arXiv subjects

Peng Huo

Publications and source records attributed to Peng Huo.

15 recordsLinked to original sources

arXiv2026

Time-Aware Diffusion based on Preference Disentanglement for Generative Recommendation

Recently, Generative Recommenders (GRs) have emerged as a transformative recommendation paradigm by replacing traditional item IDs with semantic indices (SIDs). Owing to the exceptional generative capabilities of diffusion models, a few pioneering works explore developing GRs with diffusion architectures as the backbone. However, a fatal limitation of existing diffusion-based GRs is that the diffusion process applies uniformly to all items within the historical interactions. In contrast, the user preference is shaped by multifaceted time-evolving factors and thus exhibits a non-stationary distribution in the temporal aspect. To bridge this gap, this study proposes a novel GR framework, named TDPM, by designing the time-aware diffusion on SID tokens. Specifically, TDPM explicitly integrates the impact of time-evolving user preferences into the diffusion process. In detail, the user preference is disentangled into (i) the period preference, which remains consistent over a long time-span, and (ii) the point preference, which is triggered by recent focal events. Extensive experiments on three public real-world datasets demonstrate the significant superiority of TDPM over the state-of-the-art baselines. TDPM achieves average improvements of up to 29.21% and 25.45% in terms of HR@20 and NDCG@20, respectively. The ablation study further underscores the necessity of time-aware token diffusion in diffusion-based GRs.

cs.IR↗

arXiv2026

Echoes in Filter Bubble: Diagnosing and Curing Popularity Bias in Generative Recommenders

Recently, Generative Recommenders (GRs), characterized by a unified end-to-end framework, have exhibited astonishing potential in transforming the recommendation paradigm. Despite their effectiveness, we recognize that GRs are still susceptible to the long-standing issue of popularity bias that has pervaded the recommendation community. Although a few studies have attempted to extend traditional debiasing methods to GRs, their effectiveness is marginal, and the fundamental reason why GRs suffer from popularity bias remains under-explored. To bridge this gap, this study focuses on two core aspects in GRs: the optimization of generative framework and the item tokenization based on semantic index. Based on theoretical analyses, we identify that the severe popularity bias emerges from the confluence of a token-level optimization flaw and the undifferentiated property of item tokenization. Accordingly, this study develops a novel generative recommender system, called Ghost, by designing the asymmetric unlikelihood optimization and the skeleton-founded tokenization. Extensive empirical evaluations across three datasets, alongside multiple SOTA baselines, reveal that Ghost substantially alleviates popularity bias and promotes fairer recommendations, while incurring slight degradation to the overall recommendation utility.

cs.IR↗

arXiv2026

Rel-MOSS: Towards Imbalanced Relational Deep Learning on Relational Databases

In recent advances, to enable a fully data-driven learning paradigm on relational databases (RDB), relational deep learning (RDL) is proposed to structure the RDB as a heterogeneous entity graph and adopt the graph neural network (GNN) as the predictive model. However, existing RDL methods neglect the imbalance problem of relational data in RDBs and risk under-representing the minority entities, leading to an unusable model in practice. In this work, we investigate, for the first time, class imbalance problem in RDB entity classification and design the relation-centric minority synthetic over-sampling GNN (Rel-MOSS), in order to fill a critical void in the current literature. Specifically, to mitigate the issue of minority-related information being submerged by majority counterparts, we design the relation-wise gating controller to modulate neighborhood messages from each individual relation type. Based on the relational-gated representations, we further propose the relation-guided minority synthesizer for over-sampling, which integrates the entity relational signatures to maintain relational consistency. Extensive experiments on 12 entity classification datasets provide compelling evidence for the superiority of Rel-MOSS, yielding an average improvement of up to 2.46% and 4.00% in terms of Balanced Accuracy and G-Mean, compared with SOTA RDL methods and classic methods for handling class imbalance.

cs.AI↗

arXiv2024

XiHe: A Data-Driven Model for Global Ocean Eddy-Resolving Forecasting

The leading operational Global Ocean Forecasting Systems (GOFSs) use physics-driven numerical forecasting models that solve the partial differential equations with expensive computation. Recently, specifically in atmosphere weather forecasting, data-driven models have demonstrated significant potential for speeding up environmental forecasting by orders of magnitude, but there is still no data-driven GOFS that matches the forecasting accuracy of the numerical GOFSs. In this paper, we propose the first data-driven 1/12{\deg} resolution global ocean eddy-resolving forecasting model named XiHe, which is established from the 25-year France Mercator Ocean International's daily GLORYS12 reanalysis data. XiHe is a hierarchical transformer-based framework coupled with two special designs. One is the land-ocean mask mechanism for focusing exclusively on the global ocean circulation. The other is the ocean-specific block for effectively capturing both local ocean information and global teleconnection. Extensive experiments are conducted under satellite observations, in situ observations, and the IV-TT Class 4 evaluation framework of the world's leading operational GOFSs from January 2019 to December 2020. The results demonstrate that XiHe achieves stronger forecast performance in all testing variables than existing leading operational numerical GOFSs including Mercator Ocean Physical SYstem (PSY4), Global Ice Ocean Prediction System (GIOPS), BLUElinK OceanMAPS (BLK), and Forecast Ocean Assimilation Model (FOAM). Particularly, the accuracy of ocean current forecasting of XiHe out to 60 days is even better than that of PSY4 in just 10 days. Additionally, XiHe is able to forecast the large-scale circulation and the mesoscale eddies. Furthermore, it can make a 10-day forecast in only 0.35 seconds, which accelerates the forecast speed by thousands of times compared to the traditional numerical GOFSs.

physics.ao-ph↗

arXiv2023

Vertical-supercooling-controlled interfacial instability for a spreading liquid film

Thermal effect is essential to regulate the interfacial instabilities for diverse technology applications. Here we report the fingering instability at the propagation front for a spreading liquid film subjected to the supercooling at the vertical direction. We find the onset timescale of hydrodynamic instability is strongly correlated with that of the vertical solidification process. This correlation is further validated in a non-uniform geometry, demonstrating the capability of controlling fingering instability by structure design. We attribute the identified interfacial instability to a pronounced thermo-viscous effect, since the rapidly increased viscosity of propagation front undergoing solidification can significantly enhance the mobility contrast locally in the vicinity of the spreading front, consequently producing the instability analogous to viscous fingering. This work offers another valuable dimension by gating the vertical temperature to exploit the interfacial stabilities and steer liquid flow, consequently shedding light on the microfluidic cooling for electronics, and the advanced functional fibers and fabrics.

physics.flu-dyn↗

arXiv2022

Exploiting Dynamic and Fine-grained Semantic Scope for Extreme Multi-label Text Classification

Extreme multi-label text classification (XMTC) refers to the problem of tagging a given text with the most relevant subset of labels from a large label set. A majority of labels only have a few training instances due to large label dimensionality in XMTC. To solve this data sparsity issue, most existing XMTC methods take advantage of fixed label clusters obtained in early stage to balance performance on tail labels and head labels. However, such label clusters provide static and coarse-grained semantic scope for every text, which ignores distinct characteristics of different texts and has difficulties modelling accurate semantics scope for texts with tail labels. In this paper, we propose a novel framework TReaderXML for XMTC, which adopts dynamic and fine-grained semantic scope from teacher knowledge for individual text to optimize text conditional prior category semantic ranges. TReaderXML dynamically obtains teacher knowledge for each text by similar texts and hierarchical label information in training sets to release the ability of distinctly fine-grained label-oriented semantic scope. Then, TReaderXML benefits from a novel dual cooperative network that firstly learns features of a text and its corresponding label-oriented semantic scope by parallel Encoding Module and Reading Module, secondly embeds two parts by Interaction Module to regularize the text's representation by dynamic and fine-grained label-oriented semantic scope, and finally find target labels by Prediction Module. Experimental results on three XMTC benchmark datasets show that our method achieves new state-of-the-art results and especially performs well for severely imbalanced and sparse datasets.

cs.AI↗

arXiv2021

Armstrong Liquid Bridge: Formation, Evolution and Breakup

In this paper, we experimentally explore the formation, evolution and breakup of Armstrong liquid bridge. The extremely complicated evolution stage is revealed, which involves many coupled processes including the morphology change, current variation, heat transfer, and water evaporation. By focusing on the final fate of this liquid bridge, we observe that the breakup occurs once an effective length ($\tilde{L}$) is reached. This effective length increases linearly with the applied voltage, implying a threshold electric field to sustain the liquid bridge. Moreover, by an introduced external flow, the lifetime of the liquid bridge can be controlled, while the effective length associated with the breakup is independent on the external flow rate. Hence, these findings remarkably demonstrate that the breakup of liquid bridge is directly correlated with the effective length. In order to understand this correlation, a simplified model of an electrified jet is employed to take the electric field into account. By the linear stability analysis, the attained phase diagram agrees with the experiments well. Although a more comprehensive theory is required to consider other factors such as the surface charges, these results might provide a fresh perspective on the ``century-old" Armstrong liquid bridge to further elucidate its underlying physical mechanism.

physics.flu-dyn↗

arXiv2020

Electro-osmotic Instability of Concentration Enrichment in Curved Geometries for an Aqueous Electrolyte

We report that an electro-osmotic instability of concentration enrichment in curved geometries for an aqueous electrolyte, as opposed to the well-known one, is initiated exclusively at the enriched interface (anode), rather than at the depleted one (cathode). For this instability, the limitation of unrealistically high material Peclet number in planar geometry is eliminated by the strong electric field arising from the line charge singularity. In a model setup of concentric circular electrodes, we show by stability analysis, numerical simulation, and experimental visualization that instability occurs at the inner anode, below a critical radius of curvature. The stability criterion is also formulated in terms of a critical electric field and extended to arbitrary (2d) geometries by conformal mapping. This discovery suggests that transport may be enhanced in processes limited by salt enrichment, such as reverse osmosis, by triggering this instability with needle-like electrodes.

physics.flu-dyn↗

arXiv2019

Deionization Shock Driven by Electroconvection in a Circular Channel

In a circular channel passing over-limiting current (faster than diffusion), transient vortices of bulk electroconvection are observed in salt-depleted region within the horizontal plane. The spatiotemporal evolution of the salt concentration is directly visualized, revealing the propagation of a deionization shock wave driven by bulk electroconvection up to millimeter scales. This novel mechanism leads to quantitatively similar dynamics as for deionization shocks in charged porous media, which are driven instead by surface conduction and electro-osmotic flow at micron to nanometer scales. The remarkable generality of deionization shocks under over-limiting current could be used to manipulate ion transport in complex geometries for desalination and water treatment.

physics.flu-dyn↗

arXiv2018

Multiparticle azimuthal cumulants in p+Pb collisions from a multiphase transport model

A new subevent cumulant method was recently developed, which can significantly reduce the non-flow contributions in long-range correlations for small systems compared to the standard cumulant method. In this work, we study multi-particle cumulants in $p$+Pb collisions at $\sqrt{s_{\mathrm{NN}}} = 5.02$ TeV with a multiphase transport model (AMPT), including two- and four-particle cumulants ($c_{2}\{2\}$ and $c_{2}\{4\}$) and symmetric cumulants [SC(2, 3) and SC(2, 4)]. Our numerical results show that $v_{2}\{2\}$ is consistent with the experimental data, while the magnitude of $c_{2}\{4\}$ is smaller than the experimental data, which may indicate either the collectivity is underestimated or some dynamical fluctuations are absent in the AMPT model. For the symmetric cumulants, we find that the results from the standard cumulant method are consistent with the experimental data, but those from the subevent cumulant method show different behaviors. The results indicate that the measurements from the standard cumulant method are contaminated by non-flow effects, especially when the number of produced particles is small. The subevent cumulant method is a better tool to explore the $real$ collectivity in small systems.

hep-ph↗

arXiv2017

Importance of non-flow in mixed-harmonic multi-particle correlations in small collision systems

Recently CMS Collaboration measured mixed-harmonic four-particle azimuthal correlations, known as symmetric cumulants SC(n,m), in pp and pPb collisions, and interpreted the non-zero SC(n,m) as evidence for long-range collectivity in these small collision systems. Using the PYTHIA and HIJING models which do not have genuine long-range collectivity, we show that the CMS results, obtained with standard cumulant method, could be dominated by non-flow effects associated with jet and dijets, especially in $pp$ collisions. We show that the non-flow effects are largely suppressed using the recently proposed subevent cumulant methods by requiring azimuthal correlation between two or more pseudorapidity ranges. We argue that the reanalysis of SC(n,m) using the subevent method in experiments is necessary before they can used to provide further evidences for a long-range multi-particle collectivity and constraints on theoretical models in small collision systems.

nucl-ex↗

arXiv2017

Observables for longitudinal flow correlations in heavy-ion collisions

We propose several new observables/correlators, based on correlations between two or more subevents separated in pseudorapidity $η$, to study the longitudinal flow fluctuations. We show that these observables are sensitive to the event-by-event fluctuations, as a function of $η$, of the initial condition as well as the non-linear mode-mixing effects. Experimental measurement of these observables shall provide important new constraints on the boost-variant event-by-event initial conditions required by all 3+1-dimensional viscous hydrodynamics models.

nucl-th↗

arXiv2014

Elucidating the event-by-event flow fluctuations in heavy-ion collisions via the event shape selection technique

The presence of large event-by-event flow fluctuations in heavy ion collisions at RHIC and the LHC provides an opportunity to study a broad class of flow observables. This paper explores the correlations among harmonic flow coefficients $v_n$ and their phases $Φ_n$, and the rapidity fluctuation of $v_n$. The study is carried out usin Pb+Pb events generated by the AMPT model with fixed impact parameter. The overall ellipticity/triangularity of events is varied by selecting on the eccentricities $ε_n$ or the magnitudes of the flow vector $q_n$ for n=2 and 3, respectively. The responses of the $v_n$, the event-plane correlations, and the rapidity fluctuations, to the change in $ε_n$ and $q_n$ are then systematized. Strong positive correlations are observed among all even harmonics $v_2, v_4$, and $v_6$ (all increase with $q_2$), between $v_2$ and $v_5$ (both increase with $q_2$) and between $v_3$ and $v_5$ (both increase with $q_3$), consistent with the effects of nonlinear collective response. In contrast, an anti-correlation is observed between $v_2$ and $v_3$ similar to that seen between $ε_2$ and $ε_3$. These correlation patterns are found to be independent of whether selecting on $q_n$ or $ε_n$, validating the ability of $q_n$ in selecting the initial geometry. A forward/backward asymmetry of $v_n(η)$ is observed for events selected on $q_n$ but not on $ε_n$, reflecting dynamical fluctuations exposed by the $q_n$ selection. Many event-plane correlators show good agreement between $q_n$ and $ε_n$ selections, suggesting that their variations with $q_n$ are controlled by the change of $ε_n$ in the initial geometry. Hence these correlators may serve as promising observables for disentangling the fluctuations generated in various stages of the evolution of the matter created in heavy ion collisions.

nucl-ex↗

arXiv2014

A method for studying the rapidity fluctuation and decorrelation of harmonic flow in heavy-ion collisions

An "event-shape twist" technique is proposed to study the longitudinal dynamics of harmonic flow, in particular the effects of rapidity fluctuation and event-plane decorrelation. This technique can distinguish between two types of rapidity decorrelation effects: a systematic rotation versus a random fluctuation of flow angles along the rapidity direction. The technique is demonstrated and the magnitude of the two decorrelation effects is predicted using the AMPT model via a single particle analysis and two-particle correlation analysis. An observed decorrelation can be attributed to a systematic rotation of event-plane angle along the pseudorapidity, consistent with a collective response to an initial state twist of the fireball proposed by Bozek {\it et.al.}. This rotation is also observed for several higher-order harmonics with the same sign and similar magnitudes.

nucl-th↗

arXiv2014

Forward-backward eccentricity and participant-plane angle fluctuations and their influences on longitudinal dynamics of collective flow

We argue that the transverse shape of the fireball created in heavy ion collision is controlled by event-by-event fluctuations of the eccentricity vectors for the forward-going and backward-going wounded nucleons: $\vecε_n^{\mathrm{F}}\equiv ε_n^{\mathrm{F}} e^{i nΦ_n^{\mathrm{*F}}}$ and $\vecε_n^{\mathrm{B}}\equiv ε_n^{\mathrm{B}} e^{i nΦ_n^{\mathrm{*B}}}$. Due to the asymmetric energy deposition of each wounded nucleon along its direction of motion, the eccentricity vector of the produced fireball is expected to interpolate between $\vecε_n^{\mathrm{F}}$ and $\vecε_n^{\mathrm{B}}$ along the pseudorapidity, and hence exhibits sizable forward-backward(FB) asymmetry ($ε_n^{\rm B}\neqε_n^{\rm F}$) and/or FB-twist ($Φ_n^{\mathrm{*F}}\neqΦ_n^{\mathrm{*B}}$). A transport model calculation shows that these initial state longitudinal fluctuations for $n=2$ and 3 survive the collective expansion, and result in similar FB asymmetry and/or a twist in the final state event-plane angles. These novel EbyE longitudinal flow fluctuations should be accessible at RHIC and the LHC using the event-shape selection technique proposed in earlier papers. If these effects are observed experimentally, it could improve our understanding of the initial state fluctuations, particle production and collective expansion dynamics of the heavy ion collision.

nucl-th↗