arXiv subjects

Yin Wang

Publications and source records attributed to Yin Wang.

At least 19 recordsLinked to original sources

arXiv2026

CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction

Autoregressive vision language models unify heterogeneous perception tasks but are highly susceptible to compounding errors. On-policy distillation (OPD) bridges the training-inference mismatch by training students on their own rollouts. However, unreliable student predictions, especially early in training, can derail the trajectory and degrade the quality of teacher supervision. While recent interleaved distillation methods allow the teacher to verify and replace student tokens, they primarily rely on rigid ranking metrics rather than exact teacher confidence, and they overlook how intervention decisions can inform token-level supervision. To address this, we introduce Confidence-Aware On-Policy Distillation (CA-OPD), a framework that couples reliable rollout construction with adaptive supervision. CA-OPD utilizes teacher confidence to selectively correct unreliable student transitions, gradually transferring rollout control to the student via a strict-to-relaxed schedule. Crucially, CA-OPD aligns knowledge transfer with these intervention decisions: corrected positions receive direct cross-entropy supervision from the teacher's prediction, while retained positions benefit from the teacher's full predictive distribution. Evaluated in a multi-teacher setting for GUI grounding and optical character recognition, CA-OPD substantially improves the Qwen3.5-0.8B baseline across all six target benchmarks, including gains of $9.50$ points on ScreenSpot-Pro and $6.72$ points on OCRBench-v2 English. Controlled studies further show that the gains depend on intervention placement, progressive rollout control, and intervention-aligned supervision, rather than intervention frequency alone.

cs.CV↗

arXiv2026

FinRiskAtlas: Decision-Aligned Evaluation of Large Language Models for Financial Risk Review

Deploying large language models for professional financial review requires more than measuring general financial competence: models must perform the specific review operation required by a workflow and determine whether available evidence is sufficient for a defensible decision. Existing financial benchmarks cover knowledge, reasoning, compliance, and professional tasks, but their evaluation units are often organized around datasets or task formulations rather than the decisions that deployed systems support. We introduce FinRiskAtlas, a Chinese-language benchmark that evaluates financial LLMs along two complementary dimensions: operation execution under fixed evidence states and evidence-state control under evolving review conditions. The static benchmark contains 9,742 instances across 53 task families, including 42 Domain Knowledge families and eleven downstream review operations defined by explicit evaluation contracts. FinRisk-Ask extends this framework through offline replay of 680 pre-action states from 104 de-identified professional trajectories, withholding future evidence during inference and using it only to construct expert-verified evidence targets. Across 33 model configurations, operation-level evaluation yields non-redundant rankings (mean pairwise Spearman correlation 0.42 across downstream operations), and knowledge-based shortlisting can incur up to 18.01 points of regret on individual operations. FinRisk-Ask further shows that entering the Ask branch more frequently does not necessarily improve request targeting or end-to-end evidence acquisition. These results show that broad financial capability scores do not fully capture where models are reliable in professional workflows, motivating evaluation units aligned with the decisions and evidence states that deployed systems must support.

cs.AI↗

arXiv2026

EndoWAM: A Grounded World-Action Model for Generalizable Endoscopic Navigation

Autonomous endoscopic navigation can reduce clinicians' operational burden, yet robust control remains challenging due to tissue deformation, transient occlusions, and rapidly changing viewpoints. Existing learning-based policies typically predict actions from current observations without explicitly modeling future dynamics, limiting their robustness and reliability in safety-critical settings. World Action Models (WAMs) offer a promising alternative by coupling predictive visual dynamics with action generation, but extending them to robotic endoscopy remains challenging due to limited training data, restricted viewpoint diversity, deformable anatomy, and high inference latency. We present EndoWAM, which is, to our knowledge, the first WAM for generalizable robotic endoscopic navigation. EndoWAM introduces future grounding, which predicts task-relevant target regions in future observations from intermediate denoising features of a video world model. Specifically, EndoWAM couples a lightweight diffusion transformer for future target-region prediction with a discrete action expert through a shared predictive representation. This design injects target-aware supervision into predictive dynamics modeling, improving robustness to visual degradation and viewpoint changes while enabling real-time control in a single denoising pass. We further introduce EndoMotion, a robotic endoscopic motion dataset spanning three anatomically distinct procedures: ureteroscopy, esophagoscopy, and endoscopic retrograde cholangiopancreatography (ERCP). EndoWAM consistently outperforms all baselines and alternative grounding strategies, while demonstrating strong zero-shot generalization to unseen viewpoints, environments, and targets. These results establish EndoWAM as a predictive, target-grounded framework for accurate, generalizable, and long-horizon navigation in visually constrained endoscopic environments.

cs.RO↗

arXiv2026

StepX-Edge: An On-Device UI Vision-Language Model via Architecture-Training-Deployment Co-Design

Deploying a vision-language model with full UI understanding on end devices has long been trapped between accuracy and efficiency: on one side is the accuracy bar for OCR, screen understanding, visual question answering, and element grounding; on the other is the strict compute, memory, and power budget of mobile chips. Existing work either trades one for the other, or stops at simulation without real-device validation. We present StepX-Edge, a 0.9B-parameter on-device UI vision-language model that resolves this tension through three-layer co-design of architecture, training, and deployment. Architecturally, UI-aware Layered Visual Encoding (ULVE) and a Progressive Dimensionality Projection (PDP) connector target the extreme aspect ratios and fine-grained perception of screens, while standard full attention throughout ensures native compatibility with mainstream mobile NPU operators. For training, the five-stage StepX-Curriculum framework is designed around our observation of mutual-promotion effects among UI subtasks, so that all four capabilities grow synergistically under a tight parameter budget rather than interfering. For deployment, a module-wise differentiated two-stage PTQ-to-QAT quantization scheme keeps the post-quantization accuracy loss within 1%. StepX-Edge achieves the strongest overall UI understanding among <=1B models, surpassing all 2B-2.3B baselines on ScreenQA (88.76 F1) and Chinese OCRBench v2 (57.25), and matching 1.3B-2.3B general VLMs on RefCOCO (92.0%) and OCRBench v1 (831) with far fewer parameters. After W4A16+KV8 quantization, the model runs stably on Snapdragon 8 Gen5 devices with ~0.84 s TTFT, 98 tok/s decode, and 1.4 GB peak memory. We will open-source the training data, the full training recipe, and the quantization deployment pipeline.

cs.CV↗

arXiv2026

Organizational Control Layer: Governance Infrastructure at the Execution Boundary of LLM Agent Systems

LLM-based agents are increasingly deployed in workflows where generated outputs may trigger state-changing actions, such as price offers, refunds, payments, or tool calls. This creates an execution-boundary problem: a platform must decide whether an agent's proposed action is authorized before the action is executed. We introduce the Organizational Control Layer (OCL), a model-agnostic governance layer that separates proposal generation from environment-facing execution. OCL intercepts generated actions, checks them against role, policy, and economic constraints, and either approves, revises, blocks, or escalates them without modifying the underlying LLM generator. We evaluate OCL on adversarial buyer--seller negotiation environments adapted from AgenticPay. Across multiple frontier LLM backends, OCL reduces observed unsafe executions from 88% to 0% while increasing valid success from 12% to 96%. Ablations show that this gain comes from combining pre-execution enforcement with structured recovery, rather than from prompting or blocking alone. These results suggest that deployment-grade LLM agent systems require explicit governance at the boundary between language generation and executable action.

cs.MA↗

arXiv2026

Automated Extraction of Mechanical Constitutive Models from Scientific Literature using Large Language Models: Applications in Cultural Heritage Conservation

The preservation of cultural heritage is increasingly transitioning towards data-driven predictive maintenance and "Digital Twin" construction. However, the mechanical constitutive models required for high-fidelity simulations remain fragmented across decades of unstructured scientific literature, creating a "Data Silo" that hinders conservation engineering. To address this, we present an automated, two-stage agentic framework leveraging Large Language Models (LLMs) to extract mechanical constitutive equations, calibrated parameters, and metadata from PDF documents. The workflow employs a resource-efficient "Gatekeeper" agent for relevance filtering and a high-capability "Analyst" agent for fine-grained extraction, featuring a novel Context-Aware Symbolic Grounding mechanism to resolve mathematical ambiguities. Applied to a corpus of over 2,000 research papers, the system successfully isolated 113 core documents and constructed a structured database containing 185 constitutive model instances and over 450 calibrated parameters. The extraction precision reached 80.4\%, establishing a highly efficient "Human-in-the-loop" workflow that reduces manual data curation time by approximately 90\%. We demonstrate the system's utility through a web-based Knowledge Retrieval Platform, which enables rapid parameter discovery for computational modeling. This work transforms scattered literature into a queryable digital asset, laying the data foundation for the "Digital Material Twin" of built heritage.

cs.DB↗

arXiv2026

Multimodal Priors-Augmented Text-Driven 3D Human-Object Interaction Generation

We address the challenging task of text-driven 3D human-object interaction (HOI) motion generation. Existing methods primarily rely on a direct text-to-HOI mapping, which suffers from three key limitations due to the significant cross-modality gap: (Q1) sub-optimal human motion, (Q2) unnatural object motion, and (Q3) weak interaction between humans and objects. To address these challenges, we propose MP-HOI, a novel framework grounded in four core insights: (1) Multimodal Data Priors: We leverage multimodal data (text, image, pose/object) from large multimodal models as priors to guide HOI generation, which tackles Q1 and Q2 in data modeling. (2) Enhanced Object Representation: We improve existing object representations by incorporating geometric keypoints, contact features, and dynamic properties, enabling expressive object representations, which tackles Q2 in data representation. (3) Multimodal-Aware Mixture-of-Experts (MoE) Model: We propose a modality-aware MoE model for effective multimodal feature fusion paradigm, which tackles Q1 and Q2 in feature fusion. (4) Cascaded Diffusion with Interaction Supervision: We design a cascaded diffusion framework that progressively refines human-object interaction features under dedicated supervision, which tackles Q3 in interaction refinement. Comprehensive experiments demonstrate that MP-HOI outperforms existing approaches in generating high-fidelity and fine-grained HOI motions.

cs.CV↗

arXiv2026

Symphony-Coord: Adaptive Routing for Multi-Agent LLM Systems

Multi-agent large language model systems can tackle complex multi-step tasks by decomposing work and coordinating specialized behaviors. However, current coordination mechanisms typically rely on statically assigned roles and centralized controllers. As agent pools and task distributions evolve, these design choices can lead to inefficient routing, poor adaptability, and fragile fault recovery. We introduce Symphony-Coord, a task-local coordination framework with decentralized execution that transforms agent selection into an online multi-armed bandit problem. Instead of relying on a fixed task-to-role map, Symphony-Coord allows routing specializations to emerge from interaction and feedback. The framework employs a two-stage dynamic beacon protocol:(i) a lightweight candidate screening mechanism to limit communication and computation overhead; and (ii) an adaptive LinUCB selector that routes subtasks using context features derived from task requirements and agent states, updated through delayed post-execution feedback. Under candidate-conditional linear bandit assumptions, we prove sublinear regret bounds for the immediate-feedback selector and explicitly separate the deferred-update effects introduced by post-vote rewards. Validation through simulation experiments and real-world large language model benchmarks shows that Symphony-Coord improves task routing efficiency and recovery behavior under distribution shifts and agent failures.

cs.MA↗

arXiv2026

Dynamic Worlds, Dynamic Humans: Generating Virtual Human-Scene Interaction Motion in Dynamic Scenes

Scenes are continuously undergoing dynamic changes in the real world. However, existing human-scene interaction generation methods typically treat the scene as static, which deviates from reality. Inspired by world models, we introduce Dyn-HSI, the first cognitive architecture for dynamic human-scene interaction, which endows virtual humans with three humanoid components. (1)Vision (human eyes): we equip the virtual human with a Dynamic Scene-Aware Navigation, which continuously perceives changes in the surrounding environment and adaptively predicts the next waypoint. (2)Memory (human brain): we equip the virtual human with a Hierarchical Experience Memory, which stores and updates experiential data accumulated during training. This allows the model to leverage prior knowledge during inference for context-aware motion priming, thereby enhancing both motion quality and generalization. (3) Control (human body): we equip the virtual human with Human-Scene Interaction Diffusion Model, which generates high-fidelity interaction motions conditioned on multimodal inputs. To evaluate performance in dynamic scenes, we extend the existing static human-scene interaction datasets to construct a dynamic benchmark, Dyn-Scenes. We conduct extensive qualitative and quantitative experiments to validate Dyn-HSI, showing that our method consistently outperforms existing approaches and generates high-quality human-scene interaction motions in both static and dynamic settings.

cs.CV↗

arXiv2025

OptFormer: Optical Flow-Guided Attention and Phase Space Reconstruction for SST Forecasting

Sea Surface Temperature (SST) prediction plays a vital role in climate modeling and disaster forecasting. However, it remains challenging due to its nonlinear spatiotemporal dynamics and extended prediction horizons. To address this, we propose OptFormer, a novel encoder-decoder model that integrates phase-space reconstruction with a motion-aware attention mechanism guided by optical flow. Unlike conventional attention, our approach leverages inter-frame motion cues to highlight relative changes in the spatial field, allowing the model to focus on dynamic regions and capture long-range temporal dependencies more effectively. Experiments on NOAA SST datasets across multiple spatial scales demonstrate that OptFormer achieves superior performance under a 1:1 training-to-prediction setting, significantly outperforming existing baselines in accuracy and robustness.

cs.CV↗

arXiv2025

Entangled-to-packed crossover in nonlinear extensional rheology of entangled polymers

Significant challenges exist in the nonlinear extensional rheology of entangled polymers. With simulations, we show that the key to understanding this problem is to recognize the existence and importance of a strain-induced crossover from the entangled state to a packed state. This crossover, following the saturation of primitive chain stretch, takes place with massive release of entanglements via convective flow and progressive chain alignment. After the crossover, the disentangled, fully-aligned chain segments pack similarly to the random packing of rods. Meanwhile, the system enters the steady state. In this state, the tube model, built on the concept of entanglement, fails, while the stress can be quantitatively calculated by combining the intra-chain conformational contribution and a frictional contribution from the inter-chain separation along flow.

cond-mat.soft↗

arXiv2025

Fine-grained text-driven dual-human motion generation via dynamic hierarchical interaction

Human interaction is inherently dynamic and hierarchical, where the dynamic refers to the motion changes with distance, and the hierarchy is from individual to inter-individual and ultimately to overall motion. Exploiting these properties is vital for dual-human motion generation, while existing methods almost model human interaction temporally invariantly, ignoring distance and hierarchy. To address it, we propose a fine-grained dual-human motion generation method, namely FineDual, a tri-stage method to model the dynamic hierarchical interaction from individual to inter-individual. The first stage, Self-Learning Stage, divides the dual-human overall text into individual texts through a Large Language Model, aligning text features and motion features at the individual level. The second stage, Adaptive Adjustment Stage, predicts interaction distance by an interaction distance predictor, modeling human interactions dynamically at the inter-individual level by an interaction-aware graph network. The last stage, Teacher-Guided Refinement Stage, utilizes overall text features as guidance to refine motion features at the overall level, generating fine-grained and high-quality dual-human motion. Extensive quantitative and qualitative evaluations on dual-human motion datasets demonstrate that our proposed FineDual outperforms existing approaches, effectively modeling dynamic hierarchical human interaction.

cs.CV↗

arXiv2025

Data-driven transient growth analysis

Transient growth analysis has been extensively studied in asymptotically stable flows to identify their short-term amplification of perturbations. Generally, in global transient growth analyses, matrix-free methods are adopted, requiring the construction of adjoint equations, either in the discrete or continuous form. This paper introduces a data-driven algorithm that circumvents the adjoint equations by extracting the optimal initial perturbation and its energy growth over a specified time horizon from transient snapshots of perturbations. This method is validated using data from the linearised complex Ginzburg-Landau equation, backward-facing step flow, and the Batchelor vortex. Unlike model-based methods, which require $S$ sets of integrations of the linearised governing equation and its adjoint for $S$ time horizons, the proposed approach collects the snapshots of $S$ time horizons in one integration of the linearised equation. Furthermore, this study provides a robust framework for utilising proper orthogonal decomposition (POD) modes to synthesise optimal modes. The developed capacity to conduct transient growth analyses without solving the adjoint equations is expected to significantly reduce the barriers to transient dynamics research.

physics.flu-dyn↗

arXiv2025

CSRM-LLM: Embracing Multilingual LLMs for Cold-Start Relevance Matching in Emerging E-commerce Markets

As global e-commerce platforms continue to expand, companies are entering new markets where they encounter cold-start challenges due to limited human labels and user behaviors. In this paper, we share our experiences in Coupang to provide a competitive cold-start performance of relevance matching for emerging e-commerce markets. Specifically, we present a Cold-Start Relevance Matching (CSRM) framework, utilizing a multilingual Large Language Model (LLM) to address three challenges: (1) activating cross-lingual transfer learning abilities of LLMs through machine translation tasks; (2) enhancing query understanding and incorporating e-commerce knowledge by retrieval-based query augmentation; (3) mitigating the impact of training label errors through a multi-round self-distillation training strategy. Our experiments demonstrate the effectiveness of CSRM-LLM and the proposed techniques, resulting in successful real-world deployment and significant online gains, with a 45.8% reduction in defect ratio and a 0.866% uplift in session purchase rate.

cs.IR↗

arXiv2025

MOST: Motion Diffusion Model for Rare Text via Temporal Clip Banzhaf Interaction

We introduce MOST, a novel motion diffusion model via temporal clip Banzhaf interaction, aimed at addressing the persistent challenge of generating human motion from rare language prompts. While previous approaches struggle with coarse-grained matching and overlook important semantic cues due to motion redundancy, our key insight lies in leveraging fine-grained clip relationships to mitigate these issues. MOST's retrieval stage presents the first formulation of its kind - temporal clip Banzhaf interaction - which precisely quantifies textual-motion coherence at the clip level. This facilitates direct, fine-grained text-to-motion clip matching and eliminates prevalent redundancy. In the generation stage, a motion prompt module effectively utilizes retrieved motion clips to produce semantically consistent movements. Extensive evaluations confirm that MOST achieves state-of-the-art text-to-motion retrieval and generation performance by comprehensively addressing previous challenges, as demonstrated through quantitative and qualitative results highlighting its effectiveness, especially for rare prompts.

cs.CV↗

arXiv2025

Period-LLM: Extending the Periodic Capability of Multimodal Large Language Model

Periodic or quasi-periodic phenomena reveal intrinsic characteristics in various natural processes, such as weather patterns, movement behaviors, traffic flows, and biological signals. Given that these phenomena span multiple modalities, the capabilities of Multimodal Large Language Models (MLLMs) offer promising potential to effectively capture and understand their complex nature. However, current MLLMs struggle with periodic tasks due to limitations in: 1) lack of temporal modelling and 2) conflict between short and long periods. This paper introduces Period-LLM, a multimodal large language model designed to enhance the performance of periodic tasks across various modalities, and constructs a benchmark of various difficulty for evaluating the cross-modal periodic capabilities of large models. Specially, We adopt an "Easy to Hard Generalization" paradigm, starting with relatively simple text-based tasks and progressing to more complex visual and multimodal tasks, ensuring that the model gradually builds robust periodic reasoning capabilities. Additionally, we propose a "Resisting Logical Oblivion" optimization strategy to maintain periodic reasoning abilities during semantic alignment. Extensive experiments demonstrate the superiority of the proposed Period-LLM over existing MLLMs in periodic tasks. The code is available at https://github.com/keke-nice/Period-LLM.

cs.CV↗

arXiv2025

Neural nanophotonic object detector with ultra-wide field-of-view

Intelligent object detection, which extracts crucial information like targets categories and locations, plays a vital role in emerging technologies including autonomous driving, the Internet of Things, and next-generation mobile communication systems. With the advancement of intelligent object detectors towards higher integration and miniaturization, their portability and adaptability to a broader range of scenarios have been significantly enhanced. However, this progress comes at the cost of reduced detection quality and narrower field-of-view, which severely impacts overall performances. Here we present a neural nanophotonic object detector based on a metalens array, capable of delivering high-quality imaging with an ultra-wide field-of-view of 135{\deg}. The combined neural network not only further improves the imaging quality, but also enables the detector to achieve high-precision target recognition and localization. Moreover, we integrated the neural nanophotonic object detector into a miniature unmanned aerial vehicle to enable wide-angle imaging and intelligent recognition of various real-world dynamic objects, demonstrating the high mobility and flexibility of our neural nanophotonic object detector. Our study presents a systematic framework for advancing revolutionary intelligent detection systems, offering significant potential for a wide range of future applications.

physics.optics↗

arXiv2025

AFD-STA: Adaptive Filtering Denoising with Spatiotemporal Attention for Chaotic System Prediction

This paper presents AFD-STA Net, a neural framework integrating adaptive filtering and spatiotemporal dynamics learning for predicting high-dimensional chaotic systems governed by partial differential equations. The architecture combines: 1) An adaptive exponential smoothing module with position-aware decay coefficients for robust attractor reconstruction, 2) Parallel attention mechanisms capturing cross-temporal and spatial dependencies, 3) Dynamic gated fusion of multiscale features, and 4) Deep projection networks with dimension-scaling capabilities. Numerical experiments on nonlinear PDE systems demonstrate the model's effectiveness in maintaining prediction accuracy under both smooth and strongly chaotic regimes while exhibiting noise tolerance through adaptive filtering. Component ablation studies confirm critical contributions from each module, particularly highlighting the essential role of spatiotemporal attention in learning complex dynamical interactions. The framework shows promising potential for real-world applications requiring simultaneous handling of measurement uncertainties and high-dimensional nonlinear dynamics.

cs.LG↗