arXiv subjects

Hui Wang

Publications and source records attributed to Hui Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Metamodel-Guided Model Generation with Layered Constraints

Large language models (LLMs) enable natural-language interaction in engineering modeling, but generated models may violate structural constraints, domain rules, or task requirements. We propose a metamodel-guided model generation method that coordinates generation-time constraints and post-generation validation. The method transforms metamodel information, uses its terminology to guide structured constraint extraction from specifications, and links constraints to metamodel elements while recording their sources in an Integrated Constraint Model (ICM). For each task, relevant constraints are bound to concrete objects, values, and references. The generation-time constraint layer (L1) restricts candidate content. The post-generation validation layer (L2) checks constructed models and serialized artifacts, and task acceptance checks retain the original requirements throughout repair. Deterministic procedures construct and serialize models, while LLMs propose candidate content and repairs. Validation uses existing domain tools and checkers written by humans with LLM assistance. Experiments cover AUTOSAR, railway models, and structured decisions in private international law. All 60 AUTOSAR generation runs passed acceptance within the declared task scope, and all 255 resulting ARXML files passed XSD validation. In a separate controlled AUTOSAR repair experiment, all 85 core fault units and 15 prespecified substitute units were restored within one repair round. A local AUTOSAR experiment recorded interventions during stepwise generation. The results support coordinating generation constraints, domain checks, and task acceptance to construct models and guide bounded repair.

cs.SE↗

arXiv2026

Action-level characterization of gravitational-wave propagation in dynamical Barbero--Immirzi gravity

We investigate which operators in first-order dynamical Barbero--Immirzi (BI) gravity control cosmological tensor propagation at the action level. Within a bosonic, two-derivative, curvature-linear Einstein--Cartan class, eliminating the algebraic Lorentz connection reveals a separation between the scalar sector and the tensor kinetic normalization. In particular, the Holst-to-Palatini ratio determines the scalar kinetic structure, whereas the transverse-traceless tensor mode is normalized by the parity-even Hilbert--Palatini coefficient. As a consequence, a minimal dynamical Holst sector with fixed parity-even normalization remains exactly on the general-relativistic tensor-propagation surface and does not generate anomalous gravitational-wave friction. We then construct a regular nonminimal parity-even realization in which the tensor normalization evolves cosmologically, derive the corresponding standard-siren observable, and apply current GWTC-3, GWTC-4.0, and GWTC-5.0 information as an observational test of this action-level tensor sector. The resulting constraints should therefore be interpreted as limits on the evolution of the tensor kinetic normalization rather than direct constraints on minimal BI torsion dynamics.

gr-qc↗

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

Cyclic Riemannian nilmanifolds are naturally reductive

We study the structure of cyclic Riemannian nilmanifolds, that is, connected nilpotent Lie groups $N$ endowed with a left-invariant cyclic metric $\langle\cdot,\cdot\rangle$. We prove that a Riemannian nilmanifold $(N,\langle\cdot,\cdot\rangle)$ is cyclic if and only if its Lie algebra $\mathfrak{n}$ is at most two-step nilpotent and the associated family of skew-symmetric endomorphisms $J_{\mathfrak{z}}=\{J_Z : Z\in \mathfrak{z}\}\subset so(\mathfrak{a})$ is Abelian. As a direct consequence, every cyclic Riemannian nilmanifold is naturally reductive. We also determine the full isometry group of a connected and simply connected cyclic Riemannian nilmanifold without Euclidean factor.

math.DG↗

arXiv2026

ThinkFlow: Self-Evolving Probabilistic Latent Memory for Lifelong Conversational Agents

Lifelong conversational agents rely on memory systems to maintain deep, context-aware interactions with users. However, existing explicit textual memory pipelines suffer from a severe information bottleneck, often losing subtle behavioral patterns and emotional shifts. Furthermore, being typically static post-deployment, they cannot autonomously adapt to personal habits and preferences without manual feedback. Cognitive science, however, suggests that humans maintain mental models purely in a latent space and continuously refine them through predictive coding. Inspired by this, we propose \textbf{ThinkFlow}, a novel end-to-end latent memory framework for lifelong conversational agents. ThinkFlow bypasses the text bottleneck by dynamically compressing conversational flows into probabilistic latent memory skills, autonomously consolidating complex user states into disentangled, continuous vectors without semantic interference. To break this barrier, we introduce a test-time evolution paradigm. By coupling teacher-guided latent alignment to bootstrap the initial state with a self-supervised next-user-utterance prediction task for continuous refinement, the framework successfully overcomes cold-start challenges and achieves label-free lifelong personalization. Extensive experiments on long-term conversation benchmarks demonstrate that ThinkFlow significantly outperforms prevailing memory systems, providing highly personalized and contextually accurate responses over extended multi-session interactions.

cs.AI↗

arXiv2026

Interactive Memory Learning for Long-Term Conversations

Recent advancements in large language models have significantly enhanced the capabilities of agents in modeling long-term conversations. Despite these successes, existing approaches typically adopt a static heuristic paradigm, where information is passively archived without adaptive memory valuation. Consequently, these methods fail to self-evolve or align their memory management with evolving user needs. To address this, we propose ICML (InteraCtive Memory Learning), a multi-agent framework that transforms the memory mechanism from a passive archive into a learnable, interactive memory policy. Specifically, we first employ a session synthesis pipeline to generate expert data, facilitating rapid test-time adaptation in unseen scenarios. Building on this, ICML utilizes an online reinforcement learning mechanism where a Planner agent selectively encodes high-value information and a Trigger agent dynamically retrieves it to optimize response quality, whereby the two agents co-evolve through continuous interaction feedback. Crucially, both agents are synchronized through a delayed reward mechanism that propagates future feedback back to earlier storage decisions, ensuring memory policies are precisely aligned with user expectations. Experimental results demonstrate that ICML significantly outperforms strong baselines, exhibiting the unique capability to continuously improve response quality as interactions accumulate.

cs.AI↗

arXiv2026

LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence

We introduce LimiX-2, a new model in the LimiX family, developed through model and data scaling guided by our previously established scaling laws. LimiX-2 adopts the Contextual Mechanism Networks (CMNs) paradigm and is pretrained with Context-Conditional Masked Modeling (CCMM). CMNs shifts the organizing principle of in-context learning from target-centric prediction to mechanism-oriented joint modeling. Rather than centering the network on the $p(y \mid x, D_{\mathrm{context}})$ objective of conventional tabular PFNs, it is designed around learning $p(x, y \mid D_{\mathrm{context}})$, a context-dependent representation of the joint structure underlying data generation. Pretraining uses synthetic datasets generated by structural causal models (SCMs) spanning diverse graph structures, functional mechanisms, and observation processes. Evaluations on TabArena, TALENT, and BCCO show that LimiX-2 outperforms current dataset-specific models and tabular foundation models. Beyond predictive performance, the CMN paradigm also promotes causal awareness in LimiX-2: its feature attention encodes direct causal relationships, enabling accurate causal skeleton recovery.

cs.AI↗

arXiv2026

Boundary layer analysis for the 2D Chemotaxis-Navier-Stokes system with logarithmic sensitivity, Part II. viscous vanishing limit

This is the second part of a two-part work concerning boundary layer solutions to the coupled Chemotaxis-Navier-Stokes system in the two-dimensional half-space. In the present work, we address the convergence of boundary layer solutions to singular chemotaxis-fluid equations under slip boundary conditions with respect to the chemical diffusion-viscosity parameter $\varepsilon$ in the two-dimensional half-plane. More precisely, we show that the boundary layer for $\varepsilon>0$ (viscous convection coefficient) converges to the superposition of the outer layer (solution with $\varepsilon=0$) and the inner layer as $\varepsilon\rightarrow0$. The outer and inner profiles are explicitly derived as in the first part\cite{WWZ}. Furthermore, the well-posedness results of the coupled Chemotaxis-Navier-Stokes system in conormal Sobolev spaces will be presented in Appendix. They answer the question mentioned in the first part of the two-part work. This study could help the understanding of the chemotactic movement of aerobic bacteria to the water-air surface observed experimentally in fluids, and enrich the theoretical results of boundary layer in chemotactic fluid models.

math.AP↗

arXiv2026

AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

We introduce AuK, an open-source foundational model that unifies speech generation and editing through a common interface of natural-language instructions and audio context. To support this broad capability set, we construct approximately 3.03 billion instruction--audio instances and 1.95 million hours of effective supervision across five task families: speech generation, content editing, enhancement and separation, paralinguistic editing, and acoustic editing. AuK combines a multimodal large language model for semantic conditioning, an VAE jointly trained on speech, general audio, and music for acoustic conditioning, and a hybrid rectified-flow Transformer that performs dual-stream MMDiT blocks followed by unified single-stream DiT blocks for generation. Training begins with generation-only warm-up and proceeds to joint generation--editing pre-training. We then apply complementary post-training strategies: human-feedback preference optimization for open-ended editing and reward-based reinforcement learning for speech generation. To reduce inference cost, we further distill the model with consistency initialization and task-routed Decoupled DMD. The resulting AuK-Flash performs 4-step inference without classifier-free guidance and achieves a 4.5 wall-clock speedup over the full model under matched conditions. Experiments demonstrate leading performance on zero-shot and instruction-controlled speech generation and general instruction-guided editing, while remaining competitive on signal-level restoration tasks. We release both the source code and model weights to support reproducibility and further research.

cs.SD↗

arXiv2026

DL$^3$M: A Vision-to-Language Framework for Expert-Level Medical Reasoning through Deep Learning and Large Language Models

Medical image classifiers detect gastrointestinal diseases well, but they do not explain their decisions. Large language models can generate clinical text, yet they struggle with visual reasoning and often produce unstable or incorrect explanations. This leaves a gap between what a model sees and the type of reasoning a clinician expects. We introduce a framework that links image classification with structured clinical reasoning. A new hybrid model, MobileCoAtNet, is designed for endoscopic images and achieves high accuracy across eight stomach-related classes. Its outputs are then used to drive reasoning by several LLMs. To judge this reasoning, we build two expert-verified benchmarks covering causes, symptoms, treatment, lifestyle, and follow-up care. Thirty-two LLMs are evaluated against these gold standards. Strong classification improves the quality of their explanations, but none of the models reach human-level stability. Even the best LLMs change their reasoning when prompts vary. Our study shows that combining DL with LLMs can produce useful clinical narratives, but current LLMs remain unreliable for high-stakes medical decisions. The framework provides a clearer view of their limits and a path for building safer reasoning systems. The complete source code and datasets used in this study are available at https://github.com/souravbasakshuvo/DL3M.

cs.CV↗

arXiv2026

Fact-Checking Comments Precede Community Notes to Misleading Posts on X

Community-based fact-checking systems, such as X's Community Notes program, offer a potentially scalable approach against online misinformation diffusion. However, their efficacy is substantially undermined by the delay of fact-check delivery. To address this challenge, we investigated the promise of organic user comments as rapid corrective signals by analyzing a large-scale dataset of over 2.2 million comments directed to 1,841 community fact-checked misleading posts. We developed a high-performance language model pipeline to identify ``fact-checking (FC) comments'', i.e., comments that correct source posts with reasoning or evidence (91% accuracy). Using this pipeline, we find that 99.4% of misleading posts receive their first FC comments before the official community note is created. Notably, the median time to the initial FC comment is only 0.1 hours, while the creation of community notes has a median latency of 9.7 hours since the publication of misleading posts. Additionally, compared to random posts, those misleading posts with displayed notes have significantly more FC comments. This suggests that FC comments are a unique pattern for misleading posts. We further identified characteristics of FC comments' intensity and speed: (i) FC comments' volume is significantly associated with content richness and misinformation type, favoring multimedia content and missing-context scenarios; (ii) the speed of FC comments remains robust across diverse topics and emotional intensities. These results suggest that platforms could use FC comments as early-warning signals, and potential resource for complementing community notes or automated synthesis.

cs.HC↗

arXiv2026

AREX: Towards a Recursively Self-Improving Agent for Deep Research

Deep research requires agents to find answers that jointly satisfy multiple constraints. Discovering such answers is costly, whereas verifying a candidate can often be decomposed into tractable constraint-wise checks. This discovery--verification asymmetry suggests that a research agent should do more than simply search longer: it should recursively improve its current answer by verifying intermediate results and using the partially verified state to guide subsequent refinement. We introduce AREX, a family of Recursively Self-Improving (RSI) deep research agents. AREX alternates between an inner research loop that gathers evidence and constructs a provisional answer, and an outer self-improvement loop that audits the answer constraint-wise, identifies unresolved claims, and launches targeted follow-up research. To sustain RSI over long horizons, AREX learns an autonomous context-update tool that compresses growing interaction history into a compact improvement state preserving verified evidence and unresolved constraints, without relying on an external model. We train AREX on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning. To mitigate sparse final rewards during long horizon learning, we emphasize key steps where decisive evidence is acquired or erroneous research directions are corrected. We instantiate a dense 4B model and a 122B-A10B Mixture-of-Experts model. Across BrowseComp, WideSearch, DeepSearchQA, Humanity's Last Exam (HLE), and other reasoning and tool-use benchmarks, AREX substantially outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.

cs.AI↗

arXiv2026

CoVer: Conflict-Aware Claim Verification

Social media fact-checking has long been challenged by evidence-level and aggregation-level conflicts, where erroneous evidence mimics authoritative news sources. To capture this challenge and support conflict verification tasks, we present ContraNote, a large-scale real-world dataset curated from X's Community Notes system. It includes 33,686 posts for evaluating evidence-level conflict resolution, and 54,474 instances for evaluating aggregation-level prioritization. Additionally, we propose CoVer, a factual adjudication framework with three-stage pipelines: evidence schema normalization, factual consensus and support verification. This prioritizes evidence over noise to prevent it from compromising the final verdict. Technical evaluations show that CoVer achieves strong performance compared with state-of-the-art baselines across ContraNote (86.0% Acc., 68.0% mac. F1, 64.5 bal. Acc. on Conflict; and 88.5% Acc., 88.5 mac. F1 and 89.2 bal. Acc. on Prioritization), CONFACT-HumC (88.4% Acc.) and CONFACT-ModC (89.4% Acc.).

cs.AI↗

arXiv2026

GRAIN: Bridging Name and Narrative Shifts in Real-World Graph Reasoning through Invariance-Rewarded Agentic RL

Despite their potential in standardized graph tasks, Large Language Models (LLMs) remain brittle to real-world shifts in node identifiers and task formulation. While deterministic graph tools are invariant to such shifts, extracting topological structures from noisy text is highly fragile for LLMs, which often overfit to surface patterns. Moreover, mitigating these parsing failures via multi-agent systems incurs prohibitive latency. To address this, we propose GRAIN, a single-agent framework optimized via reinforcement learning. GRAIN models reasoning as a semantic parsing and tool-execution pipeline, guided by a Structure Invariance Reward. By validating extracted intermediate graphs against ground-truth topologies, this reward forces the LLM to learn robust text-to-structure mappings rather than memorizing linguistic artifacts. We also introduce GRIT, a benchmark evaluating sensitivity to such linguistic shifts. GRAIN outperforms multi-agent baselines by 16.45\% in accuracy with approximately 24\% lower latency. Furthermore, it demonstrates superior structural generalization, halving the out-of-distribution (OOD) gap of SFT models (from 15.77\% to 7.80\%) and maintaining robustness on large-scale graphs beyond the training distribution.

cs.AI↗

arXiv2026

Deterministic preparation of entangled Dicke states

Dicke states $|J=N/2,m\rangle$ of a collection of $N$ atoms were central to Dicke's theory of superradiance. Except for the fully polarized end states, they are entangled many-body states; in particular, the single-excitation Dicke state is the W state well known in quantum information science. However, deterministic preparation of Dicke states with a prescribed spin projection remains challenging, especially without relying on postselection or heralding. Here we present a detuning-programmed Hamiltonian protocol using atoms or qubits in a dispersive cavity subject to a coherent transverse drive. In the absence of the drive, the off-resonant cavity produces an effective collective-spin interaction. By combining this cavity-mediated interaction with the coherent drive, tuning the atom--drive detuning enables the protocol, in principle, to target any allowed Dicke state along the symmetric Dicke ladder. Starting from the transverse-drive ground state, adiabatic ground-state interpolation prepares the selected Dicke state by ramping down the drive strength while ramping up the cavity-mediated interaction. After preparation, tuning the cavity into resonance with the atoms or qubits provides a direct way to probe the collective-emission response of the prepared Dicke state. We discuss implementation with superconducting circuit QED and show that the prepared states, especially the central Dicke state with $m=0$, provide resources for quantum sensing with Heisenberg-limited scaling.

quant-ph↗

arXiv2026

Self-Calibrating Dense Displacement Fields for Reliable Co-Registration of Large Optical Satellite Imagery

Co-registration underlies nearly every multi-temporal and multi-sensor use of optical satellite imagery, and operational products still carry documented offsets well above the fraction-of-a-pixel scale at which change detection, time series, and data fusion degrade. Real image pairs differ along several axes at once (sensor response, scene content, viewing geometry, resolution, mosaic seams), and the last of these is not a single global motion. Existing tools embed a motion model and constants tuned to their development data; a pair that fits is registered precisely, while one that does not either fails to match or returns a result wrong by tens of pixels with no failure reported. Learned matchers add a GPU requirement and carry no accuracy guarantee outside their training distribution. We present SCDF (self-calibrating displacement fields), a training-free, GPU-free estimator whose motion model is the dense per-pixel displacement field itself, so no scene motion falls outside the model. A single predict--measure--filter loop runs over a resolution pyramid: the accumulated field predicts where each patch of the moving image falls in the reference, RootSIFT matching and a correlation pass measure the displacement there to sub-pixel precision, and filters whose thresholds are all calibrated on the image pair itself decide what survives. One configuration, with no per-dataset tuning, processes full $8192^2$ scenes on a single CPU core. On 584 constructed-ground-truth pairs built from real Sentinel-2, Landsat-8/9, and NAIP imagery, against seven classical baselines and two zero-shot pretrained matchers, SCDF registers every pair with zero failures, reduces the best baseline's real-pair median end-point error from 6.83 to 4.17m, and cuts its 90th percentile from 17.8 to 7.77m.

cs.CV↗

arXiv2026

Dynamical Barbero--Immirzi field coupled to quintessence: gravitational-wave propagation constraints and next-generation forecasts

We investigate the imprints of a dynamical Barbero--Immirzi (BI) field $γ(x)$ coupled to a quintessence scalar field $ϕ$ on gravitational-wave (GW) propagation. In the framework of Einstein--Cartan--Holst gravity, promoting $γ$ to a dynamical scalar introduces a stress--energy that back-reacts on the metric, modifying the GW friction term. A minimal coupling $\proptoβ\,ϕ^2γ^2$ between the BI field and quintessence leads to a two-parameter extension of the Belgacem--Maggiore parametrization, characterized by $\xBI$ (from the isolated BI field) and $\xcp$ (from the coupling). Using the LIGO--Virgo--KAGRA GWTC-3 dark-siren constraint $Ξ_0=1.2^{+0.7}_{-0.7}$, we obtain the first simultaneous constraints: $|\xBI|\lesssim0.7$ and $|\xcp|\lesssim0.13$ at 90\% credibility. We then forecast the sensitivity of next-generation detectors Einstein Telescope (ET) and Cosmic Explorer (CE), showing that a 10-year observation campaign can improve these bounds by roughly one to two orders of magnitude depending on the parameter---a factor of $\sim\!20$ for $\xBI$ and $\sim\!20$ for $\xcp$---reaching $σ(\xBI)\sim3\times10^{-2}$ and $σ(\xcp)\sim1.2\times10^{-2}$. Translated into microscopic parameters, this corresponds to $γ_{\rm dyn}\lesssim10^{-12}$ and $β\lesssim10^{-3}$, providing a powerful new observational window into the interplay between quantum-gravity phenomenology and dark energy.

gr-qc↗

arXiv2026

Robust structure from motion for aerial-ground images via detector-free feature matching and multi-view track refinement

Integrated 3D reconstruction from aerial-ground images is essential for generating high-precision urban 3D models, yet severe variations in viewpoint, scale, and rotation make robust feature matching highly challenging. To address these limitations, this study introduces a rotation-robust detector-free matching network coupled with multi-view track refinement for incremental Structure from Motion (ISfM). The proposed workflow features four key modules. First, rotation-aware feature extraction replaces traditional convolutions with an Omnidirectional State Space Block (OSS Block) that selectively scans across eight symmetrical directions to model long-range spatial dependencies and synthesize rotation-invariant feature maps. Second, multi-scale attention transformation utilizes quadtree attention to build a hierarchical token pyramid that isolates high-association token regions and discards irrelevant areas, capturing long-range context with linear computational complexity. Third, bi-directional feature matching executes a symmetric coarse-to-fine matching scheme where coarse alignment computes dual-direction Softmax confidence matrices under mutual nearest neighbor constraints, and fine alignment uses a multi-layer perceptron to regress sub-pixel coordinate offsets. Finally, multi-view track refinement employs an integrated indexing structure to evaluate localized spatial proximity and link disjoint sub-tracks to the highest-confidence anchor point, ensuring stable feature repeatability across the ISfM pipeline. By using real aerial-ground datasets, experimental results demonstrate that the proposed method improves AUC at 5° pose error by 93.9% compared with LoFTR and achieves the highest precision in ISfM reconstruction, with the improved accuracy ranging from 27.6% to 32.7%. The proposed method provides a reliable solution for integrated 3D reconstruction of aerial-ground images.

cs.CV↗