Searcharxiv⌕ Search

arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 145 records · Page 8Linked to original sources
arXiv2026

SpecFormer: Mitigating Embedding and Attention Collapse via Spectral-Aware Transformer for Recommendation

Transformer architectures have achieved remarkable success across diverse domains; however, directly applying their standard self-attention mechanism to recommendation often yields suboptimal performance, sometimes even trailing behind well-designed simple recommendation models. In this paper, we reveal that this performance bottleneck stems from severe embedding and attention collapse unique to recommendation scenarios. The heterogeneity and long-tail nature of recommendation data lead to a severe spectral collapse dominated by a few principal singular values. We further theoretically demonstrate that this triggers a vicious cycle in recommendation model's forward and backward propagation, which accelerates embedding and attention collapse and limits the model's scaling capability with increased depth. To address these issues, we propose SpecFormer, a novel Spectral-Aware Transformer designed for mitigating embedding and attention collapse in recommendation. Specifically, SpecFormer introduces 1) a Learnable Spectral Softening module to dynamically smooth the singular values distribution of the input token embeddings; 2) a Spectrum-softened Attention mechanism to model feature interaction under a more uniform spectral distribution space; 3) a Spectral Residual Position Encoding via Taylor expansion of singular values, explicitly providing a spectral inductive bias for feature interactions. Extensive experiments on one industrial and two public datasets demonstrate that SpecFormer significantly outperforms state-of-the-art baselines. Notably, SpecFormer has been successfully deployed in a real-world commercial recommender system and exhibits exceptional scaling capabilities: stacking SpecFormer layers actively improves the attention effective rank and recommendation performance.

cs.IR↗
arXiv2026

Probing Quadruple Deformation in Transitional Nuclei via Angular Momentum Projection

Within the interacting boson model (IBM), a geometric analysis of transitional nuclei is carried out through angular momentum projection of the intrinsic coherent state. The results indicate that $K$-mixing effects in the calculations are typically negligible, validating the use of $K$-fixed projection for semiclassical analyses of the spin dependence of quadrupole deformation in the IBM. Further analysis indicates that in a rotating transitional system, quadrupole deformation is stretched with increasing angular momentum, providing a geometrically intuitive perspective on the commonly observed Jacobi-type transitions, as exemplified by the case studies of $^{160}$Gd and $^{162}$Dy. The method is additionally applied to the yrast states of $^{170}$Os to probe quadrupole deformation changes linked to the observed low-spin $B(E2)$ anomaly behavior, demonstrating the capability of IBM-based angular momentum projection in interpreting exotic collective phenomena.

nucl-th↗
arXiv2026

EntropyMoE: Entropy-Aware Sparse Expert Routing for Tokenizer-Free LLMs

Recent byte-level large language models (LLMs) have made tokenizer-free modeling increasingly competitive by grouping bytes into dynamically sized patches. However, existing byte-patch architectures still apply the same dense feed-forward computation to every patch. This uniform computation cannot adapt model capacity to variations in patch semantics and granularity. We address this limitation with EntropyMoE, a Mixture-of-Experts (MoE) architecture designed for dynamic byte patches. EntropyMoE replaces the dense feed-forward modules in the global patch Transformer with Top-K expert layers. Each dynamic patch serves as the basic unit of expert routing, and its byte coverage determines its contribution to workload accounting. The router selects experts directly from patch entropy, using the same granularity signal that underlies dynamic patch construction to organize sparse computation. Patch entropy and length jointly define the feature space for regulating expert specialization. Experiments show that EntropyMoE achieves the lowest held-out bits-per-byte among matched dense and sparse baselines while maintaining comparable downstream accuracy. These results establish patch entropy as an effective routing coordinate for sparse conditional computation and extend Mixture-of-Experts modeling beyond tokenizer-based representations.

cs.AI↗
arXiv2026

Same-sign dimuon probe of charged lepton flavor violation at electron-photon colliders

Observation of charged lepton flavor violation would constitute unambiguous evidence for physics beyond the Standard Model (SM). We identify a previously unexplored same-sign dimuon signature in electron--photon collisions, $γe^- \to e^+μ^-μ^-$, mediated by an axionlike particle (ALP) with flavor-violating $e$--$μ$ couplings. The absence of irreducible SM backgrounds and the on-shell production of the ALP render this channel intrinsically clean and highly sensitive, with only small residual backgrounds arising from detector effects. Such collisions can be realized via laser Compton backscattering at $e^+e^-$ colliders including BEPC-II with the BESIII detector, STCF, CEPC, and ILC. We find that STCF, CEPC, and ILC can probe couplings one to two orders of magnitude below existing bounds. This combination of resonant production, vanishing irreducible background, and same-sign topology would be difficult to achieve in conventional $e^+e^-$ or hadron-collider environments, establishing electron--photon collisions as a uniquely powerful probe of charged lepton flavor violation.

hep-ph↗
arXiv2026

Long-lived sterile neutrinos from axionlike particles at the Super Tau-Charm Facility

We study the search prospect of long-lived heavy neutral leptons (HNLs) pair produced in decays of axionlike particles (ALPs) at the proposed Super Tau--Charm Facility (STCF), focusing on the center-of-mass energy of $\sqrt{s}=3.773$ GeV. The ALPs are assumed to originate from $D^\pm$-meson decays in association with a charged pion. We perform both truth-level and detector-level Monte Carlo simulations and obtain the expected sensitivity reach to the mixing parameter between the HNL and the electron neutrino, $|V_{eN}|^2$, with a displaced-vertex search at STCF. We find that STCF can probe values of $|V_{eN}|^2$ about one-to-two orders of magnitude beyond the existing bounds. We also perform an approximate reinterpretation of a search for HNLs at the CHARM experiment, which is subject to model-dependent assumptions on the production and kinematic distributions, and find that beam-dump experiments may provide strong complementary constraints.

hep-ph↗
arXiv2026

Filesystem-Based Memory for LLM Agents: Organization, Evolution, and Sustainability

Deployed LLM agents increasingly keep their long-term memory as a filesystem: a directory tree of markdown files that the agent itself reads, writes, and reorganizes through generic file tools. Yet research has largely passed over this medium: prior systems design bespoke memory representations and study retrieval over them, leaving the default's two working assumptions untested: that an agent can keep a growing store organized as memories accumulate, conflict, and go stale, and that this organization pays. We present the first systematic exploration of filesystem-based memory for LLM agents. We formalize the setting as three roles around one memory filesystem: a management agent integrates and organizes incoming content, a search agent answers queries with cited sources, and an execution agent supplies task trajectories that are distilled into skills, unifying declarative memory and skills in a single store. Across long-conversation benchmarks and embodied tasks, we vary memory shape (agent-organized hierarchy, verbatim dump, chunk retrieval), stream scale, tool harness (sandboxed shell, memory-tool-style functions, varied search tooling), and the strengths of the management and search agents, tracking answer quality, cost, and store health as memory grows. What organization reliably buys is search economy: organized stores roughly halve retrieval cost where material is large. Today's agents, however, fall short of the default's promise: in our growth study, organization erodes for all but the strongest management agent, and no agent we measure converts organization itself into better answers. And the model is not the only lever over a store's shape: changing the tool set alone reshapes the store as strongly as swapping the model. The study turns the filesystem default from an assumption into a design space for agent memory.

cs.CL↗
arXiv2026

Zhinv: Real-time hub-height wind field reconstruction using only local sparse observations

The high proportion of wind power connected to the grid places higher demands on fine-grained knowledge of regional wind fields. Since the wind information directly obtainable in actual operations is mostly sparse, discrete, and irregularly distributed local observations, it is difficult to directly meet the needs of tasks such as wind power regulation, wind resource assessment, and low-altitude environmental perception of continuous regional wind fields. Therefore, we propose Zhinv, an end-to-end reconstruction framework that directly weaves sparse and irregular observations into a fine-grid wind field at hub-height. Experiments in Northeast China, Europe, and Southeast Asia demonstrate that Zhinv can accurately, robustly, and efficiently reconstruct fine-grid wind fields from sparse observations, reducing the error by about 66% compared with Kriging. With local wind-power observations as input, Zhinv enables wind power centers to bypass NWP and complex assimilation processes, supporting direct and real-time wind resource assessment from locally available data.

cs.LG↗
arXiv2026

Noise-Induced Thermalization in Quantum Systems

In the current Noisy Intermediate-Scale Quantum era, noise is widely regarded as the primary obstacle to achieving fault-tolerant quantum computation. However, certain stages of the quantum computing pipeline can, in fact, benefit from this noise. In this work, we exploit the Eigenstate Thermalization Hypothesis to show that noise generically accelerates a fundamental task in quantum computing -- the preparation of Gibbs states. We demonstrate this behavior using classical and quantum simulations with Haar-random and phase-flip noise, respectively, on a spin-1/2 chain with a local Hamiltonian. Our non-integrable model sees ~3.5x faster thermalization in the presence of noise, while our integrable model, which would not otherwise thermalize, reaches a thermal state due to noise. Since certifying a local Gibbs state is relatively easy on a quantum computer, our approach provides a new practical solution to a key problem in quantum computing. More broadly, these results establish a new paradigm in which noise can be harnessed on quantum computers, enabling practical advantages before the years of fault-tolerance.

quant-ph↗
arXiv2026

Contrastive Parameter Disentanglement for Multi-modal Remote Sensing Image Generation

Existing remote sensing image generation methods are largely confined to single-modality synthesis and therefore fail to exploit the complementary information inherent in multimodal imagery. To address this limitation, we propose a contrastive parameter disentanglement framework for multimodal remote sensing image generation, which generates semantically consistent and structurally aligned images across multiple modalities, including optical, infrared, and synthetic aperture radar (SAR), from a single text prompt. Specifically, we introduce a contrastive parameter disentanglement module that disentangles shared semantics from modality-specific attributes at the parameter level within an orthogonal core subspace. Based on this module, we develop a disentangled optimization strategy that first constrains the parameter matrix A of the LoRA adapter to capture modality-invariant semantics through a multimodal contrastive objective and then guides multiple parameter matrices B to learn modality-specific attributes under text conditioning. This strategy enables the simultaneous generation of multimodal images with consistent semantic content and distinct modality characteristics. Furthermore, to ensure structural alignment across the generated images, we devise a query-key structure transfer mechanism that jointly models multimodal sampling trajectories during inference by transferring structural correlation priors from an anchor modality to the remaining modalities. Extensive experiments demonstrate that our method outperforms state-of-the-art remote sensing image generation approaches in terms of generation quality, semantic consistency, and structural alignment, while also achieving superior performance in the downstream object classification task.

cs.CV↗
arXiv2026

EGR: Embedding-Native Generative Retrieval with a Shared LLM

Generative retrieval is increasingly popular in large-scale recommendation and advertising systems, yet current methods introduce practical complications. Semantic-ID methods rely on quantization, mutable identifier vocabularies, and token-to-item grounding; embedding-based pipelines train the item encoder separately from the query generator, which limits user-item alignment. We propose EGR, an Embedding-Native Generative Retrieval framework for recommendation and advertising. EGR uses a single shared LLM to learn item representations from item metadata and user representations from interaction histories in one embedding space. Items are indexed directly as dense vectors, and user histories are encoded as dense retrieval queries. Joint contrastive training groups related items and aligns queries with their target items. We evaluate EGR on public benchmarks, industrial data, and live deployment. EGR outperforms published baselines on Amazon Reviews; on Snap DPA, it scales with data, handles cold-start items, and benefits from multimodal input. In production, EGR delivers a +2.91% conversion-rate lift, simplifying system design while improving retrieval quality and ad performance.

cs.IR↗
arXiv2026

SMART: LLM-Augmented Hybrid Retrieval for Dynamic Product Ads

Dynamic Product Ads (DPA) require retrieving relevant items from multi-million product catalogs, balancing two competing objectives: retargeting (re-surfacing known interests) and prospecting (discovering new categories). While Large Language Models (LLMs) capture semantic intent better than traditional embedding models, deploying them at scale introduces prohibitive inference costs and lexical mismatch issues. Through controlled experiments on millions of users, we demonstrate a critical retrieval decomposition: rule-generated queries excel at retargeting on a lexical BM25 index, while LLM-generated queries excel at prospecting on a dense ANN index. Building on this, we propose SMART (SeMantic-aware Adaptive ReTrieval). To manage costs, a lightweight quality gate identifies coverage gaps in initial keyword results, adaptively routing only the ~10% of users who benefit from semantic prospecting to the LLM path. Offline evaluation demonstrates that this gated approach captures the bulk of semantic prospecting gains in Relevance Score while maintaining competitive re-targeting performance at a 90% reduction in LLM costs. Finally, in a 2-week online A/B test at Snap, SMART improved the ad conversion rate by +27.6% over a strong embedding-based baseline.

cs.IR↗
arXiv2026

The effects of a scalar singlet Leptoquark at the $Z$ factory

We evaluate the observability of the effects of a scalar singlet leptoquark (LQ) in $μ$ and $τ$-pair productions at the $Z$ factory. In the scenario addressing the charged-current anomalies, the LQ contributions to $μ$-pair final state are negligible. In contrast, a sizable contribution arises in the $τ$-pair production, which is identical in both $Z$ decay and $e^+e^-$ collider at $Z$ pole. These effects are mainly sensitive to left-handed interaction, showing a maximum deviation of about $-0.7\%$ for both 1\,TeV and 2\,TeV LQ. The suppression of new physics effects from the heavy LQ can be compensated by the enlarged couplings parameter space. For the $τ$-pair production channel, we further specify the coupling constraints corresponding to the expected measurement precision at the future $Z$ factory. Moreover, we provide an analytic function in terms of the LQ mass and couplings to quantify the LQ effects. The differential distributions in the collision process indicate that the LQ effects remain stable throughout the kinematic region. Meanwhile, the measurement sensitivity of the $τ$-pair final state at the future $Z$ factory is expected to impose further constraints on the LQ theory.

hep-ph↗
arXiv2026

When Can a Cavity Move a Mott Transition? A Spectral-Density Criterion within Gutzwiller Theory

Can vacuum electromagnetic fluctuations shift a bulk Mott transition? Within the Gutzwiller variational method, we derive a criterion that separates collective spectroscopic hybridization from thermodynamic phase control. We show that a Mott transition shifts only when the electromagnetic environment supplies finite thermodynamic spectral weight with bond-scale variation. A joint frequency--spatial Pauli--Fierz density gives the leading shift. Surface phonon polaritons yield a $d^{-3}$-to-$d^{-5}$ crossover, while finite-coordination variational Monte Carlo supports the predicted critical coefficient and $M/N$ scaling.

cond-mat.str-el↗
arXiv2026

BasketEvent: Understanding Who Did What and When in Basketball Videos

Comprehensive basketball video understanding requires resolving not only what event occurs, but also who is responsible and when the key evidence appears. However, exist- ing methods typically treat spatial perception and semantic recognition as isolated tasks, failing to ground events to individual players or pinpoint their temporal boundaries within complex collective dynamics. To bridge this gap, we introduce BasketEvent, a player- centric basketball event understanding dataset curated from real NBA broadcasts. In BasketEvent, event labels are grounded to the responsible players, and a manually an- notated subset of 1,000 samples with precise event intervals is provided to evaluate tem- poral evidence localization. Based on this data, we propose PlayNet, a player-centric reasoning framework that maps basketball videos to player-level event predictions with temporal evidence. Concretely, PlayNet tracks key entities, associates player identities, and reasons about events by modeling player-player, player-ball, and global court inter- actions, while aggregating sparse temporal evidence via gated pooling. Extensive experi- ments demonstrate that PlayNet significantly outperforms representative video-level and crop-based baselines, proving the superiority of player-centric modeling for fine-grained sports video understanding. Our data, code, and models will be made publicly available.

cs.AI↗
arXiv2026

X$^3$-OPD: Distilling Reasoning into Large Audio-Language Models via On-Policy Alignment

While large audio-language models have achieved remarkable progress in auditory perception, they still lag behind text-based large language models in deep logical reasoning, primarily due to the scarcity of high-quality audio reasoning data. To bridge this gap, we propose X$^3$-OPD, a cross-modal on-policy distillation framework that transfers reasoning capabilities from a powerful text teacher to an audio-language student. During training, the student generates reasoning trajectories conditioned on its own acoustic perception, while the teacher provides token-level guidance using matched textual inputs and verified answers. We further construct a three-tier symmetric corpus covering textual reasoning rendered into speech, audio-event reasoning grounded in complex acoustic scenes, and spoken-dialogue reasoning involving paralinguistic cues. This design extends cross-modal distillation beyond textually recoverable content to reasoning grounded in non-linguistic events, prosody, and conversational context. Experiments on MMSU, MMAU, BIG Bench Audio, and MMAR demonstrate that X$^3$-OPD substantially improves audio-grounded reasoning and chain-of-thought quality while largely preserving the model's existing capabilities under domain shift.

cs.LG↗
arXiv2026

In-the-Flow Agentic System Optimization for Effective Planning and Tool Use

Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios. Agentic systems offer a promising alternative by decomposing work across specialized modules, yet most remain training-free or rely on offline training decoupled from the live dynamics of multi-turn interaction. We introduce AgentFlow, a trainable, in-the-flow agentic framework that coordinates four modules (planner, executor, verifier, generator) through an evolving memory and directly optimizes its planner inside the multi-turn loop. To train on-policy in live environments, we propose Flow-based Group Refined Policy Optimization (Flow-GRPO), which tackles long-horizon, sparse-reward credit assignment by converting multi-turn optimization into a sequence of tractable single-turn policy updates. It broadcasts a single, verifiable trajectory-level outcome to every turn to align local planner decisions with global success and stabilizes learning with group-normalized advantages. Across ten benchmarks, AgentFlow with a 7B-scale backbone outperforms top-performing baselines with average accuracy gains of 14.9% on search, 14.0% on agentic, 14.5% on mathematical, and 4.1% on scientific tasks, even surpassing larger proprietary models like GPT-4o. Further analyses confirm the benefits of in-the-flow optimization, showing improved planning, enhanced tool-calling reliability, and positive scaling with model size and reasoning turns.

cs.AI↗
arXiv2026

Adaptive Visual Autoregressive Acceleration via Dual-Linkage Entropy Analysis

Visual AutoRegressive modeling (VAR) suffers from substantial computational cost due to the massive token count involved. Failing to account for the continuous evolution of modeling dynamics, existing VAR token reduction methods face three key limitations: heuristic stage partition, non-adaptive schedules, and limited acceleration scope, thereby leaving significant acceleration potential untapped. Since entropy variation intrinsically reflects the transition of predictive uncertainty, it offers a principled measure to capture modeling dynamics evolution. Therefore, we propose NOVA, a training-free token reduction acceleration framework for VAR models via entropy analysis. NOVA adaptively determines the acceleration activation scale during inference by online identifying the inflection point of scale entropy growth. Through scale-linkage and layer-linkage ratio adjustment, NOVA dynamically computes distinct token reduction ratios for each scale and layer, pruning low-entropy tokens while reusing the cache derived from the residuals at the prior scale to accelerate inference and maintain generation quality. Extensive experiments and analyses validate NOVA as a simple yet effective training-free acceleration framework.

cs.CV↗
arXiv2026

ClothTransformer: Unified Latent-Space Transformers for Scalable Cloth Simulation

Unified and scalable Transformers have recently achieved remarkable success in modeling diverse phenomena traditionally associated with computer graphics, such as 3D visual effects, rendering processes, and motion in videos. In this work, we take a step further by investigating whether modern Transformer techniques can tackle the challenging task of cloth simulation. To this end, we present ClothTransformer, a framework that reformulates cloth simulation as autoregressive sequence modeling in a learned latent space. Existing neural cloth simulators are largely specialized to single scenarios, intrinsically coupled to the mesh discretization, and lack robust collision handling. Our approach addresses these limitations through three contributions: (1) a unified Transformer architecture that handles diverse scenarios -- body-driven garments, robotic manipulation, and free-fall collisions -- under a single model and achieves approximately $4$--$9{\times}$ lower error than prior state-of-the-art methods across all scenarios; (2) a scalable latent-space formulation that compresses arbitrary-resolution meshes into a fixed-size set of latent tokens, making temporal dynamics computation independent of mesh resolution; and (3) a diverse-scenario high-fidelity penetration-free dataset of ${\sim}$493.4k frames spanning all three settings, which enables a differentiable Continuous Collision Detection (CCD) module to suppress penetration artifacts. Project Page: https://yucrazing.github.io/clothtransformer/

cs.GR↗