arXiv subjects

Wenyu Gao

Publications and source records attributed to Wenyu Gao.

7 recordsLinked to original sources

arXiv2026

LISA-3D: Lifting Language-Image Segmentation to 3D via Multi-View Consistency

Text-driven 3D reconstruction requires masks that understand free-form instructions and remain stable under viewpoint changes. We present LISA-3D, a two-stage framework that adapts the instruction-following segmenter LISA with geometry-aware Low-Rank Adaptation (LoRA) layers while keeping the SAM-3D reconstructor frozen. During training, paired RGB-D frames and camera poses define a differentiable reprojection loss that enforces cross-view agreement without additional 3D-text annotations. At deployment, the adapted segmenter can produce an RGBA prompt for SAM-3D from one RGB image; when registered RGB-D views are available, optional logit fusion further improves the prompt. On ScanRefer and Nr3D, geometry-aware tuning improves both 2D masks and lifted 3D reconstructions while updating only 11.6M parameters. Our results separate geometry-aware training gains from optional multi-view inference gains, providing a modular route from language grounding to object-centric 3D reconstruction.

cs.CV↗

arXiv2026

Can LLMs See Without Pixels? Benchmarking Spatial Intelligence from Textual Descriptions

Recent advancements in Spatial Intelligence (SI) have predominantly relied on Vision-Language Models (VLMs), yet a critical question remains: does spatial understanding originate from visual encoders or the fundamental reasoning backbone? Inspired by this question, we introduce SiT-Bench, a novel benchmark designed to evaluate the SI performance of Large Language Models (LLMs) without pixel-level input, comprises over 3,800 expert-annotated items across five primary categories and 17 subtasks, ranging from egocentric navigation and perspective transformation to fine-grained robotic manipulation. By converting single/multi-view scenes into high-fidelity, coordinate-aware textual descriptions, we challenge LLMs to perform symbolic textual reasoning rather than visual pattern matching. Evaluation results of state-of-the-art (SOTA) LLMs reveals that while models achieve proficiency in localized semantic tasks, a significant "spatial gap" remains in global consistency. Notably, we find that explicit spatial reasoning significantly boosts performance, suggesting that LLMs possess latent world-modeling potential. Our proposed dataset SiT-Bench serves as a foundational resource to foster the development of spatially-grounded LLM backbones for future VLMs and embodied agents. Our code and benchmark will be released at https://github.com/binisalegend/SiT-Bench .

cs.CV↗

arXiv2025

Stability Conditions and Harder-Narasimhan Filtrations for Triangulated Categories

In this paper, we investigate the relationships between Harder-Narasimhan filtrations and derived Hall algebras. We extend several results from abelian categories to triangulated categories, including Reineke inversions, wall-crossing formulas, and Joyce's elements $ε_γ$. The results in triangulated categories can be summarized via a diagram of the same form of that in abelian categories. As an application, we characterize all possibilities for stability conditions on $D^b(\mathrm{rep} A_2)$.

math.RT↗

arXiv2025

EcomBench: Towards Holistic Evaluation of Foundation Agents in E-commerce

Foundation agents have rapidly advanced in their ability to reason and interact with real environments, making the evaluation of their core capabilities increasingly important. While many benchmarks have been developed to assess agent performance, most concentrate on academic settings or artificially designed scenarios while overlooking the challenges that arise in real applications. To address this issue, we focus on a highly practical real-world setting, the e-commerce domain, which involves a large volume of diverse user interactions, dynamic market conditions, and tasks directly tied to real decision-making processes. To this end, we introduce EcomBench, a holistic E-commerce Benchmark designed to evaluate agent performance in realistic e-commerce environments. EcomBench is built from genuine user demands embedded in leading global e-commerce ecosystems and is carefully curated and annotated through human experts to ensure clarity, accuracy, and domain relevance. It covers multiple task categories within e-commerce scenarios and defines three difficulty levels that evaluate agents on key capabilities such as deep information retrieval, multi-step reasoning, and cross-source knowledge integration. By grounding evaluation in real e-commerce contexts, EcomBench provides a rigorous and dynamic testbed for measuring the practical capabilities of agents in modern e-commerce.

cs.AI↗

arXiv2025

Beyond Flatlands: Unlocking Spatial Intelligence by Decoupling 3D Reasoning from Numerical Regression

Existing Vision Language Models (VLMs) architecturally rooted in "flatland" perception, fundamentally struggle to comprehend real-world 3D spatial intelligence. This failure stems from a dual-bottleneck: input-stage conflict between computationally exorbitant geometric-aware encoders and superficial 2D-only features, and output-stage misalignment where discrete tokenizers are structurally incapable of producing precise, continuous numerical values. To break this impasse, we introduce GEODE (Geometric-Output and Decoupled-Input Engine), a novel architecture that resolves this dual-bottleneck by decoupling 3D reasoning from numerical generation. GEODE augments main VLM with two specialized, plug-and-play modules: Decoupled Rationale Module (DRM) that acts as spatial co-processor, aligning explicit 3D data with 2D visual features via cross-attention and distilling spatial Chain-of-Thought (CoT) logic into injectable Rationale Tokens; and Direct Regression Head (DRH), an "Embedding-as-Value" paradigm which routes specialized control tokens to a lightweight MLP for precise, continuous regression of scalars and 3D bounding boxes. The synergy of these modules allows our 1.5B parameter model to function as a high-level semantic dispatcher, achieving state-of-the-art spatial reasoning performance that rivals 7B+ models.

cs.CV↗

arXiv2025

Scalable Generalized Meta-Spanners Enabling Parallel Multitasking Optical Manipulation

Optical manipulation techniques offer exceptional contactless control but are fundamentally limited in their ability to perform parallel multitasking. To achieve high-density, versatile manipulation with subwavelength photonic devices, it is essential to sculpt light fields in multiple dimensions. Here, we overcome this challenge by introducing generalized optical meta-spanners (GOMSs) based on metasurfaces. Relying on complex-amplitude modulation, this platform generates lens-free, customizable optical fields that suppress diffractive losses. As a result, several advanced functionalities are simultaneously achieved, including longitudinally varying manipulation and in-plane spanner arrays, which outperforms the same operations realized by conventional donut-shaped orbital flows. Furthermore, the particle dynamics is reconfigurable simply by switching the input and output polarizations, facilitating robust multi-channel control. We experimentally validate the proposed approach by demonstrating single-particle dynamics and the parallel manipulation of particle ensembles, revealing exceptional stability for multitasking operations. These results demonstrate an ultracompact platform scalable to a much larger number of optical spanners, advancing metadevices from wavefront sculptors to particle manipulators. We envision that the GOMS will catalyze innovations in cross-disciplinary fields such as targeted drug delivery and cell-level biomechanics.

physics.optics↗

arXiv2021

Can Predominant Credible Information Suppress Misinformation in Crises? Empirical Studies of Tweets Related to Prevention Measures during COVID-19

During COVID-19, misinformation on social media affects the adoption of appropriate prevention behaviors. It is urgent to suppress the misinformation to prevent negative public health consequences. Although an array of studies has proposed misinformation suppression strategies, few have investigated the role of predominant credible information during crises. None has examined its effect quantitatively using longitudinal social media data. Therefore, this research investigates the temporal correlations between credible information and misinformation, and whether predominant credible information can suppress misinformation for two prevention measures (i.e. topics), i.e. wearing masks and social distancing using tweets collected from February 15 to June 30, 2020. We trained Support Vector Machine classifiers to retrieve relevant tweets and classify tweets containing credible information and misinformation for each topic. Based on cross-correlation analyses of credible and misinformation time series for both topics, we find that the previously predominant credible information can lead to the decrease of misinformation (i.e. suppression) with a time lag. The research findings provide empirical evidence for suppressing misinformation with credible information in complex online environments and suggest practical strategies for future information management during crises and emergencies.

cs.SI↗