Searcharxiv⌕ Search

arXiv subjects

Shu Wang

Publications and source records attributed to Shu Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

A Bayesian Finite Mixture Model Approach for Mixed-type Data Clustering and Variable Selection with Censored Biomarkers

Clustering mixed-type data remains a major challenge in biomedical research to uncover clinically meaningful subgroups within heterogeneous patient populations. Most existing clustering methods impose restrictive assumptions like local independence, fail to accommodate censored biomarkers, or unable to quantify variable importance. We propose a Bayesian finite mixture model (BFMM) clustering framework that addresses these limitations. BFMM flexibly models both continuous and categorical variables, incorporates three covariance structures to capture cluster-specific dependencies among continuous features, and handles censored observations through likelihood-based imputation. To facilitate feature prioritization, BFMM uses spike-and-slab priors to estimate variable importance on a continuous 0-1 scale. Simulation studies demonstrate that BFMM outperforms existing methods in clustering accuracy, particularly given strong within-cluster correlation or censored variables, and reliably distinguishes informative features from noise under varying conditions. We applied BFMM to two real-world datasets: (1) the SENECA cohort integrating electronic health records from patients with Sepsis; and (2) the EDEN randomized trial of patients with acute lung injury. In both settings, BFMM identified clinically interpretable phenotypes and revealed variable-specific contributions to subgroup differentiation. In the EDEN trial, it also uncovered evidence of treatment heterogeneity. These findings validate BFMM as an effective, interpretable, and practically useful clustering tool for complex biomedical datasets.

stat.AP↗
arXiv2026

Post-perihelion Coma Composition of the Interstellar Comet 3I/ATLAS from Optical Spectroscopy

We present multi-epoch optical spectroscopy of the interstellar comet 3I/ATLAS obtained between December 2025 and January 2026 (heliocentric distances 1.8-3.3 au), yielding post-perihelion production rates and mixing ratios for CN, C$_3$, C$_2$, CH, and gaseous metals (Fe I and Ni I). Our results show that the coma is less depleted in C$_2$ after perihelion than before, indicative of subsurface activation or compositional heterogeneity. The outgassing profiles reveal a pronounced perihelion asymmetry: CN and metal production rates decline more gradually outbound than inbound, consistent with the reported behavior of H$_2$O and implying a change in the comet's activity pattern across perihelion. Despite being metal-rich relative to its H$_2$O content, 3I follows the metal-CO correlation observed in comets of diverse origins, suggesting that gaseous metal release is more closely linked to a CO-bearing volatile reservoir than to H$_2$O, potentially in the form of metal carbonyls. In addition, the [O I] $\lambda6300$ emission shows a significant residual after subtracting the expected contributions from H$_2$O, CO$_2$, and CO, which may reflect systematic uncertainties in the photodissociation yields of those molecules or a contribution from additional oxygen-bearing parents.

astro-ph.EP↗
arXiv2026

A Planetary Nebula from a 5.7 $M_{\odot}$ Progenitor in a 90 Myr M31 Star Cluster

Planetary nebulae (PNe) trace the late evolution of low-to-intermediate-mass stars, yet the masses of their progenitors are rarely measured directly. Here we present a PN physically associated with a young star cluster in M31, providing an unprecedented extragalactic empirical anchor in the poorly constrained high-mass regime of PN progenitors. High-resolution Hubble Space Telescope imaging shows that the nebula lies near the cluster center, and spectral decomposition of the blended cluster-plus-nebula spectrum yields consistent stellar and nebular radial velocities, strongly supporting a physical association. Isochrone fitting to the color-magnitude diagram indicates a cluster age of ~90 Myr and a near-solar metallicity, implying a progenitor initial mass of $5.66^{+0.42}_{-0.37}\,M_{\odot}$. This value is among the highest empirical progenitor-mass constraints yet reported for any PN and approaches the lower boundary of the super-asymptotic giant branch (super-AGB) regime. We further find that the nebula is strongly nitrogen-enhanced, with an N/O ratio ~7 times the solar value, broadly consistent with hot bottom burning in a relatively massive AGB progenitor. This system therefore provides a rare opportunity to test PN formation and nucleosynthesis at the high-mass end of the PN progenitor distribution.

astro-ph.SR↗
arXiv2026

MT-EditFlow: Reinforcement Learning for Multi-Turn Image Editing with Flow Matching

Recent breakthroughs in instruction-based image editing have captured significant attention, as models are now capable of handling real-world editing demands with the practicality required by everyday users. However, editing models trained primarily for single-turn edits often break down in multi-turn editing--the natural interactive setting where a user iteratively refines an image based on the model's own previous outputs. This failure stems from the all-or-nothing requirement, where a single failed turn compromises the entire sequence, and error propagation, where exposure bias leads to compounding editing errors. To address these challenges, we introduce MT-EditFlow, a flow-matching reinforcement learning framework designed to optimize reward signals for sequential image editing. MT-EditFlow integrates a multi-turn perspective with a multi-reward formulation to provide a unified structure applicable to both GRPO and NFT-based reinforcement learning methods. We systematically analyze and optimize the reward signal by investigating effective scoring strategies for turn-level aggregation, VLM reasoning modes to trade off reward bias and variance, and advantage fusion levels to prevent reward hacking. Our findings reveal that broadcasting the aggregated advantage across the entire editing trajectory effectively bridges the gap between local planning and global multi-turn task success. Extensive experiments demonstrate that MT-EditFlow significantly improves performance across diverse base models. Notably, it boosts FLUX.1-Kontext-dev by 6.85 points in turn-3 overall performance, surpassing state-of-the-art open-source models such as Qwen-Image-Edit. By maintaining high marginal success rates and reducing exposure bias, MT-EditFlow provides a foundation for more reliable and natural human-AI collaboration in visual content creation.

cs.CV↗
arXiv2026

New black hole mass calibrations and the fundamental plane of the broad-line region size, luminosity, and velocity

We present a new calibration of the broad-line region (BLR) size-luminosity-velocity relation using a sample of 157 AGNs with reliable Hbeta time-delay (\lag) measurements from Wang & Woo 2024. By incorporating the Eddington ratio as a third parameter, we effectively correct the systematic offset of high-Eddington AGNs in the traditional BLR size-luminosity relation. The resulting three-parameter fit defines a fundamental plane in the 3-D space of the \lag, optical luminosity, and Hbeta velocity, with an intrinsic scatter of 0.21 dex. This tight correlation reflects the coupled effects of gas kinematics, photoionization, and BLR geometry. In turn, we develop a new method to infer \lag\ from the combination of optical luminosity and Hbeta velocity, and derive single-epoch black hole mass estimators by adopting either the full-width-at-half-maximum (FWHM) or line dispersion ($σ$) of the Hbeta line profile as the velocity indicator. The derived \lag shows a ~0.1 dex scatter, depending on the choice of calibrations. We show that the previous mass estimates based on the two-parameter size-luminosity relation with a 0.5 slope can be overestimated by up to 0.5 dex, demonstrating that the new mass estimator substantially changes the cosmic black hole mass density and the growth of black hole seeds in the early universe.

astro-ph.GA↗
arXiv2026

Mean Field Competition of Optimal Switching: The Vanishing Entropy Regularization Approach

This paper studies a type of rank-based mean field game in which competing agents strategically switch among multiple effort regimes. We propose an entropy regularized auxiliary problem where the switching decisions are randomized to the control of transition probability for a continuous-time finite-state Markov chain. We first establish the existence of regularized equilibrium in this auxiliary problem. Assuming the convexity of reward scheme, we then prove that the equilibrium is unique and can be approximated by a fictitious play iteration scheme. Furthermore, as the entropy regularization vanishes, we establish the convergence analysis of the regularized equilibrium towards the relaxed equilibrium in the original MFG of optimal switching. The uniqueness of the population ranking distribution under the relaxed equilibrium is also obtained given a strictly convex reward scheme.

math.OC↗
arXiv2026

Implementing the principal stratum strategy for intercurrent events with survival outcomes: a tutorial

The International Council for Harmonization (ICH) E9 (R1) addendum provides the estimand framework to formulate treatment effects in a clinical trial. One of the attributes of an estimand the framework describes is intercurrent events. Among the five strategies to intercurrent events the guidance lists, the principal stratum strategy is the most conceptually and technically challenging because it defines treatment effects on unobserved strata. Its application to survival outcomes is particularly inaccessible to practitioners. This tutorial reviews the methodology and implementation of the estimand framework with the principal stratum strategy to address intercurrent events with survival outcomes. We illustrate using a clinical trial in oncology and focus on a simple case with binary treatment and a single binary intercurrent event of discontinuation of the assigned treatment. We define the causal effects and review two main methods for estimating the effects: the mixture model method and the weighting method. For each method, we elaborate the associated assumptions, models, sensitivity analysis, software and provide example R code. We conduct simulation studies that mimic the real study to study the operation characteristics of these methods.

stat.ME↗
arXiv2026

DBES: A Systematic Benchmark and Metric Suite for Evaluating Expert Specialization in Large-Scale MoEs

Expert specialization in Mixture-of-Experts (MoE) models remains poorly understood, with traditional evaluations conflating architectural load-balancing with functional specialization. We introduce DBES, a comprehensive diagnostic framework combining a multi-domain benchmark with five theoretically grounded metrics: Routing Specialization, Normalized Effective Rank, Domain Isolation, Routing Stiffness Score, and N-gram Expertise measures. Critical findings demonstrate distinct specialization paradigms across models: Qwen-series exhibit modular specialization with high domain isolation, while DeepSeek and GLM employ distributed collaboration. However, we emphasize that specialization is a diagnostic dimension, necessary but not sufficient for downstream performance. Most crucially, interventional evidence validates the actionability of these metrics: by using DBES to identify high-specialization expert paths during domain-specific post-training, we achieved 66% to 94.48% improvement in specialized domains with only 15% of original training resources, demonstrating that these diagnostic tools can be converted into concrete optimization operators. This work provides the first systematic methodology for evaluating expert specialization independently of accuracy metrics, offering crucial insights for the design and post-training optimization of next-generation MoE systems.

cs.LG↗
arXiv2026

SkillRAE: Agent Skill-Based Context Compilation for Retrieval-Augmented Execution

Large Language Model (LLM)-based agents (e.g., OpenClaw) increasingly rely on reusable skill libraries to solve artifact-rich tasks such as document-centric workflows and data-intensive analysis. As these libraries grow, a few works have attempted to study the Retrieval-Augmented Execution (RAE), which often first retrieves some external skills and other knowledge, then compiles the context using retrieved skills, and finally executes the task. Existing works mainly focus on optimizing skill retrieval and task execution, and they pay little attention to how to effectively organize the selected skill evidence in a form that is compact, grounded, and immediately usable for the downstream executors to complete tasks. To fill this gap, we propose SkillRAE, a two-stage RAE approach focusing on skill-based context compilation, which consists of the offline and online stages. Specifically, in the offline indexing stage, it builds a multi-level skill graph over skill communities, skills, and reusable subunits, for capturing their relationships. In the online retrieval stage, it first performs skill-ranked retrieval with selected-subunit evidence export in the graph, and then applies rescue-aware compact compilation to recover the key evidence. Together, these components compile a coarse-ranked skill set into a task-specific context that is compact, grounded, and immediately usable. Experiments on two public benchmarks show that SkillRAE achieves a significant improvement over baselines for RAE. For example, on SkillsBench, it achieves an improvement of 11.7% over the SOTA method. Ablation studies further show that our context compilation is crucial, instead of a mere prompt addition.

cs.CL↗
arXiv2026

ASTRA-QA: A Benchmark for Abstract Question Answering over Documents

Document-based question answering (QA) increasingly includes abstract questions that require synthesizing scattered information from long documents or across multiple documents into coherent answers. However, this setting is still poorly supported by existing benchmarks and evaluation methods, which often lack stable abstract references or rely on coarse similarity metrics and unstable head-to-head comparisons. To alleviate this issue, we introduce ASTRA-QA, a benchmark for AbSTRAct Question Answering over documents. ASTRA-QA contains 869 QA instances over academic papers and news documents, covering five abstract question types and three controlled retrieval scopes. Each instance is equipped with explicit evaluation annotations, including answer topic sets, curated unsupported topics, and aligned evidence. Building on these annotations, ASTRA-QA assesses whether answers cover required key points and avoid unsupported content by directly scoring topic coverage and curated unsupported content, enabling scalable evaluation without exhaustive head-to-head comparisons. Experiments with representative Retrieval-Augmented Generation (RAG) methods spanning vanilla, graph-based, and hierarchical retrieval settings show that ASTRA-QA provides reference-grounded diagnostics for coverage, hallucination, and retrieval-scope robustness. Our dataset and code are available at https://xinyangsally.github.io/astra-benchmark.

cs.CL↗
arXiv2026

ArchRAG: Attributed Community-based Hierarchical Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) has proven effective in integrating external knowledge into large language models (LLMs) for solving question-answer (QA) tasks. The state-of-the-art RAG approaches often use the graph data as the external data since they capture the rich semantic information and link relationships between entities. However, existing graph-based RAG approaches cannot accurately identify the relevant information from the graph and also consume large numbers of tokens in the online retrieval process. To address these issues, we introduce a novel graph-based RAG approach, called Attributed Community-based Hierarchical RAG (ArchRAG), by augmenting the question using attributed communities, and also introducing a novel LLM-based hierarchical clustering method. To retrieve the most relevant information from the graph for the question, we build a novel hierarchical index structure for the attributed communities and develop an effective online retrieval method. Experimental results demonstrate that ArchRAG outperforms existing methods in both accuracy and token cost.

cs.IR↗
arXiv2026

A Search for High Frequency Oscillations in TESS Cycle 7

High-quality, short-cadence photometry from TESS enables the detection of rapid oscillators with unprecedented sensitivity. In this work, we conduct a homogeneous search for high-frequency variability using 20-second cadence light curves from TESS Cycle 7 (Sectors 84--96). From $\sim 3.9\times10^{4}$ light curves, we compute Lomb-Scargle periodograms and select candidates exhibiting at least one significant signal with $\mathrm{FAP}\le 10^{-4}$ at frequencies $f\ge 50~\mathrm{d^{-1}}$. After excluding previously reported objects and performing pixel-level and light-curve vetting to mitigate contamination, we identify 73 rapid oscillators, including 24 pulsating white dwarfs, 31 hot subdwarfs, and 18 A-F stars. Using an iterative prewhitening procedure, we carry out a detailed frequency analysis for each target and derive the oscillation frequencies and amplitudes. We further investigate the physical origins of the detected frequency content and present statistical characterizations of the rapid-oscillator sample. We highlight one white dwarf and one subdwarf that exhibit clear frequency multiplets consistent with rotational splitting. This work enlarges the sample of rapid oscillators accessible with TESS data and provides a uniformly measured frequency-amplitude catalog, establishing a consistent basis for future asteroseismic and population studies.

astro-ph.SR↗
arXiv2026

A Search for Variability of Ultracool Dwarfs with the Zwicky Transient Facility

Rotationally modulated photometric variability of ultracool dwarfs encodes key information about cloud structure and temperature contrasts. Large homogeneous optical datasets are crucial for linking atmospheric heterogeneity to fundamental parameters such as rotation, mass, and age. We present a search for rotation periods in ultracool dwarfs using Zwicky Transient Facility (ZTF) optical light curves. By propagating the coordinates to the ZTF epoch and applying Lomb-Scargle analysis, we identified 226 periodic variables, including 32 robust detections and 194 tentative cases. Among the robust detections, 12 have no previously published periods, while 20 have literature counterparts, most of which are consistent with the published values. Most robust detections are M dwarfs, reflecting the optical sensitivity limits of ZTF. We find a trend of decreasing periods toward later spectral types in relatively old dwarfs (> 100 Myr), suggesting faster rotation for late-M types than for mid-M types. The age-period relation of our sample is broadly consistent with angular-momentum-conservation models at higher-mass regime of brown dwarfs, consistent with the M-dwarf bias of our catalog. Many additional candidates remain to be confirmed due to sparse sampling or low S/N. Future high-cadence, multi-wavelength monitoring and systematic mining of ZTF and upcoming surveys will be crucial for validating these periods, extend sensitivity to later (L/T) types, and better connect rotation with cloud physics across the stellar-substellar boundary.

astro-ph.SR↗
arXiv2026

In-depth Analysis of Graph-based RAG in a Unified Framework

Graph-based Retrieval-Augmented Generation (RAG) has proven effective in integrating external knowledge into large language models (LLMs), improving their factual accuracy, adaptability, interpretability, and trustworthiness. A number of graph-based RAG methods have been proposed in the literature. However, these methods have not been systematically and comprehensively compared under the same experimental settings. In this paper, we first summarize a unified framework to incorporate all graph-based RAG methods from a high-level perspective. We then extensively compare representative graph-based RAG methods over a range of questing-answering (QA) datasets -- from specific questions to abstract questions -- and examine the effectiveness of all methods, providing a thorough analysis of graph-based RAG approaches. As a byproduct of our experimental analysis, we are also able to identify new variants of the graph-based RAG methods over specific QA and abstract QA tasks respectively, by combining existing techniques, which outperform the state-of-the-art methods. Finally, based on these findings, we offer promising research opportunities. We believe that a deeper understanding of the behavior of existing methods can provide new valuable insights for future research.

cs.IR↗
arXiv2026

StarCLR: Contrastive Learning Representation for Astronomical Light Curves

With the rapid development of time-domain surveys, the availability of massive light curve data offers new opportunities for studying stellar evolution and variable star classification, while simultaneously posing challenges for feature extraction and modeling. We present StarCLR, a contrastive pretraining framework for large-scale light curves. By constructing positive pairs from partially overlapping sub-sequences, StarCLR encourages the model to learn temporal representations. We pretrain StarCLR on the TESS dataset and fine-tune it for variable star classification on three surveys with distinct observational characteristics, namely TESS (18 types), ZTF (11 types), and Gaia (24 types). StarCLR achieves macro-F1 scores of 84.35%, 87.82%, and 92.73%, and micro-F1 scores of 94.46%, 92.83%, and 99.49%, respectively. Compared with LSTM and Transformer trained from scratch, StarCLR performs better on TESS and ZTF, with the largest gains on sparsely sampled ZTF light curves, demonstrating promising generalization. For Gaia, which involves a broader class space, the evaluation is not directly comparable, and performance is likely influenced by astrophysical features, resulting in a more limited contribution from the pretrained backbone. Systematic ablations on embedding design, pooling strategy, and pretraining settings further indicate that the pretrained representations provide performance gains by capturing informative temporal characteristics of light curves. Looking ahead, with standardized datasets and more diverse labeling schemes, the generalization ability of StarCLR can be further enhanced.

astro-ph.SR↗
arXiv2026

Dynamical Modeling of the Broad-Line Region with High-Mass Active Galactic Nuclei and Constraints on the Virial Factor

We present the results of broad-line region (BLR) dynamical modeling for eight high-mass active galactic nuclei (AGNs) from the Seoul National University AGN Monitoring Project, by constraining BLR geometry and kinematics as well as black hole (BH) mass ($M_{\rm BH}$). We find that the H$β$-emitting BLRs are best described as thick disks viewed at intermediate inclinations, with emission preferentially originating from the far side of the BLR. BLR kinematics show a combination of rotational, inflowing and outflowing components. By comparing the $M_{\rm BH}$ from dynamical modeling with the virial products based on reverberation lags and line widths, we determine the virial factor $f$ for individual AGNs. Combining our sample with those $M_{\rm BH}$ consistently determined from BLR dynamical modeling, yielding a total of 38 objects, we derive a virial factor for future $M_{\rm BH}$ estimation of log$_{10}({f})_{\rm pred}=0.69\pm0.21$ based on $σ_{\rm line,rms}$ and $-0.08\pm0.23$ based on FWHM$_{\rm mean}$. The derived virial factor is consistent with that inferred by aligning the reverberation-mapped AGNs with quiescent galaxies in the $M_{\rm BH}$-$σ_{\ast}$relation, supporting the assumption that local active and inactive galaxies follow the same $M_{\rm BH}$-$σ_{\ast}$ relation. Our updated $f$ values exhibit an intrinsic dispersion of $\sim0.2$ dex, which allows for a more precise $M_{\rm BH}$ estimates than those based on the $M_{\rm BH}$-$σ_{\ast}$ relation. Our sample extends the dynamical modeling-based reverberation sample to $M_{\rm BH}$ $\sim$ [$10^8$, $10^{8.5}$] $M_{\odot}$ range, where the virial factor from the the AGN $M_{\rm BH}$-$σ_{\ast}$ relation remains poorly constrained, underscoring the unique value of dynamical modeling analysis in constraining the $M_{\rm BH}$ of the most massive BHs.

astro-ph.GA↗
arXiv2026

Can BLR line profile shape improve single-epoch black hole mass estimates?

The virial coefficient ($f$), which is meant to encapsulate broad-line region (BLR) geometry and kinematics, remains one of the largest sources of systematic uncertainty in black hole mass estimates for Active Galactic Nuclei (AGNs). While the use of a sample average $\langle f \rangle$ enables black hole mass estimates across large samples and cosmological distances, individual AGNs may deviate from this average due to differences in BLR structure and viewing angle. In previous work, we reported marginal evidence for a correlation between $f$ and the shape of the broad H$β$ emission line, $\log_{10}(\mathrm{FWHM}/σ)$. In this work, we update our sample to include ten new sources with CARAMEL BLR dynamical modeling, increasing both the black hole mass range and statistical power of our analysis. We find marginal evidence for a correlation between $f$ and $\log_{10}(\mathrm{FWHM}/σ)$, with a slope and intrinsic scatter consistent with previous results. The confirmation of this trend across a larger sample further supports the idea that line profile shape may reflect BLR properties in a way that directly impacts $f$. If confirmed with future BLR dynamical modeling of sources within a wider range of $\log_{10}(\mathrm{FWHM}/σ)$, this relationship could enable empirical estimates of the virial coefficient and improve single-epoch black hole mass estimates across cosmic time.

astro-ph.GA↗
arXiv2026

LLM+Graph@VLDB'2025 Workshop Summary

The integration of large language models (LLMs) with graph-structured data has become a pivotal and fast evolving research frontier, drawing strong interest from both academia and industry. The 2nd LLM+Graph Workshop, co-located with the 51st International Conference on Very Large Data Bases (VLDB 2025) in London, focused on advancing algorithms and systems that bridge LLMs, graph data management, and graph machine learning for practical applications. This report highlights the key research directions, challenges, and innovative solutions presented by the workshop's speakers.

cs.DB↗