Searcharxiv⌕ Search

arXiv subjects

Zhiqiang Shen

Publications and source records attributed to Zhiqiang Shen.

At least 55 records · Page 3Linked to original sources
arXiv2026

Verification of the Polarimetric Capability of the East Asia VLBI Network

The East Asia VLBI Network (EAVN) has recently enabled dual-polarization observations at $22$ and $43\,\mathrm{GHz}$. We present the first systematic verification of its polarimetric performance using EAVN observations of M87, 3C 279, 3C 273, and OJ 287, calibrated with the GPCAL pipeline and evaluated against near-contemporaneous VLBA images at comparable frequencies. Most stations show stable polarimetric leakages with amplitudes of $5$-$10\%$ over monthly timescales. While several VERA stations exhibit D-term phase variations between epochs, we attribute these to field-rotator (FR) offsets and demonstrate that phase stability is restored after applying the analytically derived FR corrections. The resulting linear-polarization morphologies and EVPAs broadly agree with the VLBA results within uncertainties; fractional polarization measured by the EAVN tends to be slightly higher near polarization peaks. Although exact one-to-one comparisons are limited by moderate frequency and epoch differences, the combined evidence indicates robust EAVN polarimetric calibration and imaging capabilities at $22$ and $43\,\mathrm{GHz}$. These results support the scientific capability of EAVN polarimetry and lay the groundwork for expanded, higher-fidelity polarimetric studies in East Asia.

astro-ph.IM↗
arXiv2026

From Masks to Pixels and Meaning: A New Taxonomy, Benchmark, and Metrics for VLM Image Tampering

Existing tampering detection benchmarks largely rely on object masks, which severely misalign with the true edit signal: many pixels inside a mask are untouched or only trivially modified, while subtle yet consequential edits outside the mask are treated as natural. We reformulate VLM image tampering from coarse region labels to a pixel-grounded, meaning and language-aware task. First, we introduce a taxonomy spanning edit primitives (replace/remove/splice/inpaint/attribute/colorization, etc.) and their semantic class of tampered object, linking low-level changes to high-level understanding. Second, we release a new benchmark with per-pixel tamper maps and paired category supervision to evaluate detection and classification within a unified protocol. Third, we propose a training framework and evaluation metrics that quantify pixel-level correctness with localization to assess confidence or prediction on true edit intensity, and further measure tamper meaning understanding via semantics-aware classification and natural language descriptions for the predicted regions. We also re-evaluate the existing strong segmentation/localization baselines on recent strong tamper detectors and reveal substantial over- and under-scoring using mask-only metrics, and expose failure modes on micro-edits and off-mask changes. Our framework advances the field from masks to pixels, meanings and language descriptions, establishing a rigorous standard for tamper localization, semantic classification and description. Code and benchmark data are available at https://github.com/VILA-Lab/PIXAR.

cs.CV↗
arXiv2026

Multiscale Structure-Guided Latent Diffusion for Multimodal MRI Translation

Although diffusion models have achieved remarkable progress in multi-modal magnetic resonance imaging (MRI) translation tasks, existing methods still tend to suffer from anatomical inconsistencies or degraded texture details when handling arbitrary missing-modality scenarios. To address these issues, we propose a latent diffusion-based multi-modal MRI translation framework, termed MSG-LDM. By leveraging the available modalities, the proposed method infers complete structural information, which preserves reliable boundary details. Specifically, we introduce a style--structure disentanglement mechanism in the latent space, which explicitly separates modality-specific style features from shared structural representations, and jointly models low-frequency anatomical layouts and high-frequency boundary details in a multi-scale feature space. During the structure disentanglement stage, high-frequency structural information is explicitly incorporated to enhance feature representations, guiding the model to focus on fine-grained structural cues while learning modality-invariant low-frequency anatomical representations. Furthermore, to reduce interference from modality-specific styles and improve the stability of structure representations, we design a style consistency loss and a structure-aware loss. Extensive experiments on the BraTS2020 and WMH datasets demonstrate that the proposed method outperforms existing MRI synthesis approaches, particularly in reconstructing complete structures. The source code is publicly available at https://github.com/ziyi-start/MSG-LDM.

eess.IV↗
arXiv2026

BiGain: Unified Token Compression for Joint Generation and Classification

Acceleration methods for diffusion models (e.g., token merging or downsampling) typically optimize synthesis quality under reduced compute, yet often ignore discriminative capacity. We revisit token compression with a joint objective and present BiGain, a training-free, plug-and-play framework that preserves generation quality while improving classification in accelerated diffusion models. Our key insight is frequency separation: mapping feature-space signals into a frequency-aware representation disentangles fine detail from global semantics, enabling compression that respects both generative fidelity and discriminative utility. BiGain reflects this principle with two frequency-aware operators: (1) Laplacian-gated token merging, which encourages merges among spectrally smooth tokens while discouraging merges of high-contrast tokens, thereby retaining edges and textures; and (2) Interpolate-Extrapolate KV Downsampling, which downsamples keys/values via a controllable interextrapolation between nearest and average pooling while keeping queries intact, thereby conserving attention precision. Across DiT- and U-Net-based backbones and ImageNet-1K, ImageNet-100, Oxford-IIIT Pets, and COCO-2017, our operators consistently improve the speed-accuracy trade-off for diffusion-based classification, while maintaining or enhancing generation quality under comparable acceleration. For instance, on ImageNet-1K, with 70% token merging on Stable Diffusion 2.0, BiGain increases classification accuracy by 7.15% while improving FID by 0.34 (1.85%). Our analyses indicate that balanced spectral retention, preserving high-frequency detail and low/mid-frequency semantics, is a reliable design rule for token compression in diffusion models. To our knowledge, BiGain is the first framework to jointly study and advance both generation and classification under accelerated diffusion, supporting lower-cost deployment.

cs.CV↗
arXiv2026

Diff-ES: Stage-wise Structural Diffusion Pruning via Evolutionary Search

Diffusion models have achieved remarkable success in high-fidelity image generation but remain computationally demanding due to their multi-step denoising process and large model sizes. Although prior work improves efficiency either by reducing sampling steps or by compressing model parameters, existing structured pruning approaches still struggle to balance real acceleration and image quality preservation. In particular, prior methods such as MosaicDiff rely on heuristic, manually tuned stage-wise sparsity schedules and stitch multiple independently pruned models during inference, which increases memory overhead. However, the importance of diffusion steps is highly non-uniform and model-dependent. As a result, schedules derived from simple heuristics or empirical observations often fail to generalize and may lead to suboptimal performance. To this end, we introduce \textbf{Diff-ES}, a stage-wise structural \textbf{Diff}usion pruning framework via \textbf{E}volutionary \textbf{S}earch, which optimizes the stage-wise sparsity schedule and executes it through memory-efficient weight routing without model duplication. Diff-ES divides the diffusion trajectory into multiple stages, automatically discovers an optimal stage-wise sparsity schedule via evolutionary search, and activates stage-conditioned weights dynamically without duplicating model parameters. Our framework naturally integrates with existing structured pruning methods for diffusion models including depth and width pruning. Extensive experiments on DiT and SDXL demonstrate that Diff-ES consistently achieves wall-clock speedups while incurring minimal degradation in generation quality, establishing state-of-the-art performance for structured diffusion model pruning.

cs.CV↗
arXiv2026

Exploring 3D Dataset Pruning

Dataset pruning has been widely studied for 2D images to remove redundancy and accelerate training, while particular pruning methods for 3D data remain largely unexplored. In this work, we study dataset pruning for 3D data, where its observed common long-tail class distribution nature make optimization under conventional evaluation metrics Overall Accuracy (OA) and Mean Accuracy (mAcc) inherently conflicting, and further make pruning particularly challenging. To address this, we formulate pruning as approximating the full-data expected risk with a weighted subset, which reveals two key errors: coverage error from insufficient representativeness and prior-mismatch bias from inconsistency between subset-induced class weights and target metrics. We propose representation-aware subset selection with per-class retention quotas for long-tail coverage, and prior-invariant teacher supervision using calibrated soft labels and embedding-geometry distillation. The retention quota also serves as a switch to control the OA-mAcc trade-off. Extensive experiments on 3D datasets show that our method can improve both metrics across multiple settings while adapting to different downstream preferences. Our code is available at https://github.com/XiaohanZhao123/3D-Dataset-Pruning.

cs.CV↗
arXiv2026

ParamMem: Augmenting Language Agents with Parametric Reflective Memory

Self-reflection enables language agents to iteratively refine solutions, yet often produces repetitive outputs that limit reasoning performance. Recent studies have attempted to address this limitation through various approaches, among which increasing reflective diversity has shown promise. Our empirical analysis reveals a strong positive correlation between reflective diversity and task success, further motivating the need for diverse reflection signals. We introduce ParamMem, a parametric memory module that encodes cross-sample reflection patterns into model parameters, enabling diverse reflection generation through temperature-controlled sampling. Building on this module, we propose ParamAgent, a reflection-based agent framework that integrates parametric memory with episodic and cross-sample memory. Extensive experiments on code generation, mathematical reasoning, and multi-hop question answering demonstrate consistent improvements over state-of-the-art baselines. Further analysis reveals that ParamMem is sample-efficient, enables weak-to-strong transfer across model scales, and supports self-improvement without reliance on stronger external model, highlighting the potential of ParamMem as an effective component for enhancing language agents.

cs.LG↗
arXiv2026

HieraMAS: Optimizing Intra-Node LLM Mixtures and Inter-Node Topology for Multi-Agent Systems

Multi-agent systems (MAS) built on large language models (LLMs) have shown strong performance across many tasks. Most existing approaches improve only one aspect at a time, such as the communication topology, role assignment, or LLM routing, while treating each agent as a single, indivisible unit. This misses the opportunity to use mixtures of LLMs within an agent to strengthen role-specific abilities. We propose HieraMAS, a hierarchical collaboration framework that combines intra-node LLM mixtures with an inter-node communication topology. HieraMAS introduces supernodes, where each functional role is implemented by multiple heterogeneous LLMs using a propose-synthesis structure. Optimizing HieraMAS creates unique credit-assignment challenges: final task performance depends heavily on the underlying LLMs' capabilities, which can lead reinforcement methods to incorrectly reward suboptimal configurations. To address this, we use a two-stage algorithm: (1) multi-level reward attribution, which provides fine-grained feedback at both the node level and the overall system level; (2) graph classification for topology selection, which treats choosing the communication structure as a holistic decision rather than optimizing edges one by one. Experiments on reasoning and coding benchmarks show that HieraMAS substantially outperforms existing methods while also delivering better cost-performance trade-offs.

cs.MA↗
arXiv2026

Sink-Aware Pruning for Diffusion Language Models

Diffusion Language Models (DLMs) incur high inference cost due to iterative denoising, motivating efficient pruning. Existing pruning heuristics largely inherited from autoregressive (AR) LLMs, typically preserve attention sink tokens because AR sinks serve as stable global anchors. We show that this assumption does not hold for DLMs: the attention-sink position exhibits substantially higher variance over the full generation trajectory (measured by how the dominant sink locations shift across timesteps), indicating that sinks are often transient and less structurally essential than in AR models. Based on this observation, we propose ${\bf \texttt{Sink-Aware Pruning}}$, which automatically identifies and prunes unstable sinks in DLMs (prior studies usually keep sinks for AR LLMs). Without retraining, our method achieves a better quality-efficiency trade-off and outperforms strong prior pruning baselines under matched compute. Our code is available at https://github.com/VILA-Lab/Sink-Aware-Pruning.

cs.CL↗
arXiv2026

Fast and Scalable Analytical Diffusion

Analytical diffusion models offer a mathematically transparent path to generative modeling by formulating the denoising score as an empirical-Bayes posterior mean. However, this interpretability comes at a prohibitive cost: the standard formulation necessitates a full-dataset scan at every timestep, scaling linearly with dataset size. In this work, we present the first systematic study addressing this scalability bottleneck. We challenge the prevailing assumption that the entire training data is necessary, uncovering the phenomenon of Posterior Progressive Concentration: the effective golden support of the denoising score is not static but shrinks asymptotically from the global manifold to a local neighborhood as the signal-to-noise ratio increases. Capitalizing on this, we propose Dynamic Time-Aware Golden Subset Diffusion (GoldDiff), a training-free framework that decouples inference complexity from dataset size. Instead of static retrieval, GoldDiff uses a coarse-to-fine mechanism to dynamically pinpoint the ''Golden Subset'' for inference. Theoretically, we derive rigorous bounds guaranteeing that our sparse approximation converges to the exact score. Empirically, GoldDiff achieves a $\bf 71 \times$ speedup on AFHQ while matching or achieving even better performance than full-scan baselines. Most notably, we demonstrate the first successful scaling of analytical diffusion to ImageNet-1K, unlocking a scalable, training-free paradigm for large-scale generative modeling.

cs.LG↗
arXiv2026

Dataset Distillation via Committee Voting

Dataset distillation aims to synthesize a compact yet representative dataset that preserves the essential characteristics of the original data for efficient model training. Existing methods mainly focus on improving data-synthetic alignment or scaling distillation to large datasets. In this work, we propose $\textbf{C}$ommittee $\textbf{V}$oting for $\textbf{D}$ataset $\textbf{D}$istillation ($\textbf{CV-DD}$), an orthogonal approach that leverages the collective knowledge of multiple models to produce higher-quality distilled data. We first establish a strong baseline that achieves state-of-the-art performance through modern architectural and optimization choices. By integrating distributions and predictions from multiple models and generating high-quality soft labels, our method captures a broader range of data characteristics, reduces model-specific bias and the impact of distribution shifts, and significantly improves generalization. This voting-based strategy enhances diversity and robustness, alleviates overfitting, and improves post-evaluation performance. Extensive experiments across multiple datasets and IPC settings demonstrate that CV-DD consistently outperforms single- and multi-model distillation methods and generalizes well to non-training-based frameworks and challenging synthetic-to-real transfer tasks. Code is available at: https://github.com/Jiacheng8/CV-DD.

cs.CV↗
arXiv2026

The Interstellar Scintillation of the Radio-Loud Magnetar XTE J1810-197

We present a comprehensive interstellar scintillation (ISS) study of the radio-loud magnetar XTE~J1810$-$197, based on six years of multi-frequency monitoring (2018$-$2024) with the Shanghai Tian Ma Radio Telescope (TMRT) at 7.0, 8.6, and 14.0~GHz. The scintillation parameters--decorrelation bandwidth $Δν_{\rm d}$, decorrelation time $Δτ_{\rm d}$, and drift rate $dt/dν$--are fully characterized. Our measured $Δτ_{\rm d}$ implies $Δτ_{\rm d} < 4$~s at 575-725~MHz under a Kolmogorov spectrum, which is shorter than the magnetar's 5.54~s spin period. This result naturally explains the previously reported absence of pulse-to-pulse coherence at these frequencies. Kinematic modeling locates the dominant scattering screen at $1.6\pm0.1$~kpc away from the Earth, within the Sagittarius Arm. The screen coincides with the HII region JCMTSE~J180921.2$-$201932 and is unrelated to the magnetar's 2018 outburst suggested by earlier studies. A scintillation arc detected at 14.0~GHz represents the highest-frequency arc observed to date. The asymmetry of arcs is linearly correlated with a dispersion-measure gradient across the screen ($r = 0.959$, $p < 10^{-8}$). We also measure its refractive scintillation timescale, which is only $1.21\pm0.19$~d. Clear DISS at 14~GHz effectively resolves the debate over a possible strong-to-weak scattering transition at this frequency. These results extend the ISS characterization of magnetars to previously unexplored frequencies and provide a precise probe of the ionized interstellar medium in the Sagittarius Arm.

astro-ph.HE↗
arXiv2026

ShapeCond: Fast Shapelet-Guided Dataset Condensation for Time Series Classification

Time series data supports many domains (e.g., finance and climate science), but its rapid growth strains storage and computation. Dataset condensation can alleviate this by synthesizing a compact training set that preserves key information. Yet most condensation methods are image-centric and often fail on time series because they miss time-series-specific temporal structure, especially local discriminative motifs such as shapelets. In this work, we propose ShapeCond, a novel and efficient condensation framework for time series classification that leverages shapelet-based dataset knowledge via a shapelet-guided optimization strategy. Our shapelet-assisted synthesis cost is independent of sequence length: longer series yield larger speedups in synthesis (e.g., 29$\times$ faster over prior state-of-the-art method CondTSC for time-series condensation, and up to 10,000$\times$ over naively using shapelets on the Sleep dataset with 3,000 timesteps). By explicitly preserving critical local patterns, ShapeCond improves downstream accuracy and consistently outperforms all prior state-of-the-art time series dataset condensation methods across extensive experiments. Code is available at https://github.com/lunaaa95/ShapeCond.

cs.LG↗
arXiv2026

Duality Models: An Embarrassingly Simple One-step Generation Paradigm

Consistency-based generative models like Shortcut and MeanFlow achieve impressive results via a target-aware design for solving the Probability Flow ODE (PF-ODE). Typically, such methods introduce a target time $r$ alongside the current time $t$ to modulate outputs between a local multi-step derivative ($r = t$) and a global few-step integral ($r = 0$). However, the conventional "one input, one output" paradigm enforces a partition of the training budget, often allocating a significant portion (e.g., 75% in MeanFlow) solely to the multi-step objective for stability. This separation forces a trade-off: allocating sufficient samples to the multi-step objective leaves the few-step generation undertrained, which harms convergence and limits scalability. To this end, we propose Duality Models (DuMo) via a "one input, dual output" paradigm. Using a shared backbone with dual heads, DuMo simultaneously predicts velocity $v_t$ and flow-map $u_t$ from a single input $x_t$. This applies geometric constraints from the multi-step objective to every sample, bounding the few-step estimation without separating training objectives, thereby significantly improving stability and efficiency. On ImageNet 256 $\times$ 256, a 679M Diffusion Transformer with SD-VAE achieves a state-of-the-art (SOTA) FID of 1.79 in just 2 steps. Code is available at: https://github.com/LINs-lab/DuMo

cs.LG↗
arXiv2026

CH3CCH as a thermometer in warm molecular gas

Kinetic temperature is a fundamental parameter in molecular clouds. Symmetric top molecules, such as NH$_3$ and CH$_3$CCH, are often used as thermometers. However, at high temperatures, NH$_3$(2,2) can be collisionally excited to NH$_3$(2,1) and rapidly decay to NH$_3$(1,1), which can lead to an underestimation of the kinetic temperature when using rotation temperatures derived from NH$_3$(1,1) and NH$_3$(2,2). In contrast, CH$_3$CCH is a symmetric top molecule with lower critical densities of its rotational levels than those of NH$_3$, which can be thermalized close to the kinetic temperature at relatively low densities of about 10$^{4}$ cm$^{-3}$. To compare the rotation temperatures derived from NH$_3$(1,1)$\&$(2,2) and CH$_3$CCH rotational levels in warm molecular gas, we used observational data toward 55 massive star-forming regions obtained with Yebes 40m and TMRT 65m. Our results show that rotation temperatures derived from NH$_3$(1,1)$\&$(2,2) are systematically lower than those from CH$_3$CCH 5-4. This suggests that CH$_3$CCH rotational lines with the same $J$+1$\rightarrow$$J$ quantum number may be a more reliable thermometer than NH$_3$(1,1)$\&$(2,2) in warm molecular gas located in the surroundings of massive young stellar objects or, more generally, in massive star-forming regions.

astro-ph.GA↗
arXiv2026

Does the radio-active phase of XTE~J1810$-$197 recur following the same evolutionary pattern?

Magnetars are the most strongly magnetized compact objects known in the Universe and are regarded as one of the primary engines powering a variety of enigmatic, high-energy transients. However, our understanding of magnetars remains highly limited, constrained by observational sample size and radiative variability. XTE~J1810$-$197, which re-entered a radio-active phase in 2018, is one of only six known radio-pulsating magnetars. Leveraging the distinctive capability for simultaneous dual-frequency observations, we utilized the Shanghai Tianma Radio Telescope (TMRT) to monitor this magnetar continuously at both 2.25 and 8.60~GHz, capturing its entire evolution from radio activation to quenching. This enabled precise characterization of the evolution in its integrated profile, spin frequency, flux density, and spectral index ($α$, defined by $S \propto f^α$). The first time derivative of its spin frequency $\dotν$ passed through four distinct phases -- rapid decrease, violent oscillation, steady decline, and stable recovery -- before returning to its pre-outburst value concomitant with the cessation of radio emission. Remarkably, both the amplitudes and the characteristic time-scales of these $\dotν$ variations match those observed during the previous outburst that began in 2003, providing the first demonstration that post-outburst rotational evolution and radiative behavior in a magnetar are repeatable. A twisted-magnetosphere model can qualitatively account for this repeatability as well as for the progressive narrowing and abrupt disappearance of the radio pulse radiation, thereby receiving strong observational support.

astro-ph.HE↗
arXiv2025

Attention Is All You Need for KV Cache in Diffusion LLMs

This work studies how to adaptively recompute key-value (KV) caches for diffusion large language models (DLMs) to maximize prediction accuracy while minimizing decoding latency. Prior methods' decoders recompute QKV for all tokens at every denoising step and layer, despite KV states changing little across most steps, especially in shallow layers, leading to substantial redundancy. We make three observations: (1) distant ${\bf MASK}$ tokens primarily act as a length-bias and can be cached block-wise beyond the active prediction window; (2) KV dynamics increase with depth, suggesting that selective refresh starting from deeper layers is sufficient; and (3) the most-attended token exhibits the smallest KV drift, providing a conservative lower bound on cache change for other tokens. Building on these, we propose ${\bf Elastic-Cache}$, a training-free, architecture-agnostic strategy that jointly decides ${when}$ to refresh (via an attention-aware drift test on the most-attended token) and ${where}$ to refresh (via a depth-aware schedule that recomputes from a chosen layer onward while reusing shallow-layer caches and off-window MASK caches). Unlike fixed-period schemes, Elastic-Cache performs adaptive, layer-aware cache updates for diffusion LLMs, reducing redundant computation and accelerating decoding with negligible loss in generation quality. Experiments on LLaDA-Instruct, LLaDA-1.5, and LLaDA-V across mathematical reasoning and code generation tasks demonstrate consistent speedups: $8.7\times$ on GSM8K (256 tokens), and $45.1\times$ on longer sequences, while consistently maintaining higher accuracy than the baseline. Our method achieves significantly higher throughput ($6.8\times$ on GSM8K) than existing confidence-based approaches while preserving generation quality, enabling practical deployment of diffusion LLMs.

cs.CL↗
arXiv2025

Do Not Merge My Model! Safeguarding Open-Source LLMs Against Unauthorized Model Merging

Model merging has emerged as an efficient technique for expanding large language models (LLMs) by integrating specialized expert models. However, it also introduces a new threat: model merging stealing, where free-riders exploit models through unauthorized model merging. Unfortunately, existing defense mechanisms fail to provide effective protection. Specifically, we identify three critical protection properties that existing methods fail to simultaneously satisfy: (1) proactively preventing unauthorized merging; (2) ensuring compatibility with general open-source settings; (3) achieving high security with negligible performance loss. To address the above issues, we propose MergeBarrier, a plug-and-play defense that proactively prevents unauthorized merging. The core design of MergeBarrier is to disrupt the Linear Mode Connectivity (LMC) between the protected model and its homologous counterparts, thereby eliminating the low-loss path required for effective model merging. Extensive experiments show that MergeBarrier effectively prevents model merging stealing with negligible accuracy loss.

cs.CR↗