Searcharxiv⌕ Search

arXiv subjects

Haibo Wang

Publications and source records attributed to Haibo Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

Learning Geometric Representations from Videos for Spatial Intelligent Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) excel at 2D semantic understanding but lack intrinsic 3D awareness, resulting in representations that fail to maintain geometric and spatial consistency across video frames. Given the scarcity of large-scale 3D data, we present GeoVR, a novel framework that learns geometric representations using purely 2D video sequences. This approach effectively restructures the semantic latent space within MLLMs to unlock spatial intelligence. Rather than employing superficial feature mixing, GeoVR reshapes the internal representations of the MLLM by distilling geometry knowledge from pre-trained 3D foundation models. This is accomplished through a multi-objective learning strategy driven by four complementary geometric targets: (1) estimating inter-frame camera poses to embed varying viewpoint dynamics, (2) regressing dense depth maps to anchor physical distances, (3) predicting a metric scale factor for real-world calibration, and (4) distilling multi-scale 3D features to align the intermediate feature space. Guided by these explicit physical and geometric constraints, the model's internal representations naturally develop strong 3D awareness. Extensive experiments on spatial reasoning benchmarks demonstrate that GeoVR achieves state-of-the-art performance, establishing a new paradigm for endowing foundation models with spatial intelligence.

cs.CV↗
arXiv2026

Accelerating MoE with Dynamic In-Switch Computing on Multi-GPUs

Mixture-of-Experts (MoE) has been adopted by many leading large models to reduce computational requirements. However, frequent inter-GPU communication in MoE expert parallelism (EP) becomes a performance challenge. We observe substantial redundant inter-GPU data transfers in MoE that can be potentially addressed by in-switch computing. Unfortunately, the existing solution, NVLink SHARP (NVLS), can only support static collectives with regular patterns, incapable of dynamic communication with irregular patterns in MoE. To bridge the functionality gap, we propose DySHARP, an integral dynamic in-switch computing solution to accelerate MoE, encompassing both communication primitives and communication-aware scheduling: 1) Dynamic multimem addressing co-designs ISA, architecture, and runtime, as a dynamic extension to NVLS, reducing redundant traffic. However, the resulting traffic reduction is inherently asymmetric between two directions, preventing it from directly translating into speedup. 2) Token-centric kernel fusion deeply fuses the dispatch-computation-combine pipeline, resolving this asymmetry to translate traffic reduction into actual speedup. Compared with the state-of-the-art solution, DySHARP achieves up to 1.79$\times$ speedup.

cs.AR↗
arXiv2026

Towards Compute-Aware In-Switch Computing for LLMs Tensor-Parallelism on Multi-GPU Systems

Tensor parallelism (TP) in large-scale LLM inference and training introduces frequent collective operations that dominate inter-GPU communication. While in-switch computing, exemplified by NVLink SHARP (NVLS), accelerates collective operations by reducing redundant data transfer, its communication-centric design philosophy introduces the mismatch between its communication mode and the memory semantic requirement of LLM's computation kernel. Such a mismatch isolates the compute and communication phases, resulting in underutilized resources and limited overlap in multi-GPU systems. To address the limitation, we propose CAIS, the first Compute-Aware In-Switch computing framework that aligns communication modes with computation's memory semantics requirement. CAIS consists of three integral techniques: (1) compute-aware ISA and microarchitecture extension to enable compute-aware in-switch computing. (2) merging-aware TB (Thread Block) coordination to improve the temporal alignment for efficient request merging. (3) graph-level dataflow optimizer to achieve a tight cross-kernel overlap. Evaluations on LLM workloads show that CAIS achieves 1.38$\times$ average end-to-end training speedup over the SOTA NVLS-enabled solution, and 1.61$\times$ over T3, the SOTA compute-communicate overlap solutions but do not leverage NVLS, demonstrating its effectiveness in accelerating TP on multi-GPU systems.

cs.AR↗
arXiv2026

MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems

The Mixture-of-Experts (MoE) architecture is crucial for scaling large language models, but its scalability is severely limited by inter-GPU communication bottlenecks in multi-GPU systems. Although overlapping communication with computation is a widely recognized optimization, its effective deployment still remains challenging, both in terms of performance and programmability. In this work, we identify the root cause as a fundamental abstraction mismatch between MoE's dynamic, irregular token-to-expert mapping and the static, address-centric communication model of modern GPUs, which necessitates a complex software mediation phase to resolve addresses before data transfers, limiting performance and software flexibility. To resolve this, we propose MoE-Hub, a hardware-software co-design that introduces a destination-agnostic communication paradigm. MoE-Hub decouples data transmission from address management, allowing producers to send data immediately after routing using only a logical destination, while address allocation and data-flow orchestration are handled transparently by lightweight hardware in the GPU hub. By hardware-accelerating the entire communication control plane, MoE-Hub enables seamless and transparent overlap. Our evaluation shows that MoE-Hub achieves 1.40x-3.08x per-layer and 1.21x-1.98x end-to-end speedup over state-of-the-art systems.

cs.AR↗
arXiv2026

Open-Ended Video Game Glitch Detection with Agentic Reasoning and Temporal Grounding

Open-ended video game glitch detection aims to identify glitches in gameplay videos, describe them in natural language, and localize when they occur. Unlike conventional game glitch understanding tasks which have largely been framed as image-level recognition or closed-form question answering, this task requires reasoning about game-specific dynamics such as mechanics, physics, rendering, animation, and expected state transitions directly over continuous gameplay videos and distinguishing true glitches from unusual but valid in-game events. To support this task, we introduce VideoGlitchBench, the first benchmark for open-ended video game glitch detection with temporal localization. VideoGlitchBench contains 5,238 gameplay videos from 120 games, each annotated with detailed glitch descriptions and precise temporal spans, enabling unified evaluation of semantic understanding and temporal grounding. We further propose GliDe, an agentic framework with three key components: a game-aware contextual memory for informed reasoning, a debate-based reflector for multi-perspective glitch detection and verification, and an event-level grounding module that recovers complete glitch intervals from fragmented temporal evidence. We also design a task-specific evaluation protocol that jointly measures semantic fidelity and temporal accuracy. Experiments show that this task remains highly challenging for current multimodal models, while GliDe achieves substantially stronger performance than corresponding vanilla model baselines.

cs.MA↗
arXiv2026

Ethics Testing: Proactive Identification of Generative AI System Harms

Generative Artificial Intelligence (GAI) systems that can automatically generate content in the form of source code or other contents (e.g., images) has seen increasing popularity due to the emergence of tools such as ChatGPT which rely on Large Language Models (LLMs). Misuse of the automatically generated content can incur serious consequences due to potential harms in the generated content. Despite the importance of ensuring the quality of automatically generated content, there is little to no approach that can systematically generate tests for identifying software harms in the content generated by these GAI systems. In this article, we introduce the novel concept of ethics testing which aims to systematically generate tests for identifying software harms. Different from existing testing methodologies (e.g., fairness testing that aims to identifying software discrimination), ethics testing aims to systematically detect software harms that could be induced due to unethical behavior (e.g., harmful behavior or behavior that violates intellectual property rights) in automatically generated content. We introduced the concept of ethics testing, discussed the challenges therewithin, and conducted five case studies to show how ethics testing can be performed for generative AI systems.

cs.SE↗
arXiv2026

Modeling Bank Systemic Risk of Emerging Markets under Geopolitical Shocks: Empirical Evidence from BRICS Countries

In this study, we introduce an analytics framework, the Bank Risk Interlinkage with Dynamic Graph and Event Simulations (BRIDGES), to capture the systemic risks associated with the growing economic influence of the BRICS nations. This framework includes a Dynamic Time Warping (DTW) method to construct a dynamic network of 551 BRICS banks with their annual balance sheet data from 2008 to 2024; a trend analysis in risk ratios to detect shifts in banks' behavior; a Temporal Graph Neural Network (TGNN) to detect anomalous changes in the bank network's structural relationships; and Agent-Based Model (ABM) simulations to measure the impact of anomalous changes on network stability and assess the banking system's resilience to internal financial failure and external geopolitical shocks at the individual country level and across BRICS nations. Our simulation results highlight several important insights. The failure of the largest BRICS banks can cause more systemic damage than that of financially vulnerable or anomalous banks due to the panic effects. Moreover, compared to the failure of the largest BRICS banks, a geopolitical shock with correlated country-wide propagation can cause more systemic damage, resulting in a near-total systemic collapse. Our findings suggest that the panic over the failure of the largest BRICS banks and large-scale geopolitical shocks are the primary threats to the financial stability of the BRICS nations, which traditional bank risk analysis models might not detect.

q-fin.CP↗
arXiv2026

AI-PACE: A Framework for Integrating AI into Medical Education

The integration of artificial intelligence (AI) into healthcare is accelerating, yet medical education has not kept pace with these technological advancements. This paper synthesizes current knowledge on AI in medical education through a comprehensive analysis of the literature, identifying key competencies, curricular approaches, and implementation strategies. The aim is highlighting the critical need for structured AI education across the medical learning continuum and offer a framework for curriculum development. The findings presented suggest that effective AI education requires longitudinal integration throughout medical training, interdisciplinary collaboration, and balanced attention to both technical fundamentals and clinical applications. This paper serves as a foundation for medical educators seeking to prepare future physicians for an AI-enhanced healthcare environment.

cs.CY↗
arXiv2026

GLANCE: A Global-Local Coordination Multi-Agent Framework for Music-Grounded Non-Linear Video Editing

Music-grounded mashup video creation is a challenging form of video non-linear editing, where a system must compose a coherent timeline from large collections of source videos while aligning with music rhythm, user intent, story completeness, and long-range structural constraints. Existing approaches typically rely on fixed pipelines or simplified retrieval-and-concatenation paradigms, limiting their ability to adapt to diverse prompts and heterogeneous source materials. In this paper, we present GLANCE, a global-local coordination multi-agent framework for music-grounded nonlinear video editing. GLANCE adopts a bi-loop architecture for better editing practice: an outer loop performs long-horizon planning and task-graph construction, and an inner loop adopts the "Observe-Think-Act-Verify" flow for segment-wise editing tasks and their refinements. To address the cross-segment and global conflict emerging after subtimelines composition, we introduce a dedicated global-local coordination mechanism with both preventive and corrective components, which includes a novelly designed context controller, conflict region decomposition module, and a bottom-up dynamic negotiation mechanism. To support rigorous evaluation, we construct MVEBench, a new benchmark that factorizes editing difficulty along task type, prompt specificity, and music length, and propose an agent-as-a-judge evaluation framework for scalable multi-dimensional assessment. Experimental results show that GLANCE consistently outperforms prior research baselines and open-source product baselines under the same backbone models. With GPT-4o-mini as the backbone, GLANCE improves over the strongest baseline by 33.2% and 15.6% on two task settings, respectively. Human evaluation further confirms the quality of the generated videos and validates the effectiveness of the proposed evaluation framework.

cs.MA↗
arXiv2026

Think, Act, Build: An Agentic Framework with Vision Language Models for Zero-Shot 3D Visual Grounding

3D Visual Grounding (3D-VG) aims to localize objects in 3D scenes via natural language descriptions. While recent advancements leveraging Vision-Language Models (VLMs) have explored zero-shot possibilities, they typically suffer from a static workflow relying on preprocessed 3D point clouds, essentially degrading grounding into proposal matching. To bypass this reliance, our core motivation is to decouple the task: leveraging 2D VLMs to resolve complex spatial semantics, while relying on deterministic multi-view geometry to instantiate the 3D structure. Driven by this insight, we propose "Think, Act, Build (TAB)", a dynamic agentic framework that reformulates 3D-VG tasks as a generative 2D-to-3D reconstruction paradigm operating directly on raw RGB-D streams. Specifically, guided by a specialized 3D-VG skill, our VLM agent dynamically invokes visual tools to track and reconstruct the target across 2D frames. Crucially, to overcome the multi-view coverage deficit caused by strict VLM semantic tracking, we introduce the Semantic-Anchored Geometric Expansion, a mechanism that first anchors the target in a reference video clip and then leverages multi-view geometry to propagate its spatial location across unobserved frames. This enables the agent to "Build" the target's 3D representation by aggregating these multi-view features via camera parameters, directly mapping 2D visual cues to 3D coordinates. Furthermore, to ensure rigorous assessment, we identify flaws such as reference ambiguity and category errors in existing benchmarks and manually refine the incorrect queries. Extensive experiments on ScanRefer and Nr3D demonstrate that our framework, relying entirely on open-source models, significantly outperforms previous zero-shot methods and even surpasses fully supervised baselines.

cs.CV↗
arXiv2026

Enhancing Credit Risk Prediction: A Multi-stage Ensemble Pipeline

Effective credit risk management is fundamental to financial decision-making, requiring robust models to predict default probabilities and classify financial entities. Traditional machine learning approaches face significant challenges when confronted with high-dimensional data, limited interpretability, rare-event detection, and multi-class risk imbalance. This research proposes a comprehensive multi-stage ensemble pipeline that synthesizes multiple complementary models: econometric models including Ordered logit and ordered probit, supervised learning algorithms, including XGBoost, Random Forest, Support Vector Machine, and Decision Tree; unsupervised methods such as K-Nearest Neighbors; deep learning architectures like Multilayer Perceptron; alongside LASSO regularization for feature selection and dimensionality reduction; and Error-Correcting Output Codes as an Ensemble classifier for handling imbalanced multi-class problems. We implement Permutation Feature Importance analysis for each prediction class across all constituent models to enhance model transparency. Our framework can optimize predictive performance while providing a more holistic approach to credit risk assessment. This research contributes to the development of more accurate and reliable computational models for strategic financial decision support by addressing three fundamental challenges in credit risk modeling. The empirical validation of our approach involves analyzing the Corporate Credit Ratings dataset, which contains credit ratings for 2,029 publicly listed US companies. Results demonstrate that our multi-stage ensemble pipeline significantly enhances the accuracy of financial entity classification regarding credit rating migrations (upgrades and downgrades) and default probability estimation.

cs.LG↗
arXiv2026

LiteCache: A Query Similarity-Driven, GPU-Centric KVCache Subsystem for Efficient LLM Inference

During LLM inference, KVCache memory usage grows linearly with sequence length and batch size and often exceeds GPU capacity. Recent proposals offload KV states to host memory and reduce transfers using top-k attention. But their CPU-centric management of the on-GPU cache and CPU-GPU data movement incurs high overhead and fragments the bulk GPU execution that CUDA Graph relies on. To close this gap, we observe that adjacent queries within the same attention head exhibit strong directional similarity and retrieve highly overlapping top-k KV states. This insight enables a simple head granularity cache algorithm, QSAC, in which each head reuses its previously cached KV states whenever the current query is sufficiently similar to the prior one. QSAC further simplifies cache management primitives and cuts CPU involvement almost entirely. We develop LiteCache, a KVCache subsystem that incorporates QSAC. LiteCache introduces a GPU-centric synchronization controller and speculative sparse prefetching, enabling fully overlapped data movement and computation. These mechanisms produce a stable and predictable execution pattern that remains compatible with the bulk execution mode required by CUDA Graphs. Evaluation on two widely-used LLMs indicates that LiteCache achieves comparable accuracy to baselines, while sharply minimizing CPU overhead, fully utilizing PCIe bandwidth, thus improving decoding throughput by 10.7-224.2% on both H100 and A40 GPUs and easily supporting sequence lengths beyond 1M. We opensource LiteCache at https://anonymous.4open.science/r/LiteCache-888D.

cs.LG↗
arXiv2026

A Heuristic Framework of Variable Neighborhood Descent Methods for the Large-Scale Multi-Level Facility Location Problem in Supply Chain Networks

This paper addresses the single-assignment, uncapacitated, multi-level facility location (MFL) problem, a strategic decision-making process critical to the design of long-term supply chain networks. Specifically, we examine four- and five-level facility location structures (k-LFL), modeled as a location-allocation problem where demand nodes must be assigned to open facilities across hierarchical levels. Although the MFL has been addressed in the literature, solutions to large-scale, realistic problems involving thousands of nodes are lacking. This paper proposes a heuristic framework based on the Variable Neighborhood Descent (VND) metaheuristic with a multi-start strategy. We develop and compare four variants: Basic Variable Neighborhood Descent (BVND), Pipe Variable Neighborhood Descent (PVND), Cyclic Variable Neighborhood Descent (CVND), and Union Variable Neighborhood Descent (UVND). In each case, a multi-start strategy with strong diversification components is employed. Extensive computational experiments compare the methods on large-scale instances involving up to 10,000 customers, 150 distribution centers, 50 warehouses, and 30 plants. Each algorithm settled into a unique, statistically significant computational time when solving these problems. Sensitivity analyses, supported by non-parametric statistical methods, validate the effectiveness of the proposed heuristic framework.

math.OC↗
arXiv2026

Event-Driven Market Co-Movement Dynamics in Critical Mineral Equities: An Empirical Framework Using Change Point Detection and Cross-Sectional Analysis

This study examines market behavior in critical mineral investments using a novel analytical framework that combines change-point detection (PELT algorithm) with cross-sectional analysis. This research analyzes ESG-ranked critical mineral ETFs from March 31, 2014, to April 19, 2024, using the S&P 500 as a benchmark to evaluate market co-movements. The findings demonstrate that different critical mineral investments experienced change points at distinct times, but three major dates, July 23, 2015; March 17, 2020; and December 1, 2020, were common and aligned with global events such as the oil market shock, the COVID-19 pandemic, and later market adjustments. Herding behavior among investors increased after these shocks, following the 2015 and 2020 crises, but shifted to anti-herding after positive vaccine news in late 2020 and after the Russian invasion of Ukraine in 2022. The sensitivity analysis shows that investor coordination is strongest during market downturns but exhibits greater variation during stable periods or after major developments, with these dynamics sensitive to the length of the observation period. Additionally, anti-herding became more apparent during crises, suggesting investors reacted to specific risks rather than moving in lockstep, especially in response to geopolitical shocks.

econ.EM↗
arXiv2026

Dynamic Risk in the U.S. Banking System: An Analysis of Sentiment, Policy Shocks, and Spillover Effects

The 2023 U.S. banking crisis propagated not through direct financial linkages but through a high-frequency, information-based contagion channel. This paper moves beyond exploration analysis to test the "too-similar-to-fail" hypothesis, arguing that risk spillovers were driven by perceived similarities in bank business models under acute interest rate pressure. Employing a Time-Varying Parameter Vector Autoregression (TVP-VAR) model with 30-day rolling windows, a method uniquely suited for capturing the rapid network shifts inherent in a panic, we analyze daily stock returns for the four failed institutions and a systematically selected peer group of surviving banks vulnerable to the same risks from March 18, 2022, to March 15, 2023. Our results provide strong evidence for this contagion channel: total system connectedness surged dramatically during the crisis peak, and we identify SIVB, FRC, and WAL as primary net transmitters of risk while their perceived peers became significant net receivers, a key dynamic indicator of systemic vulnerability that cannot be captured by asset-by-asset analysis. We further demonstrate that these spillovers were significantly amplified by market sentiment (as measured by the VIX) and economic policy uncertainty (EPU). By providing a clear conceptual framework and robust empirical validation, our findings confirm the persistence of systemic risks within the banking network and highlight the importance of real-time monitoring in strengthening financial stability.

econ.EM↗
arXiv2025

An Adaptive Environment-Aware Transformer Autoencoder for UAV-FSO with Dynamic Complexity Control

The rise of sixth-generation (6G) wireless networks sets high demands on UAV-assisted Free Space Optical (FSO) communications, where the channel environment becomes more complex and variable due to both atmospheric turbulence and UAV-induced vibrations. These factors increase the challenge of maintaining reliable communication and require adaptive processing methods. Autoencoders are promising as they learn optimal encodings from channel data. However, existing autoencoder designs are generic and lack the specific adaptability and computational flexibility needed for UAV-FSO scenarios. To address this, we propose AEAT-AE (Adaptive Environment-aware Transformer Autoencoder), a Transformer-based framework that integrates environmental parameters into both encoder and decoder via a cross-attention mechanism. Moreover, AEAT-AE incorporates a Deep Q-Network (DQN) that dynamically selects which layers of the Transformer autoencoder to activate based on real-time environmental inputs, effectively balancing performance and computational cost. Simulation results demonstrate that AEAT-AE outperforms conventional methods in bit error rate while maintaining efficient runtime, representing a novel tailored solution for next-generation UAV-FSO communications.

eess.SY↗
arXiv2025

Efficient Local and Tabu Search Strategies for Large-Scale Quadratic Integer Programming

This study investigates the area of general quadratic integer programming (QIP), encompassing both unconstrained (UQIP) and constrained (CQIP) variants. These NP-hard problems have far-reaching applications, yet the non-convex cases have received limited attention in the literature. To address this gap, we introduce a closed-form formula for single-variable changes, establishing novel necessary and sufficient conditions for 1-Opt local improvement in UQIP and CQIP. We develop a simple local and sophisticated tabu search with an oscillation strategy tailored for large-scale problems. Experimental results on instances with up to 8000 variables demonstrate the efficiency of these strategies, producing high-quality solutions within a short time. Our approaches significantly outperform the Gurobi 11.0.2 solver.

cs.DM↗
arXiv2025

Market Reactions and Information Spillovers in Bank Mergers: A Multi-Method Analysis of the Japanese Banking Sector

Major bank mergers and acquisitions (M&A) transform the financial market structure, but their valuation and spillover effects remain open to question. This study examines the market reaction to two M&A events: the 2005 creation of Mitsubishi UFJ Financial Group following the Financial Big Bang in Japan, and the 2018 merger involving Resona Holdings after the global financial crisis. The multi-method analysis in this research combines several distinct methods to explore these M&A events. An event study using the market model, the capital asset pricing model (CAPM), and the Fama-French three-factor model is implemented to estimate cumulative abnormal returns (CAR) for valuation purposes. Vector autoregression (VAR) models are used to test for Granger causality and map dynamic effects using impulse response functions (IRFs) to investigate spillovers. Propensity score matching (PSM) helps provide a causal estimate of the average treatment effect on the treated (ATT). The analysis detected a significant positive market reaction to the mergers. The findings also suggest the presence of prolonged positive spillovers to other banks, which may indicate a synergistic effect among Japanese banks. Combining these methods provides a unique perspective on M&A events in the Japanese banking sector, offering valuable insights for investors, managers, and regulators concerned with market efficiency and systemic stability

q-fin.CP↗