Searcharxiv⌕ Search

arXiv subjects

Ling Wang

Publications and source records attributed to Ling Wang.

At least 19 recordsLinked to original sources
arXiv2026

OmniReasoning: Pushing the Limits of Audio-Visual Joint Reasoning

Recent advances have enabled unified omni-modal models in understanding audio, vision, and language. However, existing benchmarks, training data, and learning methods largely treat the modalities independently, leaving the capability of audio-visual joint reasoning poorly evaluated and insufficiently elicited. We address this gap with a benchmark, data engine, and learning method. First, we introduce OmniReasoningBench, a benchmark where both audio and visual evidence are indispensable. It comprises 1,150 multiple-choice and open-ended questions across two tasks, reasoning over video and reasoning beyond video. Second, we develop a data engine OmniQA. It automatically constructs evidence-grounded QA pairs that explicitly necessitate audio-visual joint reasoning, together with time-stamped clue chains that guide the annotation of thinking process. Besides our benchmark, this engine produces training data OmniReasoning-SFT-112K and OmniReasoning-RL-19K. Finally, we propose an on-policy self-distillation method Modality-Factored Self-Distillation (MFSD). It evaluates each sampled response under modality-specific clue contexts, disentangling the contributions of individual clues and their cross-modal interactions for token-level credit assignment. With our training data and learning method, our model OmniReasoning-30B-A3B achieves 50.0% on OmniVideoBench and 42.5% on OmniReasoningBench, improving the base model Qwen3-Omni-30B-A3B-Thinking by 12.8 and 9.3 percentage points, respectively. Moreover, it delivers substantial gains on general and long-video benchmarks, including Video-MME-v2. We hope our work offers a solid step for facilitating future research in omni-modal joint reasoning.

cs.CV↗
arXiv2026

Omni-InteractScope: Benchmarking Proactive and Agentic Interaction in Continuous Omnimodal Streams

Omnimodal large language models (Omni-LLMs) are increasingly expected to assist users while observing continuous audiovisual streams. However, existing evaluations provide limited insight into how reliably these models connect stream evidence and user needs to timely, sustained interventions. We introduce \textbf{Omni-InteractScope}, a benchmark of 1,479 human-annotated, cross-reviewed interactions with 1,790 temporal targets across eight task groups. It links input modalities, event triggers, intervention types, and user needs, covering event-timed tool invocation and query-free assistance alongside notifications, warnings, state tracking, and continuous responses. Our experimental results on the proposed benchmark reveal three critical limitations of current Omni-LLMs: (1) \textbf{Timely tool invocation remains unreliable}: all 12 evaluated models produce an on-time first response for at most 40\% of tool-invocation targets, and joint timing-and-tool accuracy reaches at most 32.0\%. (2) \textbf{Query-free reminders remain poorly covered}: all evaluated models achieve lower timely reference-target recall on query-free reminders than on explicit questions, measured on their respective target sets. Reminder recall never exceeds 23.0\%. (3) \textbf{End-to-end guidance coverage remains limited}: nine models cover none of the 26 guidance sequences in full, and the remaining three cover only one to three sequences each. Targeted controls further show that tool-invocation scores alone do not establish event grounding and that reminder coverage depends on response history and frequency. These findings motivate developing and evaluating streaming assistants for timely action, initiative under general user needs, and coverage of entire response sequences.

cs.AI↗
arXiv2026

Carving Radiology Reports at Their Evidential Joints: Anatomical Region Sets as Generation Units

Organizing radiology reports into independently generated units offers a way to generate fine-grained clinical content and bind it to visual evidence. However, existing methods typically adopt generation-unit structures not directly specified in natural reports or their annotations, introducing additional modeling compromises. We propose RadCarve, which instead defines generation units directly from the anatomical region sets in existing statement-region correspondences, unifying unit formation with visual evidence selection. RadCarve's planner uses composition-specialized query banks to jointly predict unit existence and region composition from the image, iteratively refining both with regional evidence aggregated according to current membership predictions. The generator independently produces each unit's text conditioned on the joint visual evidence of its predicted region set. Compared with strong baselines, RadCarve improves CXRFEScore by 9.4% in unit generation given region sets and achieves 2.4 times the end-to-end binding F1. It also achieves leading report-generation results on Chest ImaGenome and PadChest-GR, using fewer than one-sixth as many parameters as large-scale baselines.

cs.CL↗
arXiv2026

Symmetry-Adapted Eigensolver from Real-Space Sampling of Tensor-Network-Bridged Quantum Circuits

We propose a symmetry-adapted eigensolver for strongly correlated systems that separates where the wave function is \emph{sampled} from where it is \emph{defined}. The lattice is divided into equal blocks: parallel quantum circuits absorb the entanglement within each block, while a matrix product state (MPS) backbone bridges the blocks and restores global correlations. The resulting real-space amplitude $ϕ(a_{\rm real})$ acts only as a sample generator---configurations are drawn from $|ϕ(a_{\rm real})|^2$ directly, block by block, without any Markov chain. The variational state itself is defined in a symmetry-adapted basis by two separate rules: the modulus of $ψ(a_{\rm symm})$ is the square root of the total weight accumulated over the entire equivalence class, $\widetilde{p}(a_{\rm symm})=\sum_{a_{\rm real}\in\{g a_{\rm repr}\}}|ϕ(a_{\rm real})|^2$, whereas its phase is inherited from the single representative amplitude $ϕ(a_{\rm repr})$. The modulus rule is what makes the scheme both exact and cheap: the distribution$\widetilde{p}(a_{\rm symm})$ sampled in the symmetric basis is exactly the one induced by the mapping from the real-space distribution $p(a_{\rm real})$, so every generated configuration is an unbiased sample of $|ψ(a_{\rm symm})|^2$; the phase rule then fixes the only freedom that remains. Because amplitudes within a class are combined in probability rather than summed with symmetry characters, $ψ$ is \emph{not} a group projection of the circuit-MPS state; the class-wide aggregation instead supplies long-range, symmetry-enforced entanglement that the MPS never has to carry itself.

quant-ph↗
arXiv2026

Global Hölder estimates for linearized Monge-Ampère equations in divergence form with applications to dual semigeostrophic equations and periodic homogenization

We establish interior and global Hölder estimates for linearized Monge-Ampère equations in divergence form in all dimensions $n\geq 3$, when the Hessian determinant of the convex Monge-Ampère potential is bounded above and below by positive constants and the vector field on the right-hand side is bounded. The estimates use the $L^p$ norm of the solution for any $p>1$. A key ingredient is an $L^1$ estimate, with appropriate decay rates, for the gradient of the Green's function of the linearized Monge-Ampère operator in small sections, uniform with respect to the pole. As applications, we obtain uniform Hölder estimates for the time derivatives of the primal and dual potentials in the three-dimensional periodic dual semigeostrophic system when the initial density is bounded away from zero and infinity, and establish a linear convergence rate for periodic homogenization of the Monge-Ampère equation.

math.AP↗
arXiv2026

From Pixel Generation to Topological Inference: Structural Dual Super-Resolution for Trustworthy Cross-Physical-Domain Trabecular Morphology Learning

Clinical CT and UHRCT cannot resolve individual trabeculae, whereas synchrotron radiation microCT (SRuCT) provides high-resolution references but is not applicable for in vivo imaging. The two domains differ by a 32x resolution gap, are only coarsely paired, and exhibit severe physical differences including partial volume effects, noise, and artifacts. Existing super-resolution networks and pretrained-prior methods (GLEAN/StyleGAN2, Stable SR/LDM) underperform because they target pixel generation---diverse details and SSIM/PSNR---and do not model these physical differences. Pixel generation for a 32x resolution gap is intrinsically ill-posed. We propose a paradigm shift from pixel generation to topological inference: deterministically predicting invariant microstructures from macro-scale low-resolution inputs, evaluated by morphological parameters. The core of our 2D morphology learning lies in training on 2D slices while evaluating on 3D morphological parameters, ensuring that the learned representations capture true three-dimensional trabecular topology rather than 2D pixel statistics. We realize this paradigm via structural dual super-resolution, coupling forward physical degradation (micro-to-macro) with inverse structural inference (macro-to-micro) through structural duality constraints. The method is an end-to-end, few-shot, compact structural dual network (SDN), comprising a bidirectional modeling network, a multi-scale structural consistency discriminator, and four structural duality constraints. On the test set, SDN achieves morphological parameters largely consistent with SRuCT across six metrics, with SSIM reaching 0.8. Trained on 3.2 um SSRF data, the model generalizes well to 3.25 um isotropic BSRF data from an independent source, validating cross-source generalization and confirming trustworthy structural inference rather than pixel generation.

cs.CV↗
arXiv2026

Nevanlinna theory of the Hahn difference operators and its applications

This paper establishes the version of Nevanlinna theory based on Hahn difference operator $\mathcal{D}_{q,c}(g)=\frac{g(qz+c)-g(z)}{(q-1)z+c}$ for meromorphic function of zero order in the complex plane $\mathbb{C}$. We first establish the logarithmic derivative lemma and the second fundamental theorem for the Hahn difference operator. Furthermore, the deficiency relation, Picard's theorem and the five-value theorem are extended to the setting of Hahn difference operators by applying the second fundamental theorem. Finally, we also consider the solutions of complex linear Hahn difference equations and Fermat type Hahn difference equations.

math.CV↗
arXiv2026

Stabilizing the Dynamic Low-Rank Training

Training neural networks directly in a low-rank parameterization is an appealing route to reducing memory, compute, and storage simultaneously during both training and inference. Dynamic low-rank training (DLRT), which confines weights to a rank-$r$ manifold via the Galerkin projection of the gradient flow, is particularly attractive because it identifies efficient subnetworks on the fly without specialized initialization or post-factorization. However, DLRT fails to find trainable networks under high compression. In this paper, we derive the gradient flow of the best rank-$r$ approximation and point out that the offset of DLRT comes from a curvature-coupling term which is large and thus non-negligible under aggressive compression. Guided by this analysis, we propose a stable dynamic low-rank training method, named SDLRT, which maintains a lightweight compensation buffer that reinjects the top neglected singular directions. Additionally, we introduce a negative feedback on the truncation tolerance to stabilize each layer's rank. Experimentally, SDLRT reliably finds trainable subnetworks where DLRT collapses and as a PEFT adapter on DeBERTa-v3, it achieves the best average score on SuperGLUE at only $2.8\%$ parameter overhead over LoRA.

cs.LG↗
arXiv2026

Omni2Web: Benchmarking Audiovisual Website Development

Screen-recorded web editing requests contain weak deictic expressions such as ``this'' and ``there,'' whose referents depend on speech, cursor trajectories, page state, and edit history. Such requests require intent recovery beyond the explicit specifications assumed by many existing web-editing benchmarks. We introduce Omni2Web, a bilingual benchmark of 918 instances spanning 13,907 edit steps. It defines three complementary tracks: Direct Editing evaluates webpage editing from recordings, Instruction Recovery measures explicit intent recovery, and Instruction Utility tests whether recovered instructions can drive a fixed code executor. We evaluate 17 open- and closed-source models. The best models attain 51.17 on the Edit Fidelity Score (EFS) for Direct Editing and 49.14 on the Instruction Recovery Score (IRS); under the fixed executor, the strongest recovered instructions reach 51.08 EFS, still far below the 89.69 EFS obtained with oracle instructions. Step-level analyses show that correct grounding does not guarantee successful edits, while some Omni models recover instructions that the fixed coding model executes substantially better than their direct edits. Controlled ablations further demonstrate the value of temporally aligned audiovisual evidence, while alternative judges preserve the leader and broad ordering. Together, these findings reveal substantial headroom in multimodal intent recovery and code execution and highlight the promise of pairing Omni rewriters with coding models.

cs.SE↗
arXiv2026

Omni Demand Understanding: A Benchmark for Contextual User-Intent Inference in Multimodal Interaction

Natural audio-visual interaction is emerging as an important interface for AI assistants, allowing users to communicate through speech and vision rather than carefully composed text prompts. However, existing benchmarks of interactive capabilities still focus primarily on response quality, leaving a more fundamental question underexplored: can a model correctly infer the user's underlying demand from complex multimodal interaction? Real-world user demands are often underspecified in speech and must be inferred from multimodal cues and dialogue history. This inference is further complicated by ambiguous or disfluent expression and noisy acoustic environments. Conversely, request-like speech may not constitute a demand to the assistant, leading to false triggers. We establish Omni Demand Understanding (ODU) as a distinct multimodal contextual inference problem: given an interaction stream, a model must detect whether a user demand is present and infer intent from multimodal and conversational context. ODU evaluates this capability along five dimensions, covering both single-turn and multi-turn interactions. We construct ODU-Bench using a challenge-driven taxonomy, taxonomy-guided agentic video generation, and human-recorded interactions, followed by media-grounded annotation and human verification. We evaluate 14 native MLLMs. Even the strongest, Gemini 3.1 Pro, recovers only 44.7% of key information that must be inferred from visual, acoustic, or conversational context. Moreover, 11 of the 14 models exhibit false-trigger rates above 50% on non-demand scenarios. These results reveal a systematic capability gap in current MLLMs' ability to infer contextual user demands. We hope ODU can establish the evaluation of a previously underexplored yet essential capability in multimodal interaction: correctly understanding user demands before generating an appropriate response.

cs.CL↗
arXiv2026

NavPatch: Evidence-Guided Object-Level Costmap Correction with Vision-Language Models

Mobile robots typically rely on geometric maps for obstacle avoidance and path planning, but the resulting obstacle representation does not always match how an object should affect navigation. A low lying cable may be missed, a flexible curtain may create spurious blockage, and a traffic cone may require an exclusion region larger than its observed footprint. We present NavPatch, an object level correction layer that assigns ADD, REMOVE, or EXTEND to navigation relevant object categories through periodic scene understanding with a vision-language model. Open vocabulary grounding localizes object instances, and LiDAR and RGB-D observations provide 3D support. Observation quality filtering and cross frame maintenance determine when each correction patch is committed, replaced, or revoked. In 50 real robot trials across five layouts, NavPatch achieves an overall success rate of 86.0%. An ablation study of four configurations with 200 runs in total shows that NavPatch improves the success rate from 70.0% to 86.0% and reduces the false commit rate from 68.4% to 40.7% compared with updates based only on the current observation.

cs.RO↗
arXiv2026

Hypergraph Enterprise Agentic Reasoner over Heterogeneous Business Systems

Applying Large Language Models (LLMs) to heterogeneous enterprise systems is hindered by hallucinations and failures in multi-hop, n-ary reasoning. Existing paradigms (e.g., GraphRAG, NL2SQL) lack the semantic grounding and auditable execution required for these complex environments. We introduce HEAR, an enterprise agentic reasoner built on a Stratified Hypergraph Ontology. Its base Graph Layer virtualizes provenance-aware data interfaces, while the Hyperedge Layer encodes n-ary business rules and procedural protocols. Operating an evidence-driven reasoning loop, HEAR dynamically orchestrates ontology tools for structured multi-hop analysis without requiring LLM retraining. Evaluations on supply-chain tasks, including order fulfillment blockage root cause analysis (RCA), show HEAR achieves up to 94.7% accuracy. Crucially, HEAR demonstrates adaptive efficiency: utilizing procedural hyperedges to minimize token costs, while leveraging topological exploration for rigorous correctness on complex queries. By matching proprietary model performance with open-weight backbones and automating manual diagnostics, HEAR establishes a scalable, auditable foundation for enterprise intelligence.

cs.AI↗
arXiv2026

Infinite-horizon time-inhomogeneous Kolmogorov equation and non-autonomous multi-scale stochastic systems

We propose a unified approach, based on the infinite-horizon time-inhomogeneous Kolmogorov equation with a parameter, to study the asymptotic behavior of non-autonomous multi-scale stochastic systems with irregular coefficients, where both the fast and the slow equations depend on the highly oscillating time component. We establish both the weak and strong convergence of double averaging principle as well as the functional central limit theorem with new homogenized coefficients given in terms of the evolution system of invariant measures of the time-inhomogeneous fast frozen equation. Moreover, rates of convergence are obtained as byproducts of our argument, which do not depend on the regularities of the coefficients with respect to the time component and the fast variable.

math.PR↗
arXiv2026

Asymptotic Expansions of the Fundamental Solution for a Mixed Local-Nonlocal Operator

In this paper, we study the fundamental solution of the mixed local-nonlocal operator $ -Δ+(-Δ)^s,$ $ 0<s<1. $ We derive precise asymptotic expansions of the fundamental solution and its gradient, up to the first nontrivial correction term, both near the origin and at infinity. The expansions describe the transition between the local and nonlocal diffusion scales and include the critical regimes in which logarithmic terms occur. As applications of these asymptotics, we establish a distributional Bôcher-type theorem for nonnegative supersolutions with an isolated singularity. We further obtain quantitative positive and antisymmetric maximum principles on punctured balls.

math.AP↗
arXiv2026

Interior $C^{1,α}$ estimates for the linearized Monge--Ampère equation in two dimensions

We prove an interior $C^{1,α}$ estimate for solutions of the homogeneous linearized Monge--Ampère equation in dimension two under the assumption \[ 0<λ\leq \det D^2φ\leqΛ<+\infty. \] No continuity assumption on the Monge--Ampère density is required. Our result is an affine-invariant analogue of the classical Morrey--Nirenberg $C^{1,α}$ estimate in two dimensions. The core of the proof is the partial Legendre transform. After the transform, the first derivatives of the solution are quotients of adjoint solutions for a uniformly elliptic non-divergence form equation. Bauman's Harnack inequality gives the Hölder control of the quotient, while the Jacobian identity of the partial Legendre transform and a Caccioppoli estimate give its local boundedness. As an application, we prove a Liouville theorem for entire solutions with at most linear growth.

math.AP↗
arXiv2026

Liouville-type theorems for coupled-drift Monge-Ampère equations

In this paper, we study entire solutions and periodic correctors for the coupled-drift Monge-Ampère equation \[ \det D^2u = \exp\{-a\cdot Du+b\cdot x+V(x)-c_0\}, \quad D^2u>0. \] For $V\equiv0$, we obtain a sharp classification of the whole-space solvability regimes: all entire smooth strictly convex solutions are quadratic when $a=b=0$; no such solution exists when $a\neq0$ and $a\cdot b\le0$; and non-quadratic entire solutions exist when $a=0$ and $b\neq0$, or when $a\cdot b>0$. For the null case $a\neq0$, $a\cdot b=0$, we give a scalar maximum-principle argument in every dimension $n\ge2$. For periodic $V$, we prove existence and uniqueness of the normalized pair $(ψ_A,c_A)$ solving the drifted cell problem \[ \det(A+D^2ψ) = \exp\{-a\cdot Dψ+V-c_A\}, \quad A+D^2ψ>0 \quad\text{on }\mathbb T^n. \] We also prove that any asymptotically quadratic entire solution must satisfy $b=Aa$. If its remainder is bounded, then the solution is the corresponding quadratic-periodic corrector up to an additive constant.

math.AP↗
arXiv2026

A Novel Approach to Temporal QoS Estimation via Extended Kalman Filter-Incorporated Latent Feature Analysis

Predicting temporal Quality of Service (QoS) data is critical for optimizing network services and rationalizing resource allocation in cloud computing and service-oriented systems. Existing mainstream methods have achieved promising predictive performance. However, their purely data-driven manner limits their ability to capture non-stationary temporal patterns, thereby leading to accuracy degradation when temporal QoS data exhibits fluctuations. To tackle this limitation, we propose a novel Extended Kalman Filter-Enhanced Latent Feature Analysis (EKL) model to perform efficient and accurate temporal QoS prediction from the perspective of bidirectional model-data-driven learning. Its main idea is three-fold: a) designing a model-driven feature producer to obtain the temporal latent features to capture the intricate temporal pattern following the principle of an Extended Kalman Filter; b) building a data-driven feature producer based on the alternating least squares algorithm to identify time-invariant latent features describing intrinsic user-service characteristics; c) exploiting a density-oriented parallel strategy that achieves workload balancing by sorting users in accordance with their service invocation density, which effectively elevates computational efficiency. In addition, we provide a rigorous theoretical analysis to formally prove the convergence of the proposed EKL. Experimental evaluations conducted on real-world temporal QoS datasets reveal that our proposed EKL surpasses existing state-of-the-art models with respect to both computational efficiency and prediction accuracy for missing temporal QoS data.

cs.LG↗
arXiv2026

Min-Max Construction of Anisotropic Minimal Surfaces with Genus Bound

We establish an anisotropic analogue of the celebrated theorem of Meeks-Simon-Yau: every minimizing sequence of surfaces within a fixed isotopy class converges to a smooth stable anisotropic minimal surface, with genus lower semicontinuity. This result also strengthens White's foundational existence theory for anisotropic minimal disks. As an application, we develop an anisotropic Simon-Smith min-max theory. In every closed $3$-manifold, we construct anisotropic min-max sequences within fixed isotopy classes whose limits are stable anisotropic minimal surfaces that are smooth except possibly at a single point. If the integrand satisfies either an ellipticity bound or a $C^3$-pinching condition, we remove the singular point by proving two independent removable singularity theorems for anisotropic minimal surfaces that are smooth and stable away from finitely many points. These removable singularity results also allow to remove the singularities arising in the anisotropic Almgren-Pitts min-max construction in $3$-manifolds of De Philippis-De Rosa and in its multiparameter variants.

math.DG↗