arXiv subjects

Yu Wang

Publications and source records attributed to Yu Wang.

At least 19 recordsLinked to original sources

arXiv2026

StatLUT: Statistical Feature-Driven Multimodal 3DLUT Generation for Photorealistic Style Transfer

Photorealistic Style Transfer (PST) aims to transfer the color and tonal style of a reference to a content image while strictly preserving its structural integrity. However, existing deep learning-based methods inherently suffer from semantic entanglement caused by pre-trained image encoders, leading to unnatural spatial distortions. Moreover, current pixel-level mapping paradigms often ignore color gamut topology, resulting in color banding, while also lacking the multimodal capability for precise, tone-specific text-driven control. To address these bottlenecks, we propose StatLUT, an innovative multimodal framework for 3D LUT generation. First, we bypass traditional encoders and introduce a Lab-Extractor to derive spatially-agnostic statistical features, fundamentally decoupling color distributions from structural semantics to ensure artifact-free rendering. Second, we formulate LUT generation as a Transformer-based Seq2Seq translation task, utilizing a Multi-dimensional Residual Mapper (MR-Mapper) to predict topologically smooth 3D LUTs. Finally, to break the unimodal barrier, we propose the H-Diffuser, a lightweight Diffusion Transformer that directly synthesizes statistical features from tone-descriptive text prompts, enabling controllable and robust color grading. Extensive experiments on standard benchmarks demonstrate that StatLUT significantly outperforms state-of-the-art methods in both visual quality and quantitative metrics, pioneering a highly robust and versatile paradigm for multimodal photorealistic style transfer.

cs.CV↗

arXiv2026

Rose: Efficient Embodied RL through Graph Optimization and GPU Compute Resource Control

Embodied reinforcement learning (RL) coordinates policy inference, simulation, and model training with highly heterogeneous resource demands. Although individual components frequently underutilize GPU compute during execution, existing systems only overlap work at coarse component boundaries (e.g., pipelining inference and simulation across simulator groups), leaving substantial low-activity GPU windows within subcomponent phases and nested loops. Exploiting these fine-grained windows manually requires tedious dependency analysis and workload-specific coordination. Furthermore, even when independent work is scheduled concurrently, cross-process GPU time slicing and streaming multiprocessor (SM) contention can negate the benefits of overlap. We present Rose, a runtime-aware system that automates fine-grained overlap and GPU resource management for embodied RL while preserving synchronous training semantics. Through a lightweight programmable interface that exposes execution hierarchy and state accesses, Rose tracks runtime data readiness and GPU activity to perform a two-stage optimization. First, a window-guided graph optimizer identifies low-activity windows and applies graph transformations to overlap complementary work across components and loop levels. Second, leveraging NVIDIA Multi-Process Service (MPS) for cross-process GPU concurrency, a slack-aware resource manager tunes process-level SM quotas by exploiting branch synchronization slack to mitigate interference on the critical path. Both stages combine local profiling with end-to-end iteration validation to guarantee throughput gains with minimal tuning overhead. Across diverse policies, simulators, RL algorithms, and GPU platforms, Rose achieves 1.20--1.92$\times$ training throughput improvement and increases average SM activity by 23.8--39.3\% compared with the state-of-the-art systems.

cs.DC↗

arXiv2026

Colorings breaking prescribed families of graph automorphisms

We prove that for every continuous action of a compact group on a discrete set, each finite-color template admits a coloring from arbitrary lists of the same size whose stabilizer is contained in a conjugate of the template stabilizer. Consequently, the ordinary and list parameters coincide whenever the ordinary parameter is finite. In particular, we answer the question of Ferrara, Flesch and Gethner negatively for finite graphs, and establish the finite case of the conjecture of Kwaśny and Stawiski whenever a distinguishing edge coloring exists. We also show that every closed locally compact permutation group with infinite motion on a countably infinite set admits continuum many inequivalent distinguishing colorings from arbitrary 2-element lists. Using compact kernels and finite bases, we extend ordinary and list equality to countable graphs with countable automorphism groups and prove that one additional list color suffices for connected locally finite graphs. Using this multiplicity and a construction fixing all vertices of infinite degree, we prove that $D'_\ell(G)\le 2$ for every countable connected graph with infinite edge motion, settling the list infinite edge-motion conjecture of Kwaśny and Stawiski.

math.CO↗

arXiv2026

Beyond Policy Patches: Distilling RL Specialists into Vision-Language-Action Policies

Vision-language-action (VLA) policies can handle noncritical stages of long-horizon manipulation yet fail during short precision-critical phases. Lightweight reinforcement learning (RL) specialists can address these failures without directly applying RL to the full VLA, but retain auxiliary controllers and switching mechanisms at deployment. We study how to transfer their capabilities into a single flow-matching VLA. Direct imitation of specialist rollouts can reproduce action jitter and leave the VLA unprepared for states encountered during its own execution. We introduce REPAIR, a value-guided distillation framework that combines offline initialization with on-policy refinement. At each critical-phase observation, REPAIR queries the corresponding specialist for multiple action candidates and weights them using its critic to construct a supervision distribution. A shared weighted flow-matching objective trains the VLA first on observations from specialist-assisted rollouts, then on observations collected through its own execution. Specialists provide supervision without taking control during on-policy collection and are removed at deployment. Across six real-robot tasks with 30 evaluation trials per task, REPAIR achieves 100% success on all six tasks, with lower mean end-effector jerk than specialist-assisted execution on every task.

cs.RO↗

arXiv2026

DepthGraft: Structural Regularization through Hierarchical Cross-Layer KV Reconstruction

Scaling depth is a key route to stronger large language models, but Pre-LN architectures often suffer from depth-wise information dilution, limiting the effective use of deep layers. While cross-layer connectivity has proven effective at mitigating this problem, existing designs usually introduce additional computation or memory access. We revisit cross-layer KV sharing as an efficient form of such connectivity. Beyond cache compression, each KV reuse edge creates a direct gradient path from deep layers to source KV projections, turning KV sharing into an implicit structural regularizer for KV representation geometry. We provide a systematic analysis of this mechanism, showing that proper sharing topologies sharpen key retrieval directions while enriching the value content subspace, whereas existing adjacent or global-anchor designs do not fully exploit this effect. Motivated by this, we propose DepthGraft, a hierarchical KV-sharing architecture that reconstructs deep-layer KV from stride-aligned shallow-middle source pairs. DepthGraft distributes non-local gradient flow across source layers, while preserving prefill early completion and reducing KV cache by roughly one third. Experiments on dense and MoE models from 0.5B to 16B parameters show consistent downstream improvements under both GQA and MLA. Notably, on the 16B MoE model, DepthGraft achieves a 7.0% absolute improvement on MMLU-Pro, along with gains of 6.6% on CMMLU and 6.4% on C3. These results validate the potential of DepthGraft as a scalable design for large-scale pretraining. The project page is available at https://github.com/jianke0604/DepthGraft.

cs.CL↗

arXiv2026

Continual Residual Memory for Gradient-Free Test-Time Adaptation in Speech Enhancement

Offline-trained speech enhancement models can degrade under deployment mismatch. Test-time adaptation (TTA) addresses this problem using unlabeled test observations, but existing approaches mainly rely on online parameter optimization and do not explicitly exploit the residual discarded by the enhancement backbone. We propose Continual Residual Memory (CRM), a gradient-free continual TTA framework for selective residual recovery. CRM first estimates bounded time-frequency residual recovery from the current utterance alone and then uses causal prototype memory from preceding utterances to refine this estimate. During deployment, adaptation is driven solely by the evolving memory state while the backbone and CRM modules remain frozen. Across DNS, EARS-D, and Libri-MUSAN, CRM improves all seven reported metrics over frozen CMGAN and achieves the best PESQ, CSIG, and COVL on all three test sets among the compared methods. CRM remains lightweight and shows cross-backbone applicability on five additional backbones. Code is available at https://github.com/Ada312/continual-residual-memory.

eess.AS↗

arXiv2026

Learning from Fleet Experience: Collective Experience Learning Across 96 Robots

Robot deployment generates experience that can support policy improvement, but continued reliance on expert corrective actions makes learning difficult to scale across growing fleets. We propose learning from fleet experience: using deployed robots' own executions as a source of supervision for further policy improvement. We call this paradigm Collective Experience Learning (CEL), in which robots pool execution experience to update a shared policy that returns to the fleet. Following expert-supervised initialization, we realize CEL through failure-guided experience retrieval fine-tuning (FERT). FERT uses failure contexts to retrieve relevant segments from successful trajectories and prioritizes their recorded actions for supervised updates, without collecting additional expert demonstrations or corrective actions. Operators provide binary episode-outcome labels and environment resets. A fault-tolerant asynchronous training system (FATS) coordinates robot execution, cloud inference and policy optimization despite local interruptions. Across five real-world manipulation tasks, the resulting policies achieve success rates of 93.3-99.2%, evaluated over 120 trials per task on robots excluded from training data collection. With collection time per robot held fixed, larger fleets provide more experience and generally improve performance across the evaluated sizes. On multi-stage phone packaging, learning with 96 robots raises success from 79.2% to 98.3%. These results support CEL as a scalable approach to improving robot policy reliability through experience generated during deployment.

cs.RO↗

arXiv2026

From Cross-Section to Cross-Omics: Conditional Cell Nucleus Image Editing in Spatial Transcriptomics

Spatial transcriptomics (ST) platforms simultaneously measure gene expression and capture cell nucleus images at single-cell resolution. A fundamental property of tissue preparation introduces a confounding relationship between the two measurement modalities: sectioning a three-dimensional tissue into a thin slice produces a two-dimensional cross-section whose area depends on the cutting angle. Larger cross-sections expose more cytoplasm, yielding higher mRNA counts, while pathomics features such as nuclear area are measured directly from the same 2D boundary. This sectioning geometry creates an apparent correlation between gene expression and pathomics that reflects the measurement process rather than independent biology. Existing gene-to-image methods generate cell images from gene features but do not support condition-driven editing of existing images across both omics layers. We apply the Manifold-Aware Diffusion (MAD) framework to this dual-omics editing setting, enabling targeted modification of one feature group while preserving the other. Evaluated on podocyte cells from CosMx SMI kidney tissue, MAD achieves MAE=0.6636 and FID=25.19 for pathomics editing, outperforming all compared baselines. A controlled simulation study further confirms that the VAE component in MAD better preserves cross-group independence than CCDM (MAD without VAE).

cs.AI↗

arXiv2026

Stability-Performance Trade-off in Maritime Integrated Sensing, Communication, and Computation: A Lyapunov-Guided Dual-Timescale Hierarchical RL Approach

Driven by the ever-growing demand for intelligent maritime services, cooperative unmanned aerial vehicle (UAV) and unmanned surface vehicle (USV) operations, together with wireless integrated sensing, communication, and computation (ISCC), have become pivotal in maritime environments. However, the distinctive maritime environment precludes a direct application of terrestrial ISCC models, while sensing, communication, and computation compete for the same spatio-temporal resources, thus yielding inherent performance trade-offs that become particularly severe in resource-constrained ocean scenarios. To address these issues, we investigate an ISCC system in which multiple UAVs detect targets at the sea surface and partially offload the collected data to a USV for cooperative processing. Characterizing the maritime-specific environment impairments, we incorporate sea clutter and wave-induced USV roll motion into the sensing and communication models, respectively. Moreover, we introduce data queueing models at both the UAVs and the USV to capture the dynamic coupling among sensing-data generation, wireless offloading, and computation execution. The optimization objective is to minimize the long-term average system energy consumption by jointly optimizing UAV-target association, UAV trajectories, beamforming, time allocation, offloading ratios, and computation resource allocation. Given that these decisions differ in update frequency and time sensitivity, we propose a Lyapunov-guided dual-timescale hierarchical soft actor-critic (LTH-SAC) framework, which transforms the long-term problem into a sequence of queue-aware online subproblems and coordinates optimization across the two timescales. Simulation results verify the effectiveness of the proposed framework and demonstrate its capability to balance the stability-performance trade-off among sensing, communication, and computation.

eess.SP↗

arXiv2026

STRATA: Reinforcement-driven Active Sampling with Tri-paradigm Adaptation for Efficient 3D Spatial Transcriptomics

Spatial transcriptomics (ST) enables joint characterization of tissue morphology and gene expression. However, 3D volumetric ST acquisition remains expensive. AI-driven methods offer a low-cost alternative but remain limited in accuracy, while recent 2D section-wise ST acquisition still rely on dense grid sampling, resulting in redundant measurements and undersampling of biologically informative regions. To bridge the gap, we propose STRATA, a reinforcement-driven active sampling framework with tri-paradigm adaptation for reconstructing dense 3D transcriptomic maps from sparse voxel-wise ST acquisition. STRATA constructs a 3D graph integrating H&E morphology with single-cell priors and employs a reinforcement learning agent to estimate morphological sparsity and transcriptomic informativeness, enabling budget-aware selection of informative spots. Given these sparse measurements, an adaptive tri-paradigm predictor combines retrieval, regression, and cross-layer imputation through lightweight fusion, dynamically weighting their contributions. Experiments across datasets demonstrate improved sampling efficiency and accuracy, particularly under low-budget settings. Code is available at https://github.com/Junchao-Zhu/STRATA.

cs.CV↗

arXiv2026

Agent-Wise Test-Time Normalization for Camera-Based Cooperative Perception with Corrupted Collaborators

Intermediate-fusion cooperative perception lets a vehicle detect objects beyond its own field of view by fusing features shared by nearby agents. When only a collaborator's cameras are degraded by fog, motion blur or exposure errors, the features it shares drift away from the training distribution while the ego input stays clean, and cooperation can become worse than not cooperating at all. Test-time batch-normalization adaptation, the standard training-free remedy, estimates statistics over the pooled batch of all agents and therefore normalizes both the clean ego and the corrupted collaborator with a mixture of two distributions. We show that this mixing harms even the clean ego: when the ego shares BN statistics with a corrupted collaborator, its detection from its own features drops by 3.58 to 6.68 AP@0.5 points although its cameras are untouched. We propose AgentNorm, an agent-wise test-time normalization that i) estimates the normalization statistics of every collaborator from its own current observation and ii) anchors the ego to its source statistics, so that the ego path remains exactly the original model. AgentNorm has no learnable parameters, needs no gradients or labels, is independent of the order of test frames, and runs locally on each agent without changing the transmitted features. On camera-based OPV2V with four fusion checkpoints and collaborator-only corruptions, AgentNorm raises AP@0.5 under severity-5 fog from 29.42-37.25 to 43.32-58.86 and outperforms pooled normalization by 5.53-12.94 points under fog and motion blur, at an extra latency of 2.3 ms per frame. On the real-world DAIR-V2X-C dataset, it outperforms pooled normalization in all ten infrastructure-corruption settings. We also report the cases it does not fix: occlusion, which removes image content rather than shifting feature statistics, and the case in which all agents share the same shift.

cs.CV↗

arXiv2026

A Transfer Principle for Elliptic Carleman Estimates: Application to Discrete Calderón Stability

We develop a transfer principle that lifts continuous elliptic Carleman estimates to mesh-uniform estimates for conforming Galerkin discretizations. The main difficulties are twofold: discrete states generally lack the $H^2$-regularity required by the continuous estimate, and the Carleman parameter $s$ must be controlled relative to the mesh size $h$. We overcome both by introducing a shifted continuous problem whose Galerkin solution is exactly the conforming reconstruction of a prescribed discrete state. Weighted approximation then transfers the inequality back to the original discrete operator without adding stabilization, and yields the range $s\le ch^{-4/5}$ for radial weights. We verify the framework for piecewise affine finite elements with consistent or lumped mass on fitted, uniformly shape-regular meshes, and for standard Cartesian finite differences through their finite-element realization. As an application, we study the discrete Calderón problem for bounded complex Schrödinger potentials in dimension $d\ge3$. Assuming agreement in a fixed connected boundary neighborhood and uniform stability of the discrete Dirichlet problems, we obtain stability from Dirichlet inputs and variational flux observations on independently prescribed, possibly disjoint boundary patches. The resulting estimate in $H^{-r}$, $r>d/2$, retains the single logarithmic data term and the $O(h^{1/2})$ mesh remainder of the full-data estimate.

math.NA↗

arXiv2026

Interpretability Is Not Just a Tax: Sparsity and Diversity Yield Legible and Composable Reward Experts for Personalization

Preference modeling aims to capture heterogeneous human values and increasingly relies on learning latent components to represent this variation without fine-grained supervision. What these components actually represent, however, is rarely examined. We audit Mixture-of-Experts (MoE) reward models along three rungs---semantic coherence, functional correspondence, and composability---in controlled and real-world settings, holding the architecture and training data fixed so that differences isolate the effect of the training objective. Our diagnosis is that the components conventional MoE reward models learn are entangled, which limits not only interpretability but also the personalization such models are built for. The entanglement is not intrinsic to the architecture but a property of the training objective: off-the-shelf sparsity and diversity regularizers already substantially reduce it. The resulting configuration, which we call sparse MoE, recovers all underlying semantic categories without domain supervision in a controlled setting, reaching 84.8% expert purity against 60.8% for conventional MoE; on real-world data its experts admit human-interpretable task- and domain-level descriptions aligned with their functional strengths. Targeted interventions on routing weights produce predictable preference changes, showing that these interpretations are behaviorally grounded rather than merely descriptive. Crucially, the disentanglement induced by this objective also makes the experts composable: lightweight router updates yield a 25.81-point improvement in few-shot personalization with 50 examples, roughly 2.7 times the gain of the identical architecture and adaptation procedure applied to entangled experts. Together, these results show that MoE reward models can learn interpretable expert structure, and that the same structure can substantially improve personalization.

cs.LG↗

arXiv2026

The Anatomy of Socio-Demographics in Pluralistic Offensiveness Modeling: When and Where It Helps, and the Role of Reasoning

Subjective judgments such as offensiveness often vary substantially across individuals, motivating pluralistic modeling approaches that incorporate annotators' socio-demographic information to capture this variation. However, prior findings on the benefits of socio-demographic information for pluralistic modeling are mixed, calling for a systematic analysis of both its effectiveness and the mechanisms through which it improves prediction. In this work, we study $\textbf{when}$ and $\textbf{where}$ socio-demographics benefit pluralistic offensiveness modeling, as well as $\textbf{the role of explicit reasoning}$ in leveraging this information. Across three multi-annotator datasets with prompted and trained LLMs, we first show that demographic gains are small and inconsistent for individual predictions but become clearer at the group and population levels, because aggregation attenuates the idiosyncratic variation that masks small but systematic demographic signals. Second, using a reasoning and prediction format that explicitly decomposes individual judgment into an overall baseline and a demographic-specific deviation, we show that the gains arise from both components: demographic information improves baseline estimation and enables meaningful demographic-specific shifts, and interventions show that better component estimates improve final predictions. Finally, comparing direct demographic conditioning with explicit reasoning about how the text interacts with annotator demographics, we find that reasoning yields more accurate baseline and shift estimates as well as final predictions, whereas conditioning alone rarely produces informative shifts. These results suggest that the mixed evidence in prior work may partly stem from individual-level evaluation, and that the value of socio-demographic information depends on both how it is evaluated and how models are guided to use it.

cs.AI↗

arXiv2026

New families of additive non-RS MDS codes and ACD codes from twisted Reed-Solomon codes

Additive codes generalize linear codes and play an important role in the construc?tion of quantum codes. As generalizations of linear MDS codes, additive MDS codes have received much research attention. As an important subclass of additive codes, ACD codes are widely used for quantum error correction. In this paper, we study single-twist additive generalized twisted Reed-Solomon (GTRS) codes over finite fields. We derive necessary and sufficient conditions for these codes to be MDS for arbitrary hook and twist parameters. Then we employ Schur squares to prove that such additive MDS codes are not monomially equivalent to additive Reed-Solomon codes. We also identify a class of additive GTRS codes whose duals remain additive GTRS codes, and show that, for all sufficiently large prime powers q whose characteristic avoids a finite set, such GTRS codes yield ACD codes.

cs.IT↗

arXiv2026

Giant third-order nonlinearity amplifies second-harmonic transport in Ca3Ru2O7

Nonlinear electrical transport enables the conversion and detection of electromagnetic signals, supporting wireless energy harvesting, sensing and frequency conversion. Achieving both a large response and electrical tunability remains a central materials challenge. Here we harness giant third-order nonlinearity to amplify and electrically control second-harmonic transport in the correlated polar metal Ca3Ru2O7. Superposing a dc bias on an ac excitation enhances the second-harmonic signal by more than an order of magnitude. The bias controls the polarity of the induced response and activates channels forbidden by symmetry at zero bias. Both transverse and longitudinal components scale quadratically with the ac excitation and are odd in the dc bias, with angular dependences consistent with third-order conductivity. The underlying third-harmonic electric-field responsivity reaches 4.5 \times 10-6 m2V-2 at 3 K, providing a quantitative measure of the strong nonlinearity. The induced response emerges below the spin-reorientation transition near 48 K, linking its enhancement to the reconstructed low-temperature electronic state. Theoretical calculations identify quantum-geometric contributions to the nonlinear conductivity and reproduce the observed response symmetry. These results demonstrate how strong higher-order nonlinearity can amplify an effective second-order response, providing a route to electrically tunable nonlinear transport in correlated materials.

cond-mat.other↗

arXiv2026

Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks

As frontier language models grow increasingly capable and autonomous, independent third-party evaluations are critical for characterizing their systemic safety risks. While closed-source releases are routinely accompanied by comprehensive system cards, open-weight models receive less scrutiny regarding frontier risks. To address this gap, we present an independent safety evaluation of DeepSeek V4.1 Flash-1017, the latest version of V4.1 Flash that departs from V4 with a new architecture, scaled-up training, and strong agentic capabilities. Drawing on the EU General-Purpose AI Code of Practice, we evaluate the model across 34 benchmarks, including 5 in-house suites, spanning evaluation integrity, loss of control, and harmful manipulation. Conforming to the AEF-1 standard, we compare V4.1 Flash-1017 against V4.1 Flash, earlier DeepSeek V4 models, and leading closed-source systems (GPT-6 Astra, GPT-5.6 Sol, and Gemini 3.8 Flash). Overall, V4.1 Flash-1017 shows substantial safety improvements over the V4 generation and approaches or matches leading closed-source baselines on several dimensions, while still exhibiting notable behaviors that warrant continued attention. Building on our findings, we highlight three key observations and recommendations. First, because safety can shift and even regress across minor updates, developers should integrate safety evaluation into incremental releases. Second, as final-turn outputs and end-state checks alone cannot reliably detect loss-of-control behaviors, deployments require process-level observability across both reasoning traces and action logs. Third, because public benchmarks elicit strong evaluation awareness and face data contamination risks, developing private benchmarks is essential for trustworthy frontier evaluation.

cs.CL↗

arXiv2026

Evaluating the Structural Basis for Polar Altermagnet Candidate Ca$_{3}$(Ru,Ti)$_{2}$O$_{7}$

The interplay between polar and altermagnetic orders remains largely unexplored in the broader landscape of correlated electron systems. Density functional theory (DFT) within the PBEsol + $U$ treatment propose Ca$_{3}$Ru$_{2}$O$_{7}$ as a polar altermagnet, reliant on the transformation of the experimentally reported $Bb2_{1}m$ phase to a lower symmetry $Pn2_{1}a$ structure. Here, we perform a targeted search for the $Pn2_{1}a$ phase using synchrotron X-ray diffraction on single crystals of Ca$_{3}$Ru$_{2}$O$_{7}$ and Ca$_{3}$(Ru$_{0.99}$Ti$_{0.01}$)$_{2}$O$_{7}$. No diffraction signature of the $Pn2_{1}a$ structure is detected down to 20 K within experimental limits. Through diffraction simulations, we translate this result into a quantitative upper bound on the symmetry breaking distortion amplitude, which we find to be significantly smaller than the DFT predicted distortion. Our DFT calculations further motivate dilute Ti substitution ($\lesssim 2\%$) as an alternative route to stabilize the low symmetry phase, for which we again find no clear evidence for this putative $Pn2_{1}a$ structure that hosts altermagnetism. Our findings indicate that PBEsol + $U$ over-stabilizes a $Pn2_{1}a$ distortion, and the measured bounds provide benchmark for future calculations. Ti substitution exceeding $\sim$3% reveals a chemically tunable polar altermagnetic phase within the $Bb2_{1}m$ structure and we theoretically demonstrate that the spin-splitting can be tuned by Ti substitution. By combining DFT, diffraction simulations and synchrotron X-ray measurements, our study demonstrates a general methodology to quantify subtle structural distortions that underlie emergent electronic and magnetic phases.

cond-mat.str-el↗