arXiv subjects

Xiaoyuan Zhang

Publications and source records attributed to Xiaoyuan Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Adaptive Online Mirror Descent for Tchebycheff Scalarization in Multi-Objective Learning

Multi-objective learning (MOL) aims to learn under multiple potentially conflicting objectives and strike a proper balance. While recent preference-guided MOL methods often rely on additional optimization objectives or constraints, we consider the classic Tchebycheff scalarization (TCH) that naturally allows for locating solutions with user-specified trade-offs. Due to its minimax formulation, directly optimizing TCH often leads to training oscillation and stagnation. In light of this limitation, we propose an adaptive online mirror descent algorithm for TCH, called (Ada)OMD-TCH. One of our main ingredients is an adaptive online-to-batch conversion that significantly improves solution optimality over traditional conversion in practice while maintaining the same theoretical convergence guarantees. We show that (Ada)OMD-TCH achieves a convergence rate of $\mathcal O(\sqrt{\log m/T})$, where $m$ is the number of objectives and $T$ is the number of rounds, providing a tighter dependency on $m$ in the offline setting compared to existing work. Empirically, we demonstrate on both synthetic problems and federated learning tasks that (Ada)OMD-TCH effectively smooths the training process and yields preference-guided, specific, diverse, and fair solutions.

cs.LG↗

arXiv2026

Precision Jet Substructure of Boosted Boson Decays with Energy Correlators

We initiate the precision study of boosted jet substructure using energy correlators, applying this framework to hadronic Higgs decays. We demonstrate that the two-body decay of the Higgs manifests as a distinct angular peak at $θ\sim \arccos(1-2/γ^2)$ for Lorentz boost factor $γ$. We show that infrared scales, such as the dead-cone effect and confinement transition, are also resolved within the boosted distribution. Precision analytic studies of boosted jet substructure may enable precision electroweak studies and open new avenues for new physics searches.

hep-ph↗

arXiv2026

The life of central radio galaxies in clusters: AGN-ICM studies of eRASS1 clusters in the ASKAP fields

The mechanical feedback from the central AGNs can be crucial for balancing the radiative cooling of the intracluster medium at the cluster centre. We aim to understand the relationship between the power of AGN feedback and the cooling of gas in the centres of galaxy clusters by correlating the radio properties of the brightest cluster galaxies (BCGs) with the X-ray properties of their host clusters. We used catalogues from the first SRG/eROSITA All-Sky Survey (eRASS1) along with ASKAP radio data. In total, we identified 134 radio sources associated with BCGs of the 151 eRASS1 clusters located in the PS1, PS2, and SWAG-X ASKAP fields. Non-detections were treated as upper limits. We correlated BCG radio luminosity, largest linear size (LLS), and BCG offset with the integrated X-ray luminosity of their host clusters. To characterise cool cores (CCs) and non-cool cores (NCCs), we used the concentration parameter $c_{R_{500}}$ and combined it with the BCG offset to assess cluster dynamical state. We analysed the correlation between radio mechanical power and X-ray luminosity within the CC subsample. We observe a potential positive trend between LLS and BCG offset, suggesting an environmental effect on radio-source morphology. We find a weak trend where more luminous central radio galaxies are found in clusters with higher X-ray luminosity. Within the CC subsample, there is a positive but highly scattered relationship between the mechanical luminosity of AGN jets and the X-ray cooling luminosity. This finding is supported by bootstrap resampling and flux-flux analyses. The correlation indicates that AGN feedback is ineffective in high-luminosity (high-mass) clusters. At a cooling luminosity of $L_{\mathrm{X},~r<R_\mathrm{cool}}\approx 5.50\times10^{43}$ erg/s, on average, AGN feedback appears to contribute only about 13%-22% of the energy needed to offset the radiative losses in the ICM.

astro-ph.CO↗

arXiv2026

Few for Many: Tchebycheff Set Scalarization for Many-Objective Optimization

Multi-objective optimization can be found in many real-world applications where some conflicting objectives can not be optimized by a single solution. Existing optimization methods often focus on finding a set of Pareto solutions with different optimal trade-offs among the objectives. However, the required number of solutions to well approximate the whole Pareto optimal set could be exponentially large with respect to the number of objectives, which makes these methods unsuitable for handling many optimization objectives. In this work, instead of finding a dense set of Pareto solutions, we propose a novel Tchebycheff set scalarization method to find a few representative solutions (e.g., 5) to cover a large number of objectives (e.g., $>100$) in a collaborative and complementary manner. In this way, each objective can be well addressed by at least one solution in the small solution set. In addition, we further develop a smooth Tchebycheff set scalarization approach for efficient optimization with good theoretical guarantees. Experimental studies on different problems with many optimization objectives demonstrate the effectiveness of our proposed method.

cs.LG↗

arXiv2025

POLAR: A Portrait OLAT Dataset and Generative Framework for Illumination-Aware Face Modeling

Face relighting aims to synthesize realistic portraits under novel illumination while preserving identity and geometry. However, progress remains constrained by the limited availability of large-scale, physically consistent illumination data. To address this, we introduce POLAR, a large-scale and physically calibrated One-Light-at-a-Time (OLAT) dataset containing over 200 subjects captured under 156 lighting directions, multiple views, and diverse expressions. Building upon POLAR, we develop a flow-based generative model POLARNet that predicts per-light OLAT responses from a single portrait, capturing fine-grained and direction-aware illumination effects while preserving facial identity. Unlike diffusion or background-conditioned methods that rely on statistical or contextual cues, our formulation models illumination as a continuous, physically interpretable transformation between lighting states, enabling scalable and controllable relighting. Together, POLAR and POLARNet form a unified illumination learning framework that links real data, generative synthesis, and physically grounded relighting, establishing a self-sustaining "chicken-and-egg" cycle for scalable and reproducible portrait illumination. Our project page: https://rex0191.github.io/POLAR/.

cs.CV↗

arXiv2025

ExLLM: Experience-Enhanced LLM Optimization for Molecular Design and Beyond

Molecular design involves an enormous and irregular search space, where traditional optimizers such as Bayesian optimization, genetic algorithms, and generative models struggle to leverage expert knowledge or handle complex feedback. Recently, LLMs have been used as optimizers, achieving promising results on benchmarks such as PMO. However, existing approaches rely only on prompting or extra training, without mechanisms to handle complex feedback or maintain scalable memory. In particular, the common practice of appending or summarizing experiences at every query leads to redundancy, degraded exploration, and ultimately poor final outcomes under large-scale iterative search. We introduce ExLLM (Experience-Enhanced LLM optimization), an LLM-as-optimizer framework with three components: (1) a compact, evolving experience snippet tailored to large discrete spaces that distills non-redundant cues and improves convergence at low cost; (2) a simple yet effective k-offspring scheme that widens exploration per call and reduces orchestration cost; and (3) a lightweight feedback adapter that normalizes objectives for selection while formatting constraints and expert hints for iteration. ExLLM sets new state-of-the-art results on PMO and generalizes strongly in our setup, it sets records on circle packing and stellarator design, and yields consistent gains across additional domains requiring only a task-description template and evaluation functions to transfer.

cs.LG↗

arXiv2025

World Models Should Prioritize the Unification of Physical and Social Dynamics

World models, which explicitly learn environmental dynamics to lay the foundation for planning, reasoning, and decision-making, are rapidly advancing in predicting both physical dynamics and aspects of social behavior, yet predominantly in separate silos. This division results in a systemic failure to model the crucial interplay between physical environments and social constructs, rendering current models fundamentally incapable of adequately addressing the true complexity of real-world systems where physical and social realities are inextricably intertwined. This position paper argues that the systematic, bidirectional unification of physical and social predictive capabilities is the next crucial frontier for world model development. We contend that comprehensive world models must holistically integrate objective physical laws with the subjective, evolving, and context-dependent nature of social dynamics. Such unification is paramount for AI to robustly navigate complex real-world challenges and achieve more generalizable intelligence. This paper substantiates this imperative by analyzing core impediments to integration, proposing foundational guiding principles (ACE Principles), and outlining a conceptual framework alongside a research roadmap towards truly holistic world models.

cs.CY↗

arXiv2025

Social World Model-Augmented Mechanism Design Policy Learning

Designing adaptive mechanisms to align individual and collective interests remains a central challenge in artificial social intelligence. Existing methods often struggle with modeling heterogeneous agents possessing persistent latent traits (e.g., skills, preferences) and dealing with complex multi-agent system dynamics. These challenges are compounded by the critical need for high sample efficiency due to costly real-world interactions. World Models, by learning to predict environmental dynamics, offer a promising pathway to enhance mechanism design in heterogeneous and complex systems. In this paper, we introduce a novel method named SWM-AP (Social World Model-Augmented Mechanism Design Policy Learning), which learns a social world model hierarchically modeling agents' behavior to enhance mechanism design. Specifically, the social world model infers agents' traits from their interaction trajectories and learns a trait-based model to predict agents' responses to the deployed mechanisms. The mechanism design policy collects extensive training trajectories by interacting with the social world model, while concurrently inferring agents' traits online during real-world interactions to further boost policy learning efficiency. Experiments in diverse settings (tax policy design, team coordination, and facility location) demonstrate that SWM-AP outperforms established model-based and model-free RL baselines in cumulative rewards and sample efficiency.

cs.CY↗

arXiv2025

Toward the Analytic Bootstrap of Energy Correlators

In this paper, we present a framework for the analytic bootstrap of three-point energy correlators, a crucial observable in $\mathcal{N}=4$ super Yang-Mills theory and quantum chromodynamics (QCD). Our approach combines spherical contour techniques, general physical constraints such as pole cancellations, and power correction data in the singular limits to determine its analytic expression. In contrast to previous bootstrap studies restricted to scattering amplitudes for supersymmetric theories, our framework makes use of the properties of Feynman integrals, marking a significant step toward bootstrapping realistic QCD observables. Using this method, we derive analytic expressions for leading-order three-point energy correlators with equal and unequal energy weights, where the latter are crucial ingredients for projected $N$-point energy correlators. We also apply the recently developed technique of analytic regression with lattice reduction as a way to bypass needing explicit expressions for the singular limits. Bridging theoretical advances in scattering amplitudes with the renewed interest in weighted cross-sections, our work opens the door to precision tests of QCD dynamics through analytic event-shape predictions.

hep-ph↗

arXiv2025

The SRG/eROSITA All-Sky Survey. The Weak-Lensing Mass Calibration and the Stellar Mass-to-Halo Mass Relation from the Hyper Suprime-Cam Subaru Strategic Program

We present the weak-lensing mass calibration and constrain the BCG (brightest cluster galaxy) stellar-mass-to-halo-mass-and-redshift ($M_{\star,\mathrm{BCG}}-M-z$) relation for a sample of $124$ galaxy clusters and groups at redshift $0.1<z<0.8$ from the first Data Release of the $eROSITA$ All-Sky Survey (eRASS1), using data from the Hyper Suprime-Cam (HSC) Subaru Strategic Program. The cluster survey is conducted by the $eROSITA$ X-ray telescope aboard the Spectrum-Roentgen-Gamma (SRG) space observatory. The cluster sample is X-ray-selected and optically confirmed with a negligibly low contamination rate ($\approx5%$). On a basis of individual clusters, the shear profiles of $96$ clusters are derived using the HSC Three-Year (HSC-Y3) weak-lensing data, while the BCG stellar masses of $101$ clusters are estimated using the SED template fitting to the HSC five-band ($grizY$) photometry. The observed X-ray photon count rate is used as the mass proxy, based on which individual halo masses are obtained at the given count rate in a population modelling while accounting for systematic uncertainties in the weak-lensing modelling through a simulation-calibrated weak-lensing mass-to-halo-mass relation. The count rate and BCG stellar mass relations are simultaneously constrained in a forward and population modelling. In agreement with the results based on the weak-lensing data from the DES and KiDS surveys, we obtain a count rate relation with a self-similar redshift scaling and a mass trend that is steeper than the self-similar prediction. Our results suggest that the BCG stellar mass at a fixed halo mass has remained stable with a moderate increase at a level of $\left(20\pm8\right)%$ since redshift $z\approx0.8$. This finding supports the picture of the ``rapid-then-slow'' BCG formation, where the majority of the stellar mass must have been assembled at much earlier cosmic time.

astro-ph.CO↗

arXiv2025

Enhancing LLM-Based Social Bot via an Adversarial Learning Framework

Developing Large Language Model (LLM) agents that exhibit human-like behavior, encompassing not only individual heterogeneity rooted in unique user profiles but also adaptive response to socially connected neighbors, is a significant research challenge. Social media platforms, with their diverse user data and explicit social structures, provide an ideal testbed for such investigations. This paper introduces EvoBot, an \textbf{Evo}lving LLM-based social \textbf{Bot} that significantly enhances human-like generative capabilities through a novel adversarial learning framework. EvoBot is initialized by Supervised Fine-Tuning (SFT) on representative data from social media and then iteratively refines its generation of sophisticated, human-like content via Direct Preference Optimization (DPO). This refinement is guided by feedback from a co-adapting \textbf{Detector} which concurrently improves its ability to distinguish EvoBot from humans, thereby creating an increasingly challenging learning environment for EvoBot. Experiments demonstrate that EvoBot generates content aligned with diverse user profiles, increasingly bypassing the co-adapting Detector through human-like expression. Moreover, it exhibits strong social responsiveness, more accurately modeling real-world opinion dynamics and information spread in multi-agent simulations. The framework also yields a more robust Detector, underscoring its broader utility for both advanced agent development and related detection tasks. The code is available at https://github.com/kfq20/EvoBot.

cs.SI↗

arXiv2025

Gradient-Based Multi-Objective Deep Learning: Algorithms, Theories, Applications, and Beyond

Many modern deep learning applications require balancing multiple objectives that are often conflicting. Examples include multi-task learning, fairness-aware learning, and the alignment of Large Language Models (LLMs). This leads to multi-objective deep learning, which tries to find optimal trade-offs or Pareto-optimal solutions by adapting mathematical principles from the field of Multi-Objective Optimization (MOO). However, directly applying gradient-based MOO techniques to deep neural networks presents unique challenges, including high computational costs, optimization instability, and the difficulty of effectively incorporating user preferences. This paper provides a comprehensive survey of gradient-based techniques for multi-objective deep learning. We systematically categorize existing algorithms based on their outputs: (i) methods that find a single, well-balanced solution, (ii) methods that generate a finite set of diverse Pareto-optimal solutions, and (iii) methods that learn a continuous Pareto set of solutions. In addition to this taxonomy, the survey covers theoretical analyses, key applications, practical resources, and highlights open challenges and promising directions for future research. A comprehensive list of multi-objective deep learning algorithms is available at https://github.com/Baijiong-Lin/Awesome-Multi-Objective-Deep-Learning.

cs.LG↗

arXiv2025

Analytic Regression of Feynman Integrals from High-Precision Numerical Sampling

In mathematics or theoretical physics one is often interested in obtaining an exact analytic description of some data which can be produced, in principle, to arbitrary accuracy. For example, one might like to know the exact analytical form of a definite integral. Such problems are not well-suited to numerical symbolic regression, since typical numerical methods lead only to approximations. However, if one has some sense of the function space in which the analytic result should lie, it is possible to deduce the exact answer by judiciously sampling the data at a sufficient number of points with sufficient precision. We demonstrate how this can be done for the computation of Feynman integrals. We show that by combining high-precision numerical integration with analytic knowledge of the function space one can often deduce the exact answer using lattice reduction. A number of examples are given as well as an exploration of the trade-offs between number of datapoints, number of functional predicates, precision of the data, and compute. This method provides a bottom-up approach that neatly complements the top-down Landau-bootstrap approach of trying to constrain the exact answer using the analytic structure alone. Although we focus on the application to Feynman integrals, the techniques presented here are more general and could apply to a wide range of problems where an exact answer is needed and the function space is sufficiently well understood.

hep-th↗

arXiv2025

The SRG/eROSITA All-Sky Survey : Subaru/HSC-SSP weak-lensing mass measurements for the eRASS1 Galaxy Clusters

We performed individual weak-lensing (WL) mass measurements for 78 eROSITA's first All-Sky Survey (eRASS1) clusters in the footprint of Hyper Suprime-Cam Subaru Strategic Program (HSC-SSP) S19A. We did not adopt priors on the eRASS1 X-ray quantities or assumption of the mass and concentration relation. In the sample, we found three clusters are misassociated with optical counterparts and 12 clusters are poorly fitted with an NFW profile. The average mass for the 12 poor-fit clusters changes from $\sim 10^{14}h_{70}^{-1}M_\odot$ to $\sim 2\times 10^{13}h_{70}^{-1}M_\odot$ when lensing contamination from surrounding mass structures is taken into account. The scaling relations between the true mass and cluster richness and X-ray count-rate agree well with the results of the eRASS1 western Galactic hemisphere region based on count-rate-inferred masses, which were calibrated with the HSC-SSP, DES, and KiDS surveys. We developed a Bayesian framework for inferring the mass-concentration relation of the cluster sample, explicitly incorporating the effects of weak-lensing mass calibration in the mass-concentration parameter space. The redshift-dependent mass and concentration relation is in excellent agreement with predictions of dark-matter-only numerical simulations and previous studies using X-ray-selected clusters. Based on the two-dimensional (2D) WL analysis, the offsets between the WL-determined centers and the X-ray centroids for 36 eRASS1 clusters with high WL S/N can be described by two Gaussian components. We find that the miscentering effect with X-ray centroids is smaller than that involving peaks in the galaxy maps. Stacked mass maps support a small miscentering effect, even for clusters with a low WL S/N. The projected halo ellipticity is $\langle \varepsilon \rangle=0.45$ at $M_{200}\sim 4\times10^{14}h_{70}^{-1}M_\odot$.

astro-ph.CO↗

arXiv2025

Discovery of Diffuse Radio Emission in a Massive z=1.709 Cool Core Cluster: A Candidate Radio Mini-Halo

Clusters of galaxies host spectacular diffuse radio sources, extending over scales from 100 kpc to several Mpcs. These sources, with extremely faint surface brightness ($μ$Jy/arcsec$^2$ level), are not tied to individual galaxies but trace synchrotron emission from large-scale magnetic fields and relativistic particles within the intracluster environment. Here, we report the discovery of a candidate radio mini-halo in SpARCS104922.6+564032.5, the most distant cool-core galaxy cluster identified to date at $z=1.709$, using deep LOFAR 120-168 MHz observations. We show that this emission originates from diffuse cluster-associated processes rather than unresolved AGN or star-forming galaxies. The diffuse radio emission coincides spatially with the X-ray emission of the hot intracluster medium and has a radio power of $P_{\rm 150~MHz}=49.8^{+14.7}_{-11.7} \times10^{24}$ W Hz$^{-1}$, exhibiting striking similarities to low-redshift radio mini-halos. This discovery doubles the redshift of previously known mini-halos, challenging models of inverse Compton losses and indicating the presence of strong magnetic fields, enhanced turbulence in high-redshift clusters, or active hadronic processes that require a cosmic ray-to-thermal energy ratio of 0.07 within 200 kpc, assuming a clumped distribution with spatial correlations among the gas, cosmic rays, and magnetic field that partially compensate for cosmological redshift dimming. It further implies that magnetic fields are efficiently amplified to $\sim$$10~μ$G levels within a Mpc$^3$ volume during the epoch of cluster formation before $z\sim2$. These findings provide critical insights into high-redshift cluster physics and emphasize the transformative potential of next-generation radio surveys, such as those with the SKA and ngVLA, in exploring the early evolution of galaxy clusters.

astro-ph.CO↗

arXiv2025

Can LLMs Understand Unvoiced Speech? Exploring EMG-to-Text Conversion with LLMs

Unvoiced electromyography (EMG) is an effective communication tool for individuals unable to produce vocal speech. However, most prior methods rely on paired voiced and unvoiced EMG signals, along with speech data, for EMG-to-text conversion, which is not practical for such individuals. Given the rise of large language models (LLMs) in speech recognition, we explore their potential to understand unvoiced speech. To this end, we address the challenge of learning from unvoiced EMG alone and propose a novel EMG adaptor module that maps EMG features into an LLM's input space, achieving an average word error rate (WER) of 0.49 on a closed-vocabulary unvoiced EMG-to-text task. Even with a conservative data availability of just six minutes, our approach improves performance over specialized models by nearly 20%. While LLMs have been shown to be extendable to new language modalities -- such as audio -- understanding articulatory biosignals like unvoiced EMG remains more challenging. This work takes a crucial first step toward enabling LLMs to comprehend unvoiced speech using surface EMG.

cs.CL↗

arXiv2025

Cross-correlation between soft X-rays and galaxies A new benchmark for galaxy evolution models

This article presents the construction and validation of complete stellar mass-selected, volume-limited galaxy samples using the Legacy Survey (data release 10) galaxy catalogs, covering $\sim16,800$ deg$^2$ of extra-galactic sky, and extending to redshift $z<0.35$. We measure the correlation function of these galaxies with tiny statistical uncertainties at the percent level and systematic uncertainties up to 5\%. A 4-parameter halo occupation distribution (HOD) model is fitted to retrieve the population of host halos, yielding results on the stellar to halo mass relation consistent with the current models of galaxy formation and evolution. Using these complete galaxy samples, we measure and analyze the cross-correlation (X-corr) between galaxies and all soft X-ray photons observed by SRG/eROSITA in the 0.5-2 keV band over $\sim13,000$ deg$^2$. The cross correlation measurements have unprecedented sub-percent statistical uncertainty and ~5-10\% systematic uncertainty. An extension to the halo model is introduced to interpret the X-corr, decomposing contributions from X-ray point sources, hot gas (CGM), satellites, and the 2-halo term. For low stellar mass thresholds ($\log M^*/M_{\odot}>$ 10, 10.25, 10.5), we find that the point source emission dominates the X-corr at small separation ($r<80$kpc). Then, in the range ($80 2$Mpc), the 2-halo term becomes dominant. Interestingly, there is no scale at which the CGM dominates. In the range ($20<r<200$kpc), the CGM contributes to more than 10\% of the signal. Progressively, with the minimum stellar mass increasing, the CGM emission increases. We constrain the $M_{500c}-L_X$ scaling relation slope, $1.629^{+0.091}_{-0.089}$, at the 5\% level using the samples with the lowest mass threshold.

astro-ph.GA↗

arXiv2025

Small radius inclusive jet production at the LHC through NNLO+NNLL

The study of hadronic jets and their substructure at hadronic colliders is crucial for improving our understanding of QCD, and searching for new physics. As such, there has been a significant effort to improve their theoretical description. In the small radius limit, inclusive jet production exhibits a universal factorization, enabling the resummation of logarithms which greatly stabilizes theoretical predictions. In this paper, we show how to combine a recently introduced framework for small-$R$ resummation with the STRIPPER subtraction formalism for fragmentation, enabling next-to-next-to-leading order calculations of small-$R$ inclusive jet production for a wide variety of processes at the LHC. We extract the two-loop constants for the jet functions, enabling for the first time next-to-next-to-leading logarithmic resummation matched to next-to-next-to-leading order perturbative calculation. We compare with CMS data for small-$R$ jet production, and find that our results greatly improve the accuracy of the predictions at small-$R$, and stabilize the perturbative convergence and error estimates at larger $R$. Our approach is applicable to a wide class of jet substructure observables exhibiting similar factorization theorems, opening the door to an NNLO jet substructure program at the LHC.

hep-ph↗