arXiv subjects

Shuo Huang

Publications and source records attributed to Shuo Huang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Fine-grained Analysis of Non-parametric Estimation for Pairwise Learning

In this paper, we are concerned with the generalization performance of non-parametric estimation for pairwise learning. Most of the existing work requires the hypothesis space to be convex or a VC-class, and the loss to be convex. However, these restrictive assumptions limit the applicability of the results in studying many popular methods, especially kernel methods and neural networks. We significantly relax these restrictive assumptions and establish a sharp oracle inequality of the empirical minimizer with a general hypothesis space for the Lipschitz continuous pairwise losses. As an example, we apply our general results to study pairwise least squares regression and derive an excess population risk bound that matches the minimax lower bound for the pointwise least squares regression. The key novelty lies in constructing a structured deep ReLU neural network to approximate the true predictor, and in designing a targeted hypothesis space composed of networks with this structure and controllable complexity. Experiments validate the effectiveness of the proposed method. This example demonstrates that the obtained general results indeed help us to explore the generalization performance on a variety of problems that cannot be handled by existing approaches.

stat.ML↗

arXiv2025

Tau Anomaly Detection in PET Imaging via Bilateral-Guided Deterministic Diffusion Model

The emergence of tau PET imaging over the last decade has enabled Alzheimer's disease (AD) researchers to examine tau pathology in vivo and more effectively characterize the disease trajectories of AD. Current tau PET analysis methods, however, typically perform inferences on large cortical ROIs and are limited in the detection of localized tau pathology that varies across subjects. In this work, we propose a novel bilateral-guided deterministic diffusion sampling method to perform anomaly detection from tau PET imaging data. By including individualized brain structure and cognitively normal (CN) template conditions, our model computes a voxel-level anomaly map based on the deterministically sampled pseudo-healthy reconstruction. We train our model on ADNI CN subjects (n=380) and evaluate anomaly localization performance on the left MCI/AD subjects (n=154) and the preclinical subjects of the A4 clinical trial (n=447). We further train a CNN classifier on the derived 3D anomaly maps from ADNI, including CN and MCI/AD, to classify subjects into two groups and test classification performance on A4. We demonstrate that our method outperforms baselines in anomaly localization. Additionally, we show that our method can successfully group preclinical subjects with significantly different cognitive functions, highlighting the potential of our approach for application in preclinical screening tests. The code will be publicly available.

eess.IV↗

arXiv2025

HunyuanVideo 1.5 Technical Report

We present HunyuanVideo 1.5, a lightweight yet powerful open-source video generation model that achieves state-of-the-art visual quality and motion coherence with only 8.3 billion parameters, enabling efficient inference on consumer-grade GPUs. This achievement is built upon several key components, including meticulous data curation, an advanced DiT architecture featuring selective and sliding tile attention (SSTA), enhanced bilingual understanding through glyph-aware text encoding, progressive pre-training and post-training, and an efficient video super-resolution network. Leveraging these designs, we developed a unified framework capable of high-quality text-to-video and image-to-video generation across multiple durations and resolutions. Extensive experiments demonstrate that this compact and proficient model establishes a new state-of-the-art among open-source video generation models. By releasing the code and model weights, we provide the community with a high-performance foundation that lowers the barrier to video creation and research, making advanced video generation accessible to a broader audience. All open-source assets are publicly available at https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5.

cs.CV↗

arXiv2025

Large gas inflow driven by a matured galactic bar in the early Universe

Bar structures are present in about half of local disk galaxies and play pivotal roles in secular galaxy evolution. Bars impose a non-axisymmetric perturbation to the rotating disk and transport gas inward to feed central starburst and, possibly, the activity of the nuclear supermassive black hole. They are believed to be long-lived structures and are now identified at redshift $z>2$. Yet, little is known about the onset and effect of bars in the early cosmic epoch because spectroscopy of distant bars at sufficient resolution is prohibitively expensive. Here, we report a kinematic study of a galactic bar at redshift 2.467, 2.6 billion years after the Big Bang. We observe the carbon monoxide and atomic carbon emission lines of the dusty star-forming galaxy J0107a and find the bar of J0107a has gas distribution and motion in a pattern identical to local bars. At the same time, the bar drives large-scale non-circular motions that dominate over disk rotation, funneling molecular gas into its center at a rate of $\approx600$ solar masses per year. Our results show that bar-driven dynamical processes and secular evolution were already at play 11.1 billion years ago, powering active star formation amid the gas-rich and far-infrared luminous growth phase in a massive disk galaxy.

astro-ph.GA↗

arXiv2025

Magnetic transition in B2 Al-Cr-Co alloys

Using Density Functional Theory (DFT) calculations and Monte-Carlo (MC) simulations, we investigate the recently reported magnetic transition in B2 Al-Cr-Co alloys. The Cr sublattice is alloyed with different amounts of Co in the antiferromagnetic (AFM) B2 AlCr binary alloy and the resulting exchange interactions are analyzed within the Heisenberg Hamiltonian framework. DFT results reveal that at low Co concentrations the system favors the AFM order, while at high Co contents a transition to the ferromagnetic (FM) state is observed. Within the FM stability field, the Curie temperature (TC), obtained within the mean-field approximation, is below ~160 K and decreases with Co concentration. The calculated exchange parameters evolve systematically with Co content, and the trends are consistent with the DFT total energies. The magnetic configurations obtained from MC simulations follow the DFT results at low Cr levels but predict a spin-glass behavior for alloys containing more than 40 at.% Co on Cr sublattice. These findings provide a fundamental understanding of how the chemistry-driven changes in exchange interactions affect magnetism in the B2 Al-Cr-Co alloys.

cond-mat.mtrl-sci↗

arXiv2025

Zero-Shot Privacy-Aware Text Rewriting via Iterative Tree Search

The increasing adoption of large language models (LLMs) in cloud-based services has raised significant privacy concerns, as user inputs may inadvertently expose sensitive information. Existing text anonymization and de-identification techniques, such as rule-based redaction and scrubbing, often struggle to balance privacy preservation with text naturalness and utility. In this work, we propose a zero-shot, tree-search-based iterative sentence rewriting algorithm that systematically obfuscates or deletes private information while preserving coherence, relevance, and naturalness. Our method incrementally rewrites privacy-sensitive segments through a structured search guided by a reward model, enabling dynamic exploration of the rewriting space. Experiments on privacy-sensitive datasets show that our approach significantly outperforms existing baselines, achieving a superior balance between privacy protection and utility preservation.

cs.CL↗

arXiv2025

Unveiling Stellar Feedback and Cloud Structure in the $ρ$ Ophiuchi A Region with ALMA and JWST: Discovery of Substellar Cores, C$^{18}$O Striations, and Protostellar Outflows

In clustered star-forming regions, stellar feedback-such as HII regions/photon-dominated regions (PDRs), and protostellar jets/outflows-shapes cloud structures and influences star formation. Using high-resolution ALMA millimeter and JWST infrared data, we analyze the cloud structure and the impact of stellar feedback in the nearest dense cluster-forming region Oph A. All 6 known Class 0/I and 2 of 6 Flat Spectrum/Class II objects are detected in the 1.3 mm dust continuum. Additionally, we newly detected 7 substellar cores, three of which show compact near-infrared emission, suggesting they are young substellar objects. The remaining cores, with masses of 0.01 Msun and high densities, are likely gravitationally bound. They appear connected by faint CO finger-like structures extending from the triple Class 0 system VLA1623-2417 Aa+Ab+B, suggesting they may have been ejected from the close binary VLA1623 Aa+Ab. 12CO and near-infrared data reveal multiple protostellar outflows. From the comparison, we identified several new outflows/jets, and shocked structures associated to the GSS30 large bipolar bubble. Strong 12CO emission traces the eastern edge of the Oph A ridge, forming part of the expanding HII/PDR bubble driven by the nearby Herbig Be star S1. The northern ridge appears blown out, with warm gas flowing toward GSS 30, injecting additional turbulent momentum. Several C18O striations in the S1 bubble align with magnetic fields, and position-velocity diagrams show wave-like patterns, possibly reflecting magnetohydrodynamic waves. Stellar feedback significantly influences Oph A's cloud structure.

astro-ph.GA↗

arXiv2025

Design of thermal hysteresis in nonstoichiometric alloys with giant magnetocaloric effect

The non-stoichiometric Fe$_2$P-type (FeMnP$_{0.5}$Si$_{0.5}$)$_{1-x}$(FeV)$_{x}$ alloys ( $x=0, 0.01$, $0.02$, and $0.03$) have been investigated as potential candidates for magnetic refrigeration near room temperature. The magnetic ordering temperature decreases with increasing FeV concentration, $x$, which can be ascribed to decreased ferromagnetic coupling strength between the magnetic atoms. The strong magnetoelastic coupling in these alloys results in large values of the isothermal entropy change ($ΔS_M$); $15.7$ J/kgK, at $2$ T magnetic field for the $x = 0$ alloy. $ΔS_M$ decreases with increasing $x$. Results from M{ö}ssbauer spectroscopy reveal that the average hyperfine field (in the ferromagnetic state) and average center shift (in the paramagnetic state) have the same decreasing trend as $ΔS_M$. The thermal hysteresis ($ΔT_{hyst}$) of the magnetic phase transition decreases with increasing $x$, while the mechanical stability of the alloys improves due to the reduced lattice volume change across the magnetoelastic phase transition. The adiabatic temperature change $ΔT_{ad}$, which highly depends on $ΔT_{hyst}$, is $1.7$ K at $1.9$ T applied field for the $x = 0.02$ alloy.

cond-mat.mtrl-sci↗

arXiv2025

ADF22+: a declining faint end in the far-infrared luminosity function in the SSA22 protocluster at z=3.09

Protoclusters represent the densest regions of cosmic large-scale structure in the early universe and are the environment where present-day massive elliptical galaxies are assembled. Millimeter continuum emission offers a powerful probe of obscured star formation at high redshifts across various environments. In this paper, we present a deep ALMA 1.17 mm mosaic of the central 8 arcmin$^2$ ($\approx30$ comoving Mpc$^{2}$) region in the SSA22 protocluster at $z=3.09$ to study the faint dusty star-forming galaxy (DSFG) population. The continuum map achieves an RMS noise level of $\approx25$ $μ$Jy beam$^{-1}$ at $\approx1\arcsec$ spatial resolution, $\approx2\times$ the depth of previous observation of this field. We detected 53 sources with a signal-to-noise ratio above 4.2, doubling the number of detections. Utilizing optical to mid-infrared ancillary data, we search for spectroscopic redshift and identify 18 of 53 as cluster members. For sources with more than two photometric data points in the near-infrared, stellar mass ($M_\star$) and star formation rate (SFR) from spectral energy distribution fitting are presented. The 1.17 mm number count shows $>2\times$ excess at flux density $\gtrsim1$ mJy but are consistent with blank field in fainter flux bins. The monochromatic far-infrared luminosity function of the SSA22 protocluster core region suggests a lack of faint DSFGs. All SSA22 protocluster member galaxies detected at 1.17 mm have SFR within the $M_\star$-SFR relation of general star-forming galaxies. Our results suggest that an early overdense environment like SSA22 protocluster predominantly enhances the formation of massive early-type galaxies in present-day galaxy clusters, but that the star formation in individual member galaxies is likely driven by gas supply along the cosmic web and occurs in a secular way.

astro-ph.GA↗

arXiv2025

A Resonant Beginning for the Solar System Terrestrial Planets

In the past two decades, transit surveys have revealed a class of planets with thick atmospheres -- sub-Neptunes -- that must have completed their accretion in protoplanet disks. When planets form in the gaseous disk, the gravitational interaction with the disk gas drives their migration and results in the trapping of neighboring planets in mean motion resonances, though these resonances can later be broken when the damping effects of disk gas or planetesimals wane. It is widely accepted that the outer Solar System gas giant planets originally formed in a resonant chain, which was later disrupted by dynamical instabilities. Here, we explore whether the early formation of the terrestrial planets in a resonance chain (including Theia) can evolve to the present configuration. Using N-body simulations, we demonstrate that the giant planet instability would also have destabilized the terrestrial resonance chain, triggering moon-forming giant impacts in 20--50\% of our simulated systems, dependent on the initial resonance architecture. After the instability, the eccentricity and inclination of the simulated planets match their present-day values. Under the proposed scenario, the current period ratio of 3.05 between Mars and Venus -- devoid of any special significance in traditional late formation models -- naturally arises as a relic of the former resonance chain.

astro-ph.EP↗

arXiv2025

Stability of a cluster-disrupted mean-motion resonance (chain) in HR 8799 and PDS 70

HR~8799 is a planetary system with four planets potentially in a mean-motion resonance chain. It is unclear from the observations if they are in mean-motion resonance. Similarly, PDS~70 has two observed planets also potentially in mean-motion resonance. We simulate HR~8799 and PDS~70 under external perturbations to study their responds if in resonance or mean-motion resonance. We integrate the equations of motion for HR~8799 and PDS~70 starting with either in resonance or in mean-motion resonance and study their in isolation and in a star cluster. In the star cluster, we take the effects of passing stars into account. The dynamics of the star cluster is resolved using the Lonely Planets module in AMUSE. HR~8799 and PDS~70 in mean-motion resonance are stable, whereas in non-resonance they dissolve in $0.303\pm0.042$Myr and $1.26\pm0.25$Myr, respectively. In a cluster, the non-resonant HR~8799 is slightly more stable than in isolation, but still dissolves in $0.300\pm0.043$Myr, whereas the resonant planetary system remains stable for at least $0.71$Myr. In contrast, a non-resonant PDS~70 system is approximately equally stable in a cluster compared to isolation, and dissolves in $1.03\pm0.20$Myr, whereas the resonant PDS~70 system remains stable for at least $0.83$Myr. Considering the more stable solutions of mean-motion resonance for HR~8799, we argue that the planetary system was born in mean-motion resonance and that the mean-motion resonance was preserved. If HR~8799 was not born in resonance, the probability that it survived until the present day is negligible. Similarly, we argue that PDS~70 was probably born in mean-motion resonance and that its state was preserved. We also find that it is almost possible for planetary systems with a broken mean-motion resonance chain to survive longer in a perturbing cluster environment compared to isolation.

astro-ph.EP↗

arXiv2025

NAP^2: A Benchmark for Naturalness and Privacy-Preserving Text Rewriting by Learning from Human

The widespread use of cloud-based Large Language Models (LLMs) has heightened concerns over user privacy, as sensitive information may be inadvertently exposed during interactions with these services. To protect privacy before sending sensitive data to those models, we suggest sanitizing sensitive text using two common strategies used by humans: i) deleting sensitive expressions, and ii) obscuring sensitive details by abstracting them. To explore the issues and develop a tool for text rewriting, we curate the first corpus, coined NAP^2, through both crowdsourcing and the use of large language models (LLMs). Compared to the prior works on anonymization, the human-inspired approaches result in more natural rewrites and offer an improved balance between privacy protection and data utility, as demonstrated by our extensive experiments. Researchers interested in accessing the dataset are encouraged to contact the first or corresponding author via email.

cs.CL↗

arXiv2025

Oort Cloud Ecology. III. The Sun left the parent star cluster shortly after the giant planets formed

The Sun was born in a clustered environment with 10,000 other stars. Being an isolated star today, the Sun must have left the nest. We do not directly know when that happened, how violent the ejection was, or how far the Solar siblings have drifted apart. The mass of the fragile outer Opic-Oort cloud, (between $r_{\rm inner} \sim 30,000$\,au and $200\,000$au from the Sun) and the orbital distribution of planetesimals in the inner Hills-Oort cloud (between $\sim 1000$\,au and $\sim 30\,000$ au) are sensitive to the dynamical processes involving the Sun in the parent cluster. We aim at understanding the extend to which observing the Oort cloud constrains the Sun's birth environment. This is achieved by a combination of theoretical arguments and N-body simulations. We show that the current mass of the Opic-Oort cloud (between 0.2 and $2.0$ Earth masses) is best explained if the Sun left the nest within $\sim 20$\,Myr after the giant planets formed and migrated. As a consequence, the possible dynamical encounter with another star carving the Kuiper belt, the Sun's abduction of Sedna, and other perturbations induced by nearby stars then must have happened shortly after the giant planets in the Solar system formed, but before the Sun left the parent cluster. Signatures of the time spend in the parent cluster must still be visible in the outer parts of the Solar system today. The strongest constraints will be the discovery of a population of relatively low-eccentricity ($e < 0.9$) inner Oort-cloud (but $500 < a < 10^4$\,au) objects.

astro-ph.EP↗

arXiv2025

Word2Minecraft: Generating 3D Game Levels through Large Language Models

We present Word2Minecraft, a system that leverages large language models to generate playable game levels in Minecraft based on structured stories. The system transforms narrative elements-such as protagonist goals, antagonist challenges, and environmental settings-into game levels with both spatial and gameplay constraints. We introduce a flexible framework that allows for the customization of story complexity, enabling dynamic level generation. The system employs a scaling algorithm to maintain spatial consistency while adapting key game elements. We evaluate Word2Minecraft using both metric-based and human-based methods. Our results show that GPT-4-Turbo outperforms GPT-4o-Mini in most areas, including story coherence and objective enjoyment, while the latter excels in aesthetic appeal. We also demonstrate the system' s ability to generate levels with high map enjoyment, offering a promising step forward in the intersection of story generation and game design. We open-source the code at https://github.com/JMZ-kk/Word2Minecraft/tree/word2mc_v0

cs.CL↗

arXiv2025

Gemini Embedding: Generalizable Embeddings from Gemini

In this report, we introduce Gemini Embedding, a state-of-the-art embedding model leveraging the power of Gemini, Google's most capable large language model. Capitalizing on Gemini's inherent multilingual and code understanding capabilities, Gemini Embedding produces highly generalizable embeddings for text spanning numerous languages and textual modalities. The representations generated by Gemini Embedding can be precomputed and applied to a variety of downstream tasks including classification, similarity, clustering, ranking, and retrieval. Evaluated on the Massive Multilingual Text Embedding Benchmark (MMTEB), which includes over one hundred tasks across 250+ languages, Gemini Embedding substantially outperforms prior state-of-the-art models, demonstrating considerable improvements in embedding quality. Achieving state-of-the-art performance across MMTEB's multilingual, English, and code benchmarks, our unified model demonstrates strong capabilities across a broad selection of tasks and surpasses specialized domain-specific models.

cs.CL↗

arXiv2025

The Dynamical History of the Kepler-221 Planet System

Kepler-221 is a G-type star hosting four planets. In this system, planets b, c, and e are in (or near) a 6:3:1 three-body resonance even though the planets' period ratios show significant departures from exact two-body commensurability. Importantly, the intermediate planet d is not part of the resonance chain. To reach this resonance configuration, we propose a scenario in which there were originally five planets in the system in a chain of first-order resonances. After disk dispersal, the resonance chain became unstable and two planets quickly merged to become the current planet d. In addition, the b/c/e three-body resonance was re-established. We run N-body simulations using REBOUND to investigate the parameter space under which this scenario can operate. We find that our envisioned scenario is possible when certain conditions are met. First, the reformation of the three-body resonance after planet merging requires convergent migration between planets b and c. Second, as has previously pointed out, an efficient damping mechanism must operate to power the expansion of the b/c/e system. We find that planet d plays a crucial role during the orbital expansion phase due to destabilizing encounters of a three-body resonance between c, d, and e. A successful orbital expansion phase puts constraints on the planet properties in the Kepler-221 system including the planet mass ratios and the tidal quality factors for the planets. Our model can also be applied to other planet systems in resonance, such as Kepler-402 and K2-138.

astro-ph.EP↗

arXiv2025

Towards a Learning Theory of Representation Alignment

It has recently been argued that AI models' representations are becoming aligned as their scale and performance increase. Empirical analyses have been designed to support this idea and conjecture the possible alignment of different representations toward a shared statistical model of reality. In this paper, we propose a learning-theoretic perspective to representation alignment. First, we review and connect different notions of alignment based on metric, probabilistic, and spectral ideas. Then, we focus on stitching, a particular approach to understanding the interplay between different representations in the context of a task. Our main contribution here is relating properties of stitching to the kernel alignment of the underlying representation. Our results can be seen as a first step toward casting representation alignment as a learning-theoretic problem.

cs.LG↗

arXiv2025

Causal Discovery Inspired Unsupervised Domain Adaptation for Emotion-Cause Pair Extraction

This paper tackles the task of emotion-cause pair extraction in the unsupervised domain adaptation setting. The problem is challenging as the distributions of the events causing emotions in target domains are dramatically different than those in source domains, despite the distributions of emotional expressions between domains are overlapped. Inspired by causal discovery, we propose a novel deep latent model in the variational autoencoder (VAE) framework, which not only captures the underlying latent structures of data but also utilizes the easily transferable knowledge of emotions as the bridge to link the distributions of events in different domains. To facilitate knowledge transfer across domains, we also propose a novel variational posterior regularization technique to disentangle the latent representations of emotions from those of events in order to mitigate the damage caused by the spurious correlations related to the events in source domains. Through extensive experiments, we demonstrate that our model outperforms the strongest baseline by approximately 11.05\% on a Chinese benchmark and 2.45\% on a English benchmark in terms of weighted-average F1 score. We have released our source code and the generated dataset publicly at: https://github.com/tk1363704/CAREL-VAE.

cs.CL↗