Searcharxiv⌕ Search

arXiv subjects

Yi Zhang

Publications and source records attributed to Yi Zhang.

At least 307 records · Page 17Linked to original sources
arXiv2026

SSR: Pushing the Limit of Spatial Intelligence with Structured Scene Reasoning

While Multimodal Large Language Models (MLLMs) excel in semantic tasks, they frequently lack the "spatial sense" essential for sophisticated geometric reasoning. Current models typically suffer from exorbitant modality-alignment costs and deficiency in fine-grained structural modeling precision.We introduce SSR, a framework designed for Structured Scene Reasoning that seamlessly integrates 2D and 3D representations via a lightweight alignment mechanism. To minimize training overhead, our framework anchors 3D geometric features to the large language model's pre-aligned 2D visual semantics through cross-modal addition and token interleaving, effectively obviating the necessity for large-scale alignment pre-training. To underpin complex spatial reasoning, we propose a novel scene graph generation pipeline that represents global layouts as a chain of independent local triplets defined by relative coordinates. This is complemented by an incremental generation algorithm, enabling the model to construct "language-model-friendly" structural scaffolds for complex environments. Furthermore, we extend these capabilities to global-scale 3D global grounding task, achieving absolute metric precision across heterogeneous data sources. At a 7B parameter scale, SSR achieves state-of-the-art performance on multiple spatial intelligence benchmarks, notably scoring 73.9 on VSI-Bench. Our approach significantly outperforms much larger models, demonstrating that efficient feature alignment and structured scene reasoning are the cornerstones of authentic spatial intelligence.

cs.CV↗
arXiv2026

Adaptive Multi-Objective Tiered Storage Configuration for KV Cache in LLM Service

The memory-for-computation paradigm of KV caching is essential for accelerating large language model (LLM) inference service, but limited GPU high-bandwidth memory (HBM) capacity motivates offloading the KV cache to cheaper external storage tiers. While this expands capacity, it introduces the challenge of dynamically managing heterogeneous storage resources to balance cost, throughput, and latency under varying workloads. We formulate this as a multi-objective optimization problem: identifying the Pareto frontier across these metrics within the storage configuration space. Using a high-fidelity end-to-end simulator, we observe that the objective functions are non-analytic and exhibit complex variable coupling, making the Pareto frontier difficult to approximate analytically. To obtain the frontier, we introduce Kareto, a KV-cache Adaptive REsource managemenT Optimizer. Kareto leverages a diminishing-return-guided pruning method to efficiently navigate the large configuration space and approximate the Pareto frontier. Additionally, it incorporates a fine-grained adaptive tuner that uses eviction policies in tier storage and KV block access patterns for group-specific cache management, improving cache efficiency. Experiments on real-world traces show that Kareto adapts to workload and can identify configurations of better cost efficiency, covering static strategies. Compared to the fixed setup with 1024 GB DRAM, Kareto can improve throughput by up to 9.3%, or reduce latency by up to 58.3%, or lower cost by up to 20.2% under respective optimization objectives.

cs.AR↗
arXiv2026

Your Code Agent Can Grow Alongside You with Structured Memory

While "Intent-oriented programming" (or "Vibe Coding") redefines software engineering, existing code agents remain tethered to static code snapshots. Consequently, they struggle to model the critical information embedded in the temporal evolution of projects, failing to leverage the "reasoning trajectories" implicit in past successful practices. This limitation results in rigid behavioral logic and a lack of autonomous adaptability, ultimately hindering their ability to tackle complex, repository-level problems. To bridge this static-dynamic mismatch, we propose MemCoder, a framework designed to enable continual human-AI co-evolution. MemCoder first structures historical human experience to distill latent intent-to-code mappings from past commits. It then employs a self-refinement mechanism driven by verification feedback to correct agent behavior in real-time. Crucially, an experience self-internalization mechanism is introduced to crystallize human-validated solutions into long-term knowledge, thereby supporting sustained evolution. Experimental results on SWE-bench Verified demonstrate that MemCoder not only achieves State-of-the-Art (SOTA) performance but also delivers a 9.4% improvement in resolved rate over the general foundation model DeepSeek-V3.2. These findings indicate that equipping agents with the capability to co-evolve with humans via project history and real-time feedback effectively unlocks the potential of general models in complex software engineering tasks.

cs.LG↗
arXiv2026

Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling

Large Language Models (LLMs) are converging towards a singular Artificial Hivemind, where shared Nature (pre-training priors) result in a profound collapse of distributional diversity, limiting the distinct perspectives necessary for creative exploration and scientific discovery. To address this, we propose to equip models with inference-time Nurture (individualized epistemic trajectories) using Epistemic Evolution paradigm, progressing through explore, internalize, and express. We instantiate this via PRISM (Pluralistic Reasoning via In-context Structure Modeling), a model-agnostic system that augments LLM with dynamic On-the-fly Epistemic Graphs. On three creativity benchmarks, PRISM achieves state-of-the-art novelty and significantly expands distributional diversity. Moreover, we evaluate the real-world utility via a challenging rare-disease diagnosis benchmark. Results demonstrate that PRISM successfully uncovers correct long-tail diagnoses that standard LLM miss, confirming that its divergence stems from meaningful exploration rather than incoherent noise. Overall, this work establishes a new paradigm for Pluralistic AI, moving beyond monolithic consensus toward a diverse ecosystem of unique cognitive individuals capable of collective, multi-perspective discovery.

cs.LG↗
arXiv2026

Probing $D_s^+ \to η^{(\prime)} \ell^+ν_\ell$ semileptonic decay within LCSR under chiral heavy quark effective field theory

Motivated by the successful application of Heavy Quark Effective Field Theory in describing decays from heavy to light mesons, this work explores its applicability to the semileptonic decays of charmed mesons. So in this paper we investigate the $D_s^+\to η^{(\prime)} \ell^+ ν_\ell$ transition form factors using the light-cone sum rules approach within the framework of heavy-quark effective field theory. To address the large uncertainties arsing from the $η^{(\prime)}$-meson twist-3 distribution amplitudes, we employ the right-handed chiral correlation function. By applying the converging simplified series expansion method, we extrapolate the form factors to the entire physical $q^2$-region. Our analysis yields the branching fractions precise predictions for semi-leptonic decays $D_s^+\to η^{(\prime)}\ell^+ν_\ell$ with : $\mathcal{B}(D_s^+\toη\ell^+ν_\ell)=2.300^{+0.230}_{-0.227}\%$ ($\ell = e$) and $2.249_{-0.206}^{+0.209}\%$ ($\ell = μ$); $\mathcal{B}(D_s^+\toη^\prime \ell^+ν_\ell)=0.861^{+0.095}_{-0.093}\%$ ($\ell = e$) and $0.821^{+0.082}_{-0.080}\%$ ($\ell = μ$). The derived lepton flavor universality ratios $R^η_{μ,e}=0.977^{+0.008}_{-0.006}$ and $R^{η^{\prime}}_{μ,e} = 0.953^{+0.011}_{-0.009}$ are consistent with lasted BESIII experimental measurements. Additionally, the forward-backward asymmetry parameters $\langle \mathcal{A}^η_{\rm FB}\rangle=-0.034^{+0.003}_{-0.003}$ and $\langle \mathcal{A}^{η^\prime}_{\rm FB}\rangle=-0.073^{+0.007}_{-0.008}$ suggest that no significant violation of lepton flavor universality in this decay process.

hep-ph↗
arXiv2026

Twisted multilayer moiré water waves topologically robust to disorder

Moiré patterns, stacking and twisting multilayer periodic lattices into superlattices, have become cornerstones of many physical systems from condensed matter to wave phenomena, but have never been properly studied in water waves. Here, we demonstrate twisted multilayer moiré water surface waves carrying robust skyrmionic topologies. Using a custom water tank of circular multi-channel phased array, we precisely generate water-wave skyrmion lattices and superimpose them into moiré superlattices with higher-order topological textures, e.g., skyrmion bags and clusters, programmed via the twist angle. We also quantitatively compare the topological robustness of bilayer and trilayer configurations under spatiotemporal perturbations. The trilayer moiré superlattices exhibit more enhanced stability, stronger field localization and energy concentration than the bilayer. Our work establishes water waves as a macroscopic, tunable, and visually accessible platform for moiré physics, towards robust particle manipulation and classical analogues of topological quantum phenomena.

physics.optics↗
arXiv2026

Distributed Virtual Model Control for Scalable Human-Robot Collaboration in Shared Workspace

We present a decentralized, agent agnostic, and safety-aware control framework for human-robot collaboration based on Virtual Model Control (VMC). In our approach, both humans and robots are embedded in the same virtual-component-shaped workspace, where motion is the result of the interaction with virtual springs and dampers rather than explicit trajectory planning. A decentralized, force-based stall detector identifies deadlocks, which are resolved through negotiation. This reduces the probability of robots getting stuck in the block placement task from up to 61.2% to zero in our experiments. The framework scales without structural changes thanks to the distributed implementation: in experiments we demonstrate safe collaboration with up to two robots and two humans, and in simulation up to four robots, maintaining inter-agent separation at around 20 cm. Results show that the method shapes robot behavior intuitively by adjusting control parameters and achieves deadlock-free operation across team sizes in all tested scenarios.

cs.RO↗
arXiv2026

From Newborn to Impact: Bias-Aware Citation Prediction

As a key to accessing research impact, citation dynamics underpins research evaluation, scholarly recommendation, and the study of knowledge diffusion. Citation prediction is particularly critical for newborn papers, where early assessment must be performed without citation signals and under highly long-tailed distributions. We identify two key research gaps: (i) insufficient modeling of implicit factors of scientific impact, leading to reliance on coarse proxies; and (ii) a lack of bias-aware learning that can deliver stable predictions on lowly cited papers. We address these gaps by proposing a Bias-Aware Citation Prediction Framework, which combines multi-agent feature extraction with robust graph representation learning. First, a multi-agent x graph co-learning module derives fine-grained, interpretable signals, such as reproducibility, collaboration network, and text quality, from metadata and external resources, and fuses them with heterogeneous-network embeddings to provide rich supervision even in the absence of early citation signals. Second, we incorporate a set of robust mechanisms: a two-stage forward process that routes explicit factors through an intermediate exposure estimate, GroupDRO to optimize worst-case group risk across environments, and a regularization head that performs what-if analyses on controllable factors under monotonicity and smoothness constraints. Comprehensive experiments on two real-world datasets demonstrate the effectiveness of our proposed model. Specifically, our model achieves around a 13% reduction in error metrics (MALE and RMSLE) and a notable 5.5% improvement in the ranking metric (NDCG) over the baseline methods.

cs.SI↗
arXiv2026

Choosing How to Remember: Adaptive Memory Structures for LLM Agents

Memory is critical for enabling large language model (LLM) based agents to maintain coherent behavior over long-horizon interactions. However, existing agent memory systems suffer from two key gaps: they rely on a one-size-fits-all memory structure and do not model memory structure selection as a context-adaptive decision, limiting their ability to handle heterogeneous interaction patterns and resulting in suboptimal performance. We propose a unified framework, FluxMem, that enables adaptive memory organization for LLM agents. Our framework equips agents with multiple complementary memory structures. It explicitly learns to select among these structures based on interaction-level features, using offline supervision derived from downstream response quality and memory utilization. To support robust long-horizon memory evolution, we further introduce a three-level memory hierarchy and a Beta Mixture Model-based probabilistic gate for distribution-aware memory fusion, replacing brittle similarity thresholds. Experiments on two long-horizon benchmarks, PERSONAMEM and LoCoMo, demonstrate that our method achieves average improvements of 9.18% and 6.14%.

cs.AI↗
arXiv2026

PLOT-CT: Pre-log Voronoi Decomposition Assisted Generation for Low-dose CT Reconstruction

Low-dose computed tomography (LDCT) reconstruction is fundamentally challenged by severe noise and compromised data fidelity under reduced radiation exposure. Most existing methods operate either in the image or post-log projection domain, which fails to fully exploit the rich structural information in pre-log measurements while being highly susceptible to noise. The requisite logarithmic transformation critically amplifies noise within these data, imposing exceptional demands on reconstruction precision. To overcome these challenges, we propose PLOT-CT, a novel framework for Pre-Log vOronoi decomposiTion-assisted CT generation. Our method begins by applying Voronoi decomposition to pre-log sinograms, disentangling the data into distinct underlying components, which are embedded in separate latent spaces. This explicit decomposition significantly enhances the model's capacity to learn discriminative features, directly improving reconstruction accuracy by mitigating noise and preserving information inherent in the pre-log domain. Extensive experiments demonstrate that PLOT-CT achieves state-of-the-art performance, attaining a 2.36dB PSNR improvement over traditional methods at the 1e4 incident photon level in the pre-log domain.

cs.CV↗
arXiv2026

Code2Worlds: Empowering Coding LLMs for 4D World Generation

Achieving spatial intelligence requires moving beyond visual plausibility to build world simulators grounded in physical laws. While coding LLMs have advanced static 3D scene generation, extending this paradigm to 4D dynamics remains a critical frontier. This task presents two fundamental challenges: multi-scale context entanglement, where monolithic generation fails to balance local object structures with global environmental layouts; and a semantic-physical execution gap, where open-loop code generation leads to physical hallucinations lacking dynamic fidelity. We introduce Code2Worlds, a framework that formulates 4D generation as language-to-simulation code generation. First, we propose a dual-stream architecture that disentangles retrieval-augmented object generation from hierarchical environmental orchestration. Second, to ensure dynamic fidelity, we establish a physics-aware closed-loop mechanism in which a PostProcess Agent scripts dynamics, coupled with a VLM-Motion Critic that performs self-reflection to iteratively refine simulation code. Evaluations on the Code4D benchmark show Code2Worlds outperforms baselines with a 41% SGS gain and 49% higher Richness, while uniquely generating physics-aware dynamics absent in prior static methods. Code: https://github.com/AIGeeksGroup/Code2Worlds. Website: https://aigeeksgroup.github.io/Code2Worlds.

cs.CV↗
arXiv2026

Zero-Sacrifice Persistent-Robustness Adversarial Defense for Pre-Trained Encoders

The widespread use of publicly available pre-trained encoders from self-supervised learning (SSL) has exposed a critical vulnerability: their susceptibility to downstream-agnostic adversarial examples (DAEs), which are crafted without knowledge of the downstream tasks but capable of misleading downstream models. While several defense methods have been explored recently, they rely primarily on task-specific adversarial fine-tuning, which inevitably limits generalizability and causes catastrophic forgetting and deteriorates benign performance. Different with previous works, we propose a more rigorous defense goal that requires only a single tuning for diverse downstream tasks to defend against DAEs and preserve benign performance. To achieve this defense goal, we introduce Zero-Sacrifice Persistent-Robustness Adversarial Defense (ZePAD), which is inspired by the inherent sensitivity of neural networks to data characteristics. Specifically, ZePAD is a dual-branch structure, which consists of a Multi-Pattern Adversarial Enhancement Branch (MPAE-Branch) that uses two adversarially fine-tuned encoders to strengthen adversarial resistance. The Benign Memory Preservation Branch (BMP-Branch) is trained on local data to ensure adversarial robustness does not compromise benign performance. Surprisingly, we find that ZePAD can directly detect DAEs by evaluating branch confidence, without introducing any adversarial exsample identification task during training. Notably, by enriching feature diversity, our method enables a single adversarial fine-tuning to defend against DAEs across downstream tasks, thereby achieving persistent robustness. Extensive experiments on 11 SSL methods and 6 datasets validate its effectiveness. In certain cases, it achieves a 29.20% improvement in benign performance and a 73.86% gain in adversarial robustness, highlighting its zero-sacrifice property.

cs.LG↗
arXiv2026

Retrieving the hot circumgalactic medium physics from the X-ray radial profile from eROSITA with an IlustrisTNG-based forward model

Recent eROSITA measurements of the radial profiles of the hot CGM in the Milky-Way stellar mass (MW-mass) regime provide us with a new benchmark to constrain the hot gas around MW-mass central and satellite galaxies and their halo mass distributions. Modelling this rich data set with state-of-the-art hydrodynamical simulations is required to further our understanding of the shortcomings in the current paradigm of galaxy formation and evolution models. We develop forward models for the stacked X-ray radial surface brightness profile measured by eROSITA around MW-mass galaxies. Our model contains two emitting components: hot gas (around central galaxies and satellite galaxies hosted by more massive halos) and X-ray point sources (X-ray binaries and Active Galactic Nuclei). We model the hot gas profile using the TNG300-based products. We generate mock observations with our TNG300-based model (matching stellar mass and redshift with observations) with different underlying halo mass distributions. We show that for the same mean stellar mass, a factor 2x increase in the mean value of the underlying halo mass distribution results in a ~4x increase in the stacked X-ray luminosity from the hot CGM. The point sources are described by a simple point-spread-function (PSF) of eROSITA, and we fit their normalization in this work. Using empirical models to derive a permissible range of AGN and XRB luminosities in the MW-mass X-ray galaxy stack, we choose our forward model best describing the hot CGM for the eROSITA observations. We find that at < 40 kpc from the galaxy centre, the hot CGM from central galaxies and the X-ray point sources emission each account for 40-50% of the total X-ray emission budget. In summary, we show that the gas physics driving the shape of the observed hot CGM (in stellar-mass-selected samples) is tightly correlated by the underlying halo-mass distribution (abridged).

astro-ph.CO↗
arXiv2026

On Reasoning Strength Planning in Large Reasoning Models

Recent studies empirically reveal that large reasoning models (LRMs) can automatically allocate more reasoning strengths (i.e., the number of reasoning tokens) for harder problems, exhibiting difficulty-awareness for better task performance. While this automatic reasoning strength allocation phenomenon has been widely observed, its underlying mechanism remains largely unexplored. To this end, we provide explanations for this phenomenon from the perspective of model activations. We find evidence that LRMs pre-plan the reasoning strengths in their activations even before generation, with this reasoning strength causally controlled by the magnitude of a pre-allocated directional vector. Specifically, we show that the number of reasoning tokens is predictable solely based on the question activations using linear probes, indicating that LRMs estimate the required reasoning strength in advance. We then uncover that LRMs encode this reasoning strength through a pre-allocated directional vector embedded in the activations of the model, where the vector's magnitude modulates the reasoning strength. Subtracting this vector can lead to reduced reasoning token number and performance, while adding this vector can lead to increased reasoning token number and even improved performance. We further reveal that this direction vector consistently yields positive reasoning length prediction, and it modifies the logits of end-of-reasoning token to affect the reasoning length. Finally, we demonstrate two potential applications of our findings: overthinking behavior detection and enabling efficient reasoning on simple problems. Our work provides new insights into the internal mechanisms of reasoning in LRMs and offers practical tools for controlling their reasoning behaviors. Our code is available at https://github.com/AlphaLab-USTC/LRM-plans-CoT.

cs.AI↗
arXiv2026

Mirage: Transmitting a Video as a Perceptual Illusion for 50,000X Speedup

The existing communication framework mainly aims at accurate reconstruction of source signals to ensure reliable transmission. However, this signal-level fidelity-oriented design often incurs high communication overhead and system complexity, particularly in video communication scenarios where mainstream frameworks rely on transmitting visual data itself, resulting in significant bandwidth consumption. To address this issue, we propose a visual data-free communication framework, Mirage, for extremely efficient video transmission while preserving semantic information. Mirage decomposes video content into two complementary components: temporal sequence information capturing motion dynamics and spatial appearance representations describing overall visual structure. Temporal information is preserved through video captioning, while key frames are encoded into compact semantic representations for spatial appearance. These representations are transmitted to the receiver, where videos are synthesized using generative video models. Since no raw visual data is transmitted, Mirage is inherently privacy-preserving. Mirage also supports personalized adaptation across deployment scenarios. The sender, network, and receiver can independently impose constraints on semantic representation, transmission, and generation, enabling flexible trade-offs between efficiency, privacy, control, and perceptual quality. Experimental results in video transmission demonstrate that Mirage achieves up to a 50000X data-level compression speedup over raw video transmission, with gains expected to scale with larger video content sizes.

cs.NI↗
arXiv2026

Learning to Self-Verify Makes Language Models Better Reasoners

Recent large language models (LLMs) achieve strong performance in generating promising reasoning paths for complex tasks. However, despite powerful generation ability, LLMs remain weak at verifying their own answers, revealing a persistent capability asymmetry between generation and self-verification. In this work, we conduct an in-depth investigation of this asymmetry throughout training evolution and show that, even on the same task, improving generation does not lead to corresponding improvements in self-verification. Interestingly, we find that the reverse direction of this asymmetry behaves differently: learning to self-verify can effectively improve generation performance, achieving accuracy comparable to standard generation training while yielding more efficient and effective reasoning traces. Building on this observation, we further explore integrating self-verification into generation training by formulating a multi-task reinforcement learning framework, where generation and self-verification are optimized as two independent but complementary objectives. Extensive experiments across benchmarks and models demonstrate performance gains over generation-only training in both generation and verification capabilities.

cs.CL↗
arXiv2026

CA-YOLO: Cross Attention Empowered YOLO for Biomimetic Localization

In modern complex environments, achieving accurate and efficient target localization is essential in numerous fields. However, existing systems often face limitations in both accuracy and the ability to recognize small targets. In this study, we propose a bionic stabilized localization system based on CA-YOLO, designed to enhance both target localization accuracy and small target recognition capabilities. Acting as the "brain" of the system, the target detection algorithm emulates the visual focusing mechanism of animals by integrating bionic modules into the YOLO backbone network. These modules include the introduction of a small target detection head and the development of a Characteristic Fusion Attention Mechanism (CFAM). Furthermore, drawing inspiration from the human Vestibulo-Ocular Reflex (VOR), a bionic pan-tilt tracking control strategy is developed, which incorporates central positioning, stability optimization, adaptive control coefficient adjustment, and an intelligent recapture function. The experimental results show that CA-YOLO outperforms the original model on standard datasets (COCO and VisDrone), with average accuracy metrics improved by 3.94%and 4.90%, respectively.Further time-sensitive target localization experiments validate the effectiveness and practicality of this bionic stabilized localization system.

cs.CV↗
arXiv2026

The Standard siren tests of viable $f(R)$ cosmologies

We constrain the Hu-Sawicki and Starobinsky $f(R)$ gravity models utilizing current electromagnetic (PP+CC, Planck and DESI2) datasets and simulate standard siren catalogs based on the resulting best-fit parameters. We demonstrate that the simulated SS data provide complementary sensitivity to the modified gravitational wave propagation friction term, thereby enhancing the discriminating power between $f(R)$ gravity and the $Λ$CDM model. However, we note that standard sirens do not offer a viable resolution to the Hubble tension in this analysis, as the inferred constraints are predominantly driven by the fiducial cosmologies adopted in the simulations. Regarding the specific models, we find that for the Hu-Sawicki scenario, several data combinations favor $F_{RR0}<0$, implying potential theoretical instabilities. And, for the Starobinsky model, while EM-only constraints are nearly symmetric between the two parameter branches ($b<0$ and $b>0$), the inclusion of SS constraints introduces mild asymmetries, revealing the sensitivity of SS observables to the curvature dependence of the theory. Future truly independent standard siren observations would be crucial for a definitive assessment of $f(R)$ gravity as an alternative to $Λ$CDM.

astro-ph.CO↗