enhancing Zero-Shot Brain Tumor Subtype Classification via Fine-Grained Patch-Text Alignment

The fine-grained classification of brain tumor subtypes from histopathological whole slide images is highly challenging due to subtle morphological variations and the scarcity of annotated data. Although vision-language models have enabled promising zero-shot classification, their ability to capture fine-grained pathological features remains limited, resulting in suboptimal subtype discrimination. To address these challenges, we propose the Fine-Grained Patch Alignment Network (FG-PAN), a novel zero-shot framework tailored for digital pathology. FG-PAN consists of two key modules: (1) a local feature refinement module that enhances patch-level visual features by modeling spatial relationships among representative patches, and (2) a fine-grained text description generation module that leverages large language models to produce pathology-aware, class-specific semantic prototypes. By aligning refined visual features with LLM-generated fine-grained descriptions, FG-PAN effectively increases class separability in both visual and semantic spaces. Extensive experiments on multiple public pathology datasets, including EBRAINS and TCGA, demonstrate that FG-PAN achieves state-of-the-art performance and robust generalization in zero-shot brain tumor subtype classification.

cs.CV↗

arXiv2025

Noema formIng Cluster survEy (NICE): A Census of Star Formation and Cold Gas Properties in Massive protoclusters at 1.5<z<4

Massive protoclusters at z~1.5-4, the peak of the cosmic star formation history, are key to understanding the formation mechanisms of massive galaxies in today's clusters. However, studies of protoclusters at these high redshifts remain limited, primarily due to small sample sizes and heterogeneous selection criteria. In this work, we conduct a systematic investigation of the star formation and cold gas properties of member galaxies of eight massive protoclusters in the COSMOS field, using the statistical and homogeneously selected sample from the Noema formIng Cluster survEy (NICE). Our analysis reveals a steep increase in the star formation rates per halo mass ($Σ_{\rm SFR} /M_{\rm halo}$) with redshifts in these intensively star-forming protoclusters, reaching values one to two orders of magnitude higher than those observed in the field at z>2. We further show that, instead of an enhancement of starbursts, this increase is largely driven by the concentration of massive and gas-rich star-forming galaxies in the protocluster cores. The member galaxies still generally follow the same star formation main sequence as in the field, with a moderate enhancement at the low mass end. Notably, the most massive protocluster galaxies ($M_\star$>8$\times$10$^{10}$M$_\odot$) exhibit higher $f_{\rm gas}$ and $τ_{\rm gas}$ than their field counterparts, while remaining on the star forming main sequence. These gas-rich, massive, and star-forming galaxies are predominantly concentrated in the protocluster cores and are likely progenitors of massive ellipticals in the center of today's clusters. These results suggest that the formation of massive galaxies in such environments is sustained by substantial gas reservoirs, which support persistent star formation and drive early mass assembly in forming cluster cores.

astro-ph.GA↗

arXiv2025

High-Density EEG Enables the Fastest Visual Brain-Computer Interfaces

Brain-computer interface (BCI) technology establishes a direct communication pathway between the brain and external devices. Current visual BCI systems suffer from insufficient information transfer rates (ITRs) for practical use. Spatial information, a critical component of visual perception, remains underexploited in existing systems because the limited spatial resolution of recording methods hinders the capture of the rich spatiotemporal dynamics of brain signals. This study proposed a frequency-phase-space fusion encoding method, integrated with 256-channel high-density electroencephalogram (EEG) recordings, to develop high-speed BCI systems. In the classical frequency-phase encoding 40-target BCI paradigm, the 256-66, 128-32, and 64-21 electrode configurations brought theoretical ITR increases of 83.66%, 79.99%, and 55.50% over the traditional 64-9 setup. In the proposed frequency-phase-space encoding 200-target BCI paradigm, these increases climbed to 195.56%, 153.08%, and 103.07%. The online BCI system achieved an average actual ITR of 472.7 bpm. This study demonstrates the essential role and immense potential of high-density EEG in decoding the spatiotemporal information of visual stimuli.

cs.HC↗

arXiv2025

JWST/MIRI reveals the true number density of massive galaxies in the early Universe

Early JWST studies reporting an unexpected abundance of massive galaxies at $z \sim 5$--$8$ challenge galaxy formation models in the $Λ$CDM framework. Previous stellar mass ($M_\star$) estimates suffered from large uncertainties due to the lack of rest-frame near-infrared data. Using deep JWST/NIRCam and MIRI photometry from PRIMER, we systematically analyze massive galaxies at $z \sim 3$--$8$, leveraging rest-frame $\gtrsim 1\,μ$m constraints. We find MIRI is critical for robust $M_\star$ measurements for massive galaxies at $z > 5$: excluding MIRI overestimates $M_\star$ by $\sim 0.4$ dex on average for $M_\star > 10^{10}\,M_\odot$ galaxies, with no significant effects at lower masses. This reduces number densities of $M_\star > 10^{10}\,M_\odot$ ($10^{10.3}\,M_\odot$) galaxies by $\sim 36\%$ ($55\%$). MIRI inclusion also reduces ``Little Red Dot'' (LRD) contamination in massive galaxy samples, lowering the LRD fraction from $\sim 32\%$ to $\sim 13\%$ at $M_\star > 10^{10.3}\,M_\odot$. Assuming pure stellar origins, LRDs exhibit $M_\star \sim 10^{9\text{--}10.5}\,M_\odot$ with MIRI constraints, rarely exceeding $10^{10.5}\,M_\odot$. Within standard $Λ$CDM, our results indicate a moderate increase in the baryon-to-star conversion efficiency ($ε$) toward higher redshifts and masses at $z > 3$. For the most massive $z \sim 8$ galaxies, $ε\sim 0.3$, compared to $ε\lesssim 0.2$ for typical galaxies at $z < 3$. This result is consistent with models where high gas densities and short free-fall times suppress stellar feedback in massive high-$z$ halos.

astro-ph.GA↗

arXiv2025

A Fast Bayesian Method for Coherent Gravitational Wave Searches with Relative Astrometry

Using relative stellar astrometry for the detection of coherent gravitational wave sources is a promising method for the microhertz range, where no dedicated detectors currently exist. Compared to other gravitational wave detection techniques, astrometry operates in an extreme high-baseline-number and low-SNR-per-baseline limit, which leads to computational difficulties when using conventional Bayesian search techniques. We extend a technique for efficiently searching pulsar timing array datasets through the precomputation of inner products in the Bayesian likelihood, showing that it is applicable to astrometric datasets. Using this technique, we are able to reduce the total dataset size by up to a factor of $\mathcal{O}(100)$, while remaining accurate to within 1% over two orders of magnitude in gravitational wave frequency. Applying this technique to simulated astrometric datasets for the Kepler Space Telescope and Nancy Grace Roman Space Telescope missions, we obtain forecasts for the sensitivity of these missions to coherent gravitational waves. Due to the low angular sky coverage of astrometric baselines, we find that coherent gravitational wave sources are poorly localized on the sky. Despite this, from $10^{-8}$ Hz to $10^{-6}$ Hz, we find that Roman is sensitive to coherent gravitational waves with an instantaneous strain above $h_0 \simeq 10^{-11.4}$, and Kepler is sensitive to strains above $h_0 \simeq $ $10^{-12.4}$. At this strain, we can detect a source with a frequency of $10^{-7}$ Hz and a chirp mass of $10^9$ $M_\odot$ at a luminosity distance of 3.6 Mpc for Kepler, and 0.3 Mpc for Roman.

astro-ph.IM↗

arXiv2025

RemoteSAM: Towards Segment Anything for Earth Observation

We aim to develop a robust yet flexible visual foundation model for Earth observation. It should possess strong capabilities in recognizing and localizing diverse visual targets while providing compatibility with various input-output interfaces required across different task scenarios. Current systems cannot meet these requirements, as they typically utilize task-specific architecture trained on narrow data domains with limited semantic coverage. Our study addresses these limitations from two aspects: data and modeling. We first introduce an automatic data engine that enjoys significantly better scalability compared to previous human annotation or rule-based approaches. It has enabled us to create the largest dataset of its kind to date, comprising 270K image-text-mask triplets covering an unprecedented range of diverse semantic categories and attribute specifications. Based on this data foundation, we further propose a task unification paradigm that centers around referring expression segmentation. It effectively handles a wide range of vision-centric perception tasks, including classification, detection, segmentation, grounding, etc, using a single model without any task-specific heads. Combining these innovations on data and modeling, we present RemoteSAM, a foundation model that establishes new SoTA on several earth observation perception benchmarks, outperforming other foundation models such as Falcon, GeoChat, and LHRS-Bot with significantly higher efficiency. Models and data are publicly available at https://github.com/1e12Leon/RemoteSAM.

cs.CV↗

arXiv2025

Restoring Real-World Images with an Internal Detail Enhancement Diffusion Model

Restoring real-world degraded images, such as old photographs or low-resolution images, presents a significant challenge due to the complex, mixed degradations they exhibit, such as scratches, color fading, and noise. Recent data-driven approaches have struggled with two main challenges: achieving high-fidelity restoration and providing object-level control over colorization. While diffusion models have shown promise in generating high-quality images with specific controls, they often fail to fully preserve image details during restoration. In this work, we propose an internal detail-preserving diffusion model for high-fidelity restoration of real-world degraded images. Our method utilizes a pre-trained Stable Diffusion model as a generative prior, eliminating the need to train a model from scratch. Central to our approach is the Internal Image Detail Enhancement (IIDE) technique, which directs the diffusion model to preserve essential structural and textural information while mitigating degradation effects. The process starts by mapping the input image into a latent space, where we inject the diffusion denoising process with degradation operations that simulate the effects of various degradation factors. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art models in both qualitative assessments and perceptual quantitative evaluations. Additionally, our approach supports text-guided restoration, enabling object-level colorization control that mimics the expertise of professional photo editing.

cs.CV↗

arXiv2025

HARDMath2: A Benchmark for Applied Mathematics Built by Students as Part of a Graduate Class

Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present HARDMath2, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.

cs.LG↗

arXiv2025

MedGUIDE: Benchmarking Clinical Decision-Making in Large Language Models

Clinical guidelines, typically structured as decision trees, are central to evidence-based medical practice and critical for ensuring safe and accurate diagnostic decision-making. However, it remains unclear whether Large Language Models (LLMs) can reliably follow such structured protocols. In this work, we introduce MedGUIDE, a new benchmark for evaluating LLMs on their ability to make guideline-consistent clinical decisions. MedGUIDE is constructed from 55 curated NCCN decision trees across 17 cancer types and uses clinical scenarios generated by LLMs to create a large pool of multiple-choice diagnostic questions. We apply a two-stage quality selection process, combining expert-labeled reward models and LLM-as-a-judge ensembles across ten clinical and linguistic criteria, to select 7,747 high-quality samples. We evaluate 25 LLMs spanning general-purpose, open-source, and medically specialized models, and find that even domain-specific LLMs often underperform on tasks requiring structured guideline adherence. We also test whether performance can be improved via in-context guideline inclusion or continued pretraining. Our findings underscore the importance of MedGUIDE in assessing whether LLMs can operate safely within the procedural frameworks expected in real-world clinical settings.

cs.CL↗

arXiv2025

B-STaR: Monitoring and Balancing Exploration and Exploitation in Self-Taught Reasoners

In the absence of extensive human-annotated data for complex reasoning tasks, self-improvement -- where models are trained on their own outputs -- has emerged as a primary method for enhancing performance. However, the critical factors underlying the mechanism of these iterative self-improving methods remain poorly understood, such as under what conditions self-improvement is effective, and what are the bottlenecks in the current iterations. In this work, we identify and propose methods to monitor two pivotal factors in this iterative process: (1) the model's ability to generate sufficiently diverse responses (exploration); and (2) the effectiveness of external rewards in distinguishing high-quality candidates from lower-quality ones (exploitation). Using mathematical reasoning as a case study, we begin with a quantitative analysis to track the dynamics of exploration and exploitation, discovering that a model's exploratory capabilities rapidly deteriorate over iterations, and the effectiveness of exploiting external rewards diminishes as well. Motivated by these findings, we introduce B-STaR, a Self-Taught Reasoning framework that autonomously adjusts configurations across iterations to Balance exploration and exploitation, thereby optimizing the self-improving effectiveness based on the current policy model and available rewards. Our experiments on mathematical reasoning, coding, and commonsense reasoning demonstrate that B-STaR not only enhances the model's exploratory capabilities throughout training but also achieves a more effective balance between exploration and exploitation, leading to superior performance.

cs.AI↗

arXiv2025

SelaFD:Seamless Adaptation of Vision Transformer Fine-tuning for Radar-based Human Activity Recognition

Human Activity Recognition (HAR) such as fall detection has become increasingly critical due to the aging population, necessitating effective monitoring systems to prevent serious injuries and fatalities associated with falls. This study focuses on fine-tuning the Vision Transformer (ViT) model specifically for HAR using radar-based Time-Doppler signatures. Unlike traditional image datasets, these signals present unique challenges due to their non-visual nature and the high degree of similarity among various activities. Directly fine-tuning the ViT with all parameters proves suboptimal for this application. To address this challenge, we propose a novel approach that employs Low-Rank Adaptation (LoRA) fine-tuning in the weight space to facilitate knowledge transfer from pre-trained ViT models. Additionally, to extract fine-grained features, we enhance feature representation through the integration of a serial-parallel adapter in the feature space. Our innovative joint fine-tuning method, tailored for radar-based Time-Doppler signatures, significantly improves HAR accuracy, surpassing existing state-of-the-art methodologies in this domain. Our code is released at https://github.com/wangyijunlyy/SelaFD.

cs.CV↗

arXiv2025

First Detection of Radio Emission from the Intermediate Mass Black Hole in POX 52: Deep Multi-Band Observations with ATCA and VLA

We present the first multi-band centimeter detection of POX 52, a nearby dwarf galaxy believed to habor a robust intermediate mass black hole (IMBH). We conducted the deep observations using the Australia Telescope Compact Array (ATCA), spanning frequencies from 4.5 to 10 GHz, as well as the sensitive observations from the Karl G. Jansky Very Large Array (VLA) operating in its most extended A-configuration at S band (2--4 GHz) and C band (4--8 GHz). In the ATCA observations, the source shows a compact morphology, with only one direction marginally resolved. The higher resolution of the VLA allowed us to slightly resolve the source, fitting it well with a two-dimensional Gaussian model. The detected radio emission confirms the presence of Active Galactic Nucleus (AGN) activity, indicating either a low-power jet or AGN-driven winds/outflows. Our dual-epoch observations with ATCA and VLA, together with previous non-detection flux density upper limits, reveal radio emission variability spanning two decades. In addition, we find that POX 52 aligns well with the low-mass extension of the fundamental plane for high-accretion, radio-quiet massive AGNs.

astro-ph.GA↗

arXiv2024

Minor-merger induced star formation rejuvenation in an elliptical radio-loud quasar host, 3C 59

We report a rare case where an elliptical radio-loud quasar host, 3C 59, rejuvenates star formation activity through minor mergers with its nearby satellite galaxies. The inferred star formation history of 3C 59 shows significant star formation rejuvenation within the past 500 Myr, before which remains rather quiescent for most of the cosmic time. Three nearest satellite galaxies of 3C 59 exhibit significant morphological disturbances, and two of them present strong tidal tails pointing towards 3C 59. In addition, all the satellite galaxies within a projected distance of 200 kpc show low star formation activities. They also have systematically lower effective radius ($R_{\rm e}$) than local late-type galaxies, while 3C 59 has significantly larger $R_{\rm e}$ than both early- and late-type galaxies. All these features suggest that ongoing minor mergers between 3C 59 and its nearby satellites could be causing gas to flow into 3C 59, which induces the star formation rejuvenation, and possibly also triggers the quasar activity. The enormous power from the large-scale radio jet of 3C 59 may in turn help keep the halo hot, prevent gas cooling, and further reduce star formation in its satellite galaxies. These results provide important insights into the mass and size growth of central galaxies and star formation quenching of satellite galaxies in galaxy groups.

astro-ph.GA↗

arXiv2024

An empirical model of the extragalactic radio background

Radio observations provide a powerful tool to constrain the assembly of galaxies over cosmic time. Recent deep and wide radio continuum surveys have improved significantly our understanding on radio emission properties of AGNs and SFGs across $0 < z < 4$. This allows us to derive an empirical model of the radio continuum emission of galaxies based on their SFR and the probability of hosting an radio AGN. We make use of the Empirical Galaxy Generator (EGG) to generate a near-infrared-selected, flux-limited multi-wavelength catalog to mimic real observations. Then we assign radio continuum flux densities to galaxies based on their SFRs and the probability of hosting a radio-AGN of specific 1.4 GHz luminosity. We also apply special treatments to reproduce the clustering signal of radio AGNs.Our empirical model successfully recovers the observed 1.4 GHz radio luminosity functions (RLFs) of both AGN and SFG populations, as well as the differential number counts at various radio bands. The uniqueness of this approach also allows us to directly link radio flux densities of galaxies to other properties, including redshifts, stellar masses, and magnitudes at various photometric bands. We find that roughly half of the radio continuum sources to be detected by SKA at $z \sim 4-6$ will be too faint to be detected in the optical survey ($r \sim 27.5$) carried out by Rubin observatory. Unlike previous studies which utilized RLFs to reproduce ERB, our work starts from a simulated galaxy catalog with realistic physical properties. It has the potential to simultaneously, and self-consistently reproduce physical properties of galaxies across a wide range of wavelengths, from optical, NIR, FIR to radio wavelengths. Our empirical model can shed light on the contribution of different galaxies to the extragalactic background light, and greatly facilitates designing future multiwavelength galaxy surveys.

astro-ph.GA↗

arXiv2024

Learning to Denoise Biomedical Knowledge Graph for Robust Molecular Interaction Prediction

Molecular interaction prediction plays a crucial role in forecasting unknown interactions between molecules, such as drug-target interaction (DTI) and drug-drug interaction (DDI), which are essential in the field of drug discovery and therapeutics. Although previous prediction methods have yielded promising results by leveraging the rich semantics and topological structure of biomedical knowledge graphs (KGs), they have primarily focused on enhancing predictive performance without addressing the presence of inevitable noise and inconsistent semantics. This limitation has hindered the advancement of KG-based prediction methods. To address this limitation, we propose BioKDN (Biomedical Knowledge Graph Denoising Network) for robust molecular interaction prediction. BioKDN refines the reliable structure of local subgraphs by denoising noisy links in a learnable manner, providing a general module for extracting task-relevant interactions. To enhance the reliability of the refined structure, BioKDN maintains consistent and robust semantics by smoothing relations around the target interaction. By maximizing the mutual information between reliable structure and smoothed relations, BioKDN emphasizes informative semantics to enable precise predictions. Experimental results on real-world datasets show that BioKDN surpasses state-of-the-art models in DTI and DDI prediction tasks, confirming the effectiveness and robustness of BioKDN in denoising unreliable interactions within contaminated KGs

cs.AI↗