arXiv subjects

Yi Zhang

Publications and source records attributed to Yi Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2026

PRISM: Refracting the Entangled User Behavior Space for E-Commerce Search

E-commerce search systems rely on modeling user behavior to estimate item relevance and user preference, which are typically assumed to be stable and independently learnable signals. However, in practice, user interactions are jointly shaped by exposure mechanisms, feedback loops, and semantic matching, leading to entangled and dynamically drifting behavioral signals. As a result, both preference estimation and relevance modeling suffer from confounding effects and semantic misalignment, which limits the robustness of downstream ranking models. To address this issue, we propose PRISM, a Preference-Relevance Interaction Semantic Modeling framework for e-commerce search behavior prediction. PRISM explicitly models the interaction between user preference and item relevance rather than treating them as independent components. Specifically, it introduces a preference rectification module to iteratively refine user preference under relevance-aware constraints, improving robustness against behavioral confounding. To ensure semantic consistency, we further incorporate a large language model (LLM)-driven semantic anchoring mechanism that leverages positive and negative prototypes to calibrate relevance representations. Finally, a preference-conditioned evidence routing module adaptively aggregates multi-source behavioral signals, enabling context-aware and preference-aligned relevance estimation. Extensive experiments on two public e-commerce benchmarks demonstrate that PRISM consistently outperforms strong baselines, validating the effectiveness of explicitly modeling preference-relevance interaction for robust and semantically grounded search behavior modeling.

cs.IR↗

arXiv2026

The SRG/eROSITA diffuse soft X-ray background II. spectra and morphology of the eROSITA bubbles in the western Galactic hemisphere

The eROSITA bubbles (eRObub) were discovered in 2020 in the first SRG/eROSITA All-Sky Survey, and are among the most extended structures in the X-ray sky. Using eROSITA all-sky maps and spatially resolved spectra, we aim to infer the three-dimensional structure and measure the hot gas properties of the eRObub. We fit spectra binned to a constant S/N and high-S/N spectra from custom regions to examine gas properties in more detail. We fit the morphology of eRObub with a parametrised geometrical model that describes a blast wave propagating into an idealised Galactic halo from the centre. We found the interior of the western eRObub is best characterised by two emission components with relatively uniform temperatures: a hotter component at $kT=0.60\pm0.02$ keV, and a colder one at $kT=0.21^{+0.03}_{-0.01}$ keV, where the latter's emission measure is about five times higher on average. Our spectra suggest sub-solar abundances ($Z=0.2\pm0.1 Z_\odot$), consistent with expectations for the Galactic halo, while we find no conclusive evidence for $α$-element enhancement. In contrast, the North Polar Spur exhibits higher abundances ($Z>0.5 Z_\odot$), which, at face value, disfavours a common origin. We spectrally confirm an apparent cool shell at $kT\sim0.18$-$0.2$ keV surrounding the northern eRObub, assuming collisional ionisation equilibrium. We found no noticeable difference in X-ray emission in regions overlapping with the Fermi Bubbles. Our geometrical model suggests that the horizontal size of both eRObub is well-constrained (semi-minor axis $\sim 6$ kpc), but their vertical extent is uncertain, as the observed X-ray emission is almost insensitive to the existence and location of a bubble cap. Additionally, a tilt ($\sim 30^{\circ}$) towards $l\sim 220^{\circ}$ is needed to reproduce the projected image of the northern eRObub, whereas the southern bubble requires little tilt.

astro-ph.HE↗

arXiv2026

MOC-3D: Manifold-Order Consistency for Text-to-3D Generation

With the burgeoning development of fields such as the Metaverse, Virtual Reality (VR), and Digital Twins, text-to-3D generation has emerged as a research hotspot in both academia and industry. Currently, optimization methods based on Score Distillation Sampling (SDS) utilizing 2D diffusion priors have become the mainstream technological paradigm in this field. However, due to the view bias of 2D priors and the mode-seeking ambiguity combined with gradient noise induced by high Classifier-Free Guidance (CFG), these methods still suffer from macro-topological inconsistency (e.g., the Janus problem) and micro-geometric discontinuity. To address these challenges, we propose MOC-3D, a text-to-3D generation method based on geometric manifold and semantic view-order consistency. Built upon the ScaleDreamer framework, our method incorporates a Semantic View-Order Constraint Module and a Manifold-based Feature Continuity Module. The former aims to rectify macro-topological inconsistency, while the latter focuses on eliminating micro-geometric discontinuity. Specifically, the Semantic View-Order Constraint Module leverages the prior knowledge of CLIP to impose a Monotonicity Rank Constraint on semantic score representations across different views, thereby providing effective guidance for the global topological structure of 3D objects. Meanwhile, the Manifold-based Feature Continuity Module employs the Riemannian Metric on the Symmetric Positive Definite (SPD) manifold. By measuring the distance of feature statistical distributions in the Riemannian space, it promotes the smooth evolution and continuity of micro-textures across multi-views in a statistical sense. Under the macro-micro synergistic optimization of these two modules, our model can simultaneously improve macro-structural consistency and micro-detail continuity.

cs.CV↗

arXiv2026

Asymmetric Invertible Threat: Learning Reversible Privacy Defense for Face Recognition

Face Recognition systems are widely deployed in real-world applications, but they also raise privacy concerns due to unauthorized collection and misuse of facial data. Existing adversarial privacy protection methods rely on input-space perturbations to obfuscate identity information, yet their protection can degrade when adversaries learn restoration or purification mappings that partially invert the transformation. We study this setting as an asymmetric adversarial attack, in which reverse manipulation becomes feasible because existing defense paradigms do not control reversibility. To address this problem, we propose Asymmetric Reversible Face Protection (ARFP), a restoration-aware extension of personalized face cloaking that integrates privacy protection, keyed recovery, and tamper indication in a single framework. ARFP consists of three components: Key-Conditioned Manifold Binding, which ties the protection transformation to a user-provided key; Adversarial Restoration-Aware Training, which introduces a surrogate restoration adversary during training to improve robustness against evaluated inverse purification attacks; and Authorized Reversible Restoration, which supports recovery with the correct key while providing nonce-based tamper indication. Extensive experiments under the threat models considered in this work show that ARFP improves resistance to the evaluated restoration attacks while preserving authorized recovery utility. These results provide empirical evidence of key-sensitive recovery behavior and tamper awareness in the tested settings.

cs.CV↗

arXiv2026

Penalized Likelihood for Dyadic Network Formation Models with Degree Heterogeneity

Estimating network formation models with degree heterogeneity raises two problems in empirical networks. First, agents that send no links, receive no links, or link to all remaining agents can make the fixed-effects MLE fail to exist. Trimming these agents changes the estimation sample and induces selection bias. Second, the incidental-parameter problem biases common parameters and average partial effects. We resolve both issues through a penalized likelihood approach. Our leading specification is a directed network model with reciprocity, nesting the standard undirected and non-reciprocal directed models. The penalty guarantees finite-sample existence and yields bias corrections for coefficients and partial effects. We establish asymptotic results without imposing compactness on the fixed-effects. Allowing the fixed effects to diverge at a logarithmic rate, our asymptotic framework accommodates the degree sparsity ubiquitous in large empirical networks. A global trade application demonstrates that our estimator avoids selection bias and recovers robust parameters where conventional methods fail.

econ.EM↗

arXiv2026

From Static Analysis to Audience Dissemination: A Training-Free Multimodal Controversy Detection Multi-Agent Framework

Multimodal controversy detection (MCD) identifies controversial content in videos and their associated user comments, to support risk management for social video platforms.Prior research frames MCD as a static representation learning task, where features are directly extracted from videos and their accompanying comments. However, these methods fail to capture the diverse perspectives and evaluations from different audience groups. Inspired by the real-world process of content dissemination among audiences, we propose AuDisAgent, a training-free multi-agent framework that reformulates MCD as a dynamic propagation process.Our framework explicitly models audience dissemination through a structured multi-agent system. First, three specialized Screening Agents (Video Agent, Comment Agent, and Interaction Agent) conduct initial assessments from visual, textual, and cross-modal perspectives, respectively. For samples where the three agents cannot reach a consensus, a Viewing Panel Agent is activated to simulate post-screening discussions among audiences with diverse backgrounds and stances. This mechanism models how different audience groups interpret and react to the same content, uncovering latent controversial content that may emerge during the dissemination process. Finally, an Arbitration Agent renders the final judgment based on the complete reasoning chain from the preceding steps.In addition, to address the "cold-start" scenario where newly released videos have few or no comments, we design a Comment Bootstrapping Strategy that leverages historical public comments from semantically similar videos as the initial comment context. Extensive experiments on a public dataset demonstrate that our framework significantly outperforms existing state-of-the-art (SOTA) methods in both rich-comment and limited-comment scenarios.

cs.LG↗

arXiv2026

PrismAgent: Illuminating Harm in Memes via a Zero-Shot Interpretable Multi-Agent Framework

The rapid spread of memes makes harmful content detection increasingly crucial, as effective identification can curb the circulation of misinformation. However, existing methods rely heavily on high-volume annotated data, which leads to substantial training costs and limited generalization. To address these challenges, we propose PrismAgent, a zero-shot, multi-agent, interpretable framework. PrismAgent conceptualizes this task as a criminal case investigation, employing four specialized agents responsible for the analysis, investigation, prosecution, and judgment stages within a structured collaborative workflow. In the first stage, the analyst agent paraphrases each meme under benevolent and malicious assumptions to probe its underlying intent. The investigator agent then retrieves supporting evidence from an unannotated dataset and constructs contextual interpretations for the meme and its variants. Next, the prosecutor agent performs three independent preliminary judgments by pairing the original meme with each of the three interpretations. Finally, the judge agent deliberates across all evidence to render a final verdict. Moreover, PrismAgent's explicit multi-stage reasoning chain makes the model inherently interpretable, as every intermediate step is explicitly explained rather than only producing a final detection result. Extensive experiments on three public datasets show that PrismAgent significantly outperforms existing zero-shot detection methods.

cs.LG↗

arXiv2026

A systematic study of AGN feedback in a disk galaxy using MACER II: predictions of X-ray surface brightness profiles and comparison with eROSITA observations

Recently, we have performed a systematic study of AGN feedback in a disk galaxy within the MACER framework. Various model predictions, including the AGN duty cycle, the correlation between black hole accretion rates and star formation rates, and the (cold) gas fraction, have been compared with observations and will be presented in a series of papers. As the second paper in this series, without adjusting any model parameters, we directly use the simulation data introduced in Paper I to compute the predicted X-ray surface brightness profile and compare it with eROSITA observations of circumgalactic medium (CGM) emission around galaxies, which provide important constraints on AGN feedback models. For this comparison, we adopt two stacked eROSITA radial profiles of X-ray surface brightness: (1) distant galaxies with log(M*/M_sun) = 10.5-11.0 at z ~ 0.02-0.10 from Y. Zhang et al. (2024), and (2) nearby L* galaxies within 50 Mpc from L. He and Z. Li (2026). We find that the average simulated profile over time is in good agreement with the stacked measurements of Y. Zhang et al. (2024) over a broad radial range (out to ~100 kpc). Our model predictions also match the results of L. He and Z. Li (2026) at projected radii from ~20 kpc to 120 kpc. While our simulations, which predict only thermal emission, are consistent with these recent X-ray observations, the limitations in our current model mean that this agreement does not preclude a potential contribution from non-thermal emission, for example, from an extended halo of cosmic rays.

astro-ph.GA↗

arXiv2026

Minimax Regret Estimation for Generalizing Heterogeneous Treatment Effects with Multisite Data

To test scientific theories and develop individualized treatment rules, researchers often wish to learn heterogeneous treatment effects that can be consistently found across diverse populations and contexts. We consider the problem of generalizing heterogeneous treatment effects (HTE) based on data from multiple sites. A key challenge is that a target population may differ from the source sites in unknown and unobservable ways. This means that the estimates from site-specific models lack external validity, and a simple pooled analysis risks bias. We develop a robust CATE (conditional average treatment effect) estimation methodology with multisite data from heterogeneous populations. We propose a minimax-regret framework that learns a generalizable CATE model by minimizing the worst-case regret over a class of target populations whose CATE can be represented as convex combinations of site-specific CATEs. Using robust optimization, the proposed methodology accounts for distribution shifts in both individual covariates and treatment effect heterogeneity across sites. We show that the resulting CATE model has an interpretable closed-form solution, expressed as a weighted average of site-specific CATE models. Thus, researchers can utilize a flexible CATE estimation method within each site and aggregate site-specific estimates to produce the final model. Through simulations and a real-world application, we show that the proposed methodology improves the robustness and generalizability of existing approaches.

stat.ME↗

arXiv2026

ProMax: Exploring the Potential of LLM-derived Profiles with Distribution Shaping for Recommender Systems

The remarkable text understanding and generation capabilities of large language models (LLMs) have revitalized the field of general recommendation based on implicit user feedback. Rather than deploying LLMs directly as recommendation models, a more flexible paradigm leverages their ability to interpret users' historical interactions and semantic contexts to extract structured profiles that characterize user preferences. These profiles can be further transformed into actionable high-dimensional representations, serving as powerful signals to augment and strengthen recommendation models. However, the mechanism by which such profiles enhance recommendation performance within the feature space remains insufficiently understood. Moreover, existing studies predominantly rely on nonlinear alignment and fusion strategies to incorporate these profiles, which often lead to semantic loss and fail to fully exploit their potential. To address these limitations, we revisit profiles from a retrieval perspective and propose a simple yet effective recommendation framework built upon distribution shaping (ProMax) in this paper. We begin by employing dense retrieval to uncover the collaborative relationships between user and item profiles within the feature space. Based on this insight, we introduce a dual distribution-reshaping process, in which the profile distribution acts as a guiding signal to steer the recommendation model toward learning user preferences for unseen items beyond the scope of observed interactions. We apply ProMax to four classic recommendation methods on three public datasets. The results indicate that ProMax substantially improves base model performance and outperforms existing LLM-based recommendation approaches.

cs.IR↗

arXiv2026

Vision SmolMamba: Spike-Guided Token Pruning for Energy-Efficient Spiking State-Space Vision Models

Spiking Transformers have shown strong potential for long-range visual modeling through spike-driven self-attention. However, their quadratic token interactions remain fundamentally misaligned with the sparse and event-driven nature of spiking neural computation. To address this limitation, we propose Vision SmolMamba, an energy-efficient spiking state-space architecture that integrates spike-driven dynamics with linear-time selective recurrence. The key idea is a Spike-Guided Spatio-Temporal Token Pruner (SST-TP), which estimates token importance using both spike activation strength and first-spike latency. This mechanism progressively removes redundant tokens while preserving salient spatio-temporal information, enabling efficient scaling with token sparsity. Based on this mechanism, the proposed SmolMamba block incorporates spike events directly into bidirectional state-space recurrence, forming a spiking state-space vision backbone for efficient long-range modeling. Extensive experiments on both static and event-based benchmarks, including ImageNet-1K, CIFAR10/100, CIFAR10-DVS, and DVS128 Gesture, demonstrate that Vision SmolMamba consistently achieves superior accuracy-efficiency trade-offs. In particular, it reduces the estimated energy cost by at least 1.5x compared with prior spiking Transformer baselines and a Spiking Mamba variant while maintaining competitive or improved accuracy. These results demonstrate that combining spike-guided token sparsity with state-space modeling offers a scalable and energy-efficient paradigm for spiking vision systems.

cs.CV↗

arXiv2026

Categorical Symmetries via Operator Algebras

We propose that the symmetry category associated to a 2D quantum field theory with 0-form $G$-symmetry with 't Hooft anomaly $k\in H^4(BG,\mathbb{Z})$ for a large class of Lie groups $G$ is the category of twisted measurable fields of Hilbert spaces over $G$ denoted by $\mathrm{Hilb}^k(G)$, which is equivalent to the category of unitary representations of $C_0(G)$ with convolution product twisted by a multiplicative bundle gerbe labeled by $k$ denoted by $\textbf{Rep}^k(C_0(G))$. We find that the Drinfeld center of the symmetry category $\mathcal{Z}(\mathrm{Hilb}^{k}(G))$ equivalent to the category of unitary representations of the groupoid $C^*$-algebra of the Fell line bundle $Σ_k$ over the conjugation action groupoid $G//_{\rm Ad} G$, denoted by $\textbf{Rep}(C^*(G//_{\rm Ad}G,Σ_k))$, where the twist is characterized by the transgression $τ(k)\in H^2(G//_{\rm Ad}G,U(1))$. To the full generality, our framework applies to a Lie group $G$ that is a direct product of a compact connected Lie group and a number of $\mathbb{R}$ or $GL(1,\mathbb{C})$ factors. We compute the braiding of anyon lines in the bulk 3D SymTFT from this formalism. Finally we provide physical examples for abelian and non-abelian $G$, and discuss the physical consequences of flat gauging continuous global symmetries.

hep-th↗

arXiv2026

Disagreement as Signals: Dual-view Calibration for Sequential Recommendation Denoising

Sequential recommendation seeks to model the evolution of user interests by capturing temporal user intent and item-level transition patterns. Transformer-based recommenders demonstrate a strong capacity for learning long-range and interpretable dependencies, yet remain vulnerable to behavioral noise that is misaligned with users' true preferences. Recent large language model (LLM)-based approaches attempt to denoise interaction histories through static semantic editing. Such methods neglect the learning dynamics of recommendation models and fail to account for the evolving nature of user interests. To address this limitation, we propose a Dual-view Calibration framework for Sequential Recommendation denoising (DC4SR). Specifically, we introduce a semantic prior, derived from an LLM fine-tuned via labeled historical interactions, to estimate the noise distribution from a semantic perspective. From the learning perspective, we further employ a model-side posterior that infers the noise distribution based on the model's learning dynamics. The disagreement between the two distributions is then leveraged to jointly refine semantic understanding and learning-aware model-side representations. Through iterative updates, dynamic dual-view calibration is achieved for both the global semantic prior and the model-side posterior, enabling consistent alignment with evolving user interests. Extensive experiments demonstrate that DC4SR consistently outperforms strong Transformer-based recommenders and LLM-based denoising methods, exhibiting enhanced robustness across training stages and noise conditions.

cs.IR↗

arXiv2026

FreeScale: Distributed Training for Sequence Recommendation Models with Minimal Scaling Cost

Modern industrial Deep Learning Recommendation Models typically extract user preferences through the analysis of sequential interaction histories, subsequently generating predictions based on these derived interests. The inherent heterogeneity in data characteristics frequently result in substantial under-utilization of computational resources during large-scale training, primarily due to computational bubbles caused by severe stragglers and slow blocking communications. This paper introduces FreeScale, a solution designed to (1) mitigate the straggler problem through meticulously load balanced input samples (2) minimize the blocking communication by overlapping prioritized embedding communications with computations (3) resolve the GPU resource competition during computation and communication overlapping by communicating through SM-Free techniques. Empirical evaluation demonstrates that FreeScale achieves up to 90.3% reduction in computational bubbles when applied to real-world workloads running on 256 H100 GPUs.

cs.LG↗

arXiv2026

Explicit Trait Inference for Multi-Agent Coordination

LLM-based multi-agent systems (MAS) show promise on complex tasks but remain prone to coordination failures such as goal drift, error cascades, and misaligned behaviors. We propose Explicit Trait Inference (ETI), a psychologically grounded method for improving coordination. ETI enables agents to infer and track partner characteristics along two established psychological dimensions--warmth (e.g., trust) and competence (e.g., skill)--from interaction histories to guide decisions. We evaluate ETI in controlled settings (economic games), where it reduces payoff loss by 45-77%, and in more realistic, complex multi-agent settings (MultiAgentBench), where it improves performance by 3-29% depending on the scenario and model, relative to a CoT baseline. Additional analysis shows that gains are closely linked to trait inference: ETI profiles predict agents' actions, and informative profiles drive improvements. These results highlight ETI as a lightweight and robust mechanism for improving coordination in diverse multi-agent settings, and provide the first systematic evidence that LLM agents can (i) reliably infer others' traits from interaction histories and (ii) leverage structured awareness of others' traits for coordination.

cs.AI↗

arXiv2026

Supplement Generation Training for Enhancing Agentic Task Performance

Training large foundation models for agentic tasks is increasingly impractical due to the high computational costs, long iteration cycles, and rapid obsolescence as new models are continuously released. Instead of post-training massive models for every new task or domain, we propose Supplement Generation Training (SGT), a more efficient and sustainable strategy. SGT trains a smaller LLM to generate useful supplemental text that, when appended to the original input, helps the larger LLM solve the task more effectively. These lightweight models can dynamically adapt supplements to task requirements, improving performance without modifying the underlying large models. This approach decouples task-specific optimization from large foundation models and enables more flexible, cost-effective deployment of LLM-powered agents in real-world applications.

cs.LG↗

arXiv2026

One Jailbreak, Many Tongues: Learning Language-Insensitive Intention Representations for Multilingual Jailbreak Detection

Large language models (LLMs) are increasingly deployed in applications for global multilingual users, yet safety training remains concentrated in dominant languages and has not progressed in parallel with multilingual capability, creating exploitable gaps for jailbreak attacks. Current jailbreak defenses are largely developed and evaluated in dominant languages, and their effectiveness is limited by the scarcity of aligned multilingual supervision and representations dispersion caused by language variation. To address this issue, we propose MLJailDe, a multilingual jailbreak detection framework designed to improve both multilingual robustness and cross-lingual generalization. MLJailDe first introduces a multilingual back-translation data augmentation algorithm to construct a semantically consistent and functionally effective dataset spanning 11 languages, consisting of 2,232 benign and 1,239 jailbreak samples. On this basis, MLJailDe employs relative-distance constraints to reduce cross-lingual representation dispersion and encourage jailbreak prompts with similar intent to form consistent clusters across languages, while an imbalance-aware classification objective is further used to alleviate class imbalance and learn more reliable multilingual decision boundaries. Experimental results show that MLJailDe outperforms state-of-the-art baselines across multiple languages, achieving an F1 score of 98.5\%, and obtains an average F1 score of 97.1\% on unseen languages, demonstrating strong effectiveness and cross-lingual generalization.

cs.CL↗

arXiv2026

TLoRA: Task-aware Low Rank Adaptation of Large Language Models

Low-Rank Adaptation (LoRA) has become a widely adopted parameter-efficient fine-tuning method for large language models, with its effectiveness largely influenced by the allocation of ranks and scaling factors, as well as initialization. Existing LoRA variants typically address only one of these factors, often at the cost of increased training complexity or reduced practical efficiency. In this work, we present Task-aware Low-Rank Adaptation (TLoRA), a unified framework that jointly optimizes initialization and resource allocation at the outset of training. TLoRA introduces a data-driven initialization strategy that aligns the LoRA $A$ matrix with task-relevant subspaces by performing singular value decomposition on the product of pre-trained weights and input activation covariance. After this, the $A$ matrix is frozen, and only the $B$ matrix is trained. Furthermore, TLoRA employs a sensitivity-based importance metric to adaptively allocate ranks and scaling factors across layers under a fixed parameter budget. We conduct extensive experiments that demonstrate TLoRA consistently performs excellently across various tasks, including natural language understanding, commonsense reasoning, math reasoning, code generation, and chat generation, while significantly reducing the number of trainable parameters.

cs.CL↗