arXiv subjects

Yan Wang

Publications and source records attributed to Yan Wang.

At least 199 records · Page 11Linked to original sources

arXiv2026

GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal Agents

We present GLM-5V-Turbo, a step toward native foundation models for multimodal agents. As foundation models are increasingly deployed in real environments, agentic capability depends not only on language reasoning, but also on the ability to perceive, interpret, and act over heterogeneous contexts such as images, videos, webpages, documents, GUIs. GLM-5V-Turbo is built around this objective: multimodal perception is integrated as a core component of reasoning, planning, tool use, and execution, rather than as an auxiliary interface to a language model. This report summarizes the main improvements behind GLM-5V-Turbo across model design, multimodal training, reinforcement learning, toolchain expansion, and integration with agent frameworks. These developments lead to strong performance in multimodal coding, visual tool use, and framework-based agentic tasks, while preserving competitive text-only coding capability. More importantly, our development process offers practical insights for building multimodal agents, highlighting the central role of multimodal perception, hierarchical optimization, and reliable end-to-end verification.

cs.CV↗

arXiv2026

A solution to Frankl and Kupavskii's conjecture concerning Erdős-Kleitman matching problem

For integers $n\ge s\ge2$, let $e(n,s)$ be the maximum size of a family $\mathcal F\subseteq2^{[n]}$ with no $s$ pairwise disjoint members. The study of determining $e(n,s)$ is closely related to its uniform counterpart, the well-known Erdős matching conjecture. Frankl and Kupavskii conjectured an exact formula for $e((m+1)s-\ell,s)$ when $1\le \ell\le \lceil s/2\rceil$. We prove that for every fixed $m\ge3$ and sufficiently large $s$, the extremal families for $e((m+1)s-\ell,s)$ are $P(m,s,\ell;L)\coloneqq\{A\subseteq [n]\colon |A|+|A\cap L|\ge m+1\}$ for some $L$ with $|L|=\ell-1$ when $1\le \ell\le (\frac{m+1}{2m+1}-o(1))s$. In particular, this confirms the Frankl--Kupavskii conjecture for every fixed $m\ge3$ and all sufficiently large $s$. For $m=3$, we determine the whole range of $\ell$ for which $P(3,s,\ell;L)$ is extremal, generalizing a theorem of Kupavskii and Sokolov.

math.CO↗

arXiv2026

MindMirror: A Local-First Multimodal State-Aware Support System for Digital Workers

Digital workers often experience fatigue, anxiety, reduced attention, and task blockage during prolonged computer-based work. Existing productivity tools mainly focus on task completion, while general-purpose AI chatbots require users to formulate clear prompts before receiving useful help. This paper presents MindMirror, a local-first multimodal state-aware support system for digital workers. MindMirror integrates camera-based facial expression cues, text input, optional speech interaction, structured blockage reflection, local large language model (LLM)-based response generation, and daily/weekly review reports. The system forms a closed workflow of state checking, manual correction, structured articulation, suggestion generation, and state review. The current prototype follows a local-first design, while optional speech services may rely on third-party APIs when enabled. It is implemented with a Web frontend, Flask backend, an emotion recognition model, an Ollama-hosted Qwen model, Chart.js visualization, and local JSON/LocalStorage records. We evaluate the emotion recognition module on an independent seven-class image-level facial expression benchmark containing 6,767 images. The fine-tuned Hugging Face model improves accuracy from 59.66% to 94.49% over a non-fine-tuned checkpoint baseline, an absolute gain of 34.83 percentage points. We further validate the prototype through endpoint-level reliability tests, voice-interaction latency tests, and a small formative user feedback study with six digital workers. Results suggest that users value the local-first design, manual correction mechanism, and structured reflection workflow. MindMirror is not intended for psychological diagnosis; instead, it serves as a lightweight, user-controllable tool for state reflection and supportive interaction.

cs.HC↗

arXiv2026

Career Mobility of Planning Alumni in the United States: Evidence from Professional Profile Data using Large Language Models

Problem, Research Strategy, and Findings: Planning professions in the United States navigate complex and dynamic career landscapes under rapid urban changes, yet comprehensive evidence regarding their career trajectories, advancement patterns, and the influence of social, spatial, organizational, and educational factors remains limited. This study draws on boundaryless career theory, social capital theory, and spatial opportunity models to analyze career mobility among more than 130,000 planning alumni. Using large language models to extract structured information from LinkedIn profiles, our results reveal that planning alumni who adopt boundaryless career patterns, specifically multisector experience or lateral and industry-switching trajectories, achieve significantly higher upward mobility. While technical competencies provide a foundational entry-level signal, soft skills leveraged through strategic lateral moves become increasingly decisive as planners reach senior stages. Geographic mobility and employment in larger, diverse metropolitan labor markets are both associated with advancement, though the latter provides modest benefits. Larger professional networks and greater organizational engagement are consistently associated with upward career transitions, while AI-related skills, now commonplace, present limited additional advantage. Limitations include reliance on LinkedIn data, which may underrepresent alumni without online profiles, and an individual-level focus that omits organizational factors.

cs.CY↗

arXiv2026

When Less is More: The LLM Scaling Paradox in Context Compression

Scaling up model parameters has long been a prevalent training paradigm driven by the assumption that larger models yield superior generation capabilities. However, under lossy context compression in a compressor--decoder setup, we find a \textbf{\textit{Size-Fidelity Paradox}}: increasing compressor size can lessen the faithfulness of reconstructed contexts though reconstruction error decreases. Across 27 compressor setups spanning model families, scales, and compression rates, we coin this paradox arising from two dominant factors: 1) \textit{knowledge overwriting}: larger models increasingly replace source facts with their own prior beliefs, \textit{e.g.}, ``the white strawberry`` $\to$ ``the red strawberry``; and 2) \textit{semantic drift}: larger models tend to paraphrase or restructure content instead of reproducing it verbatim, \textit{e.g.}, ``Alice hit Bob`` $\to$ ``Bob hit Alice``. Interestingly, this paradox persists across varied settings, with mid-sized compressors often outperforming larger ones in faithful recovery. By analyzing the compressed memory via embedding geometry and reconstruction determinacy, we further reveal that compressors tend to organize memory across broader semantic subspaces, yielding more ambiguous representations prone to overwriting, drift, and weakened recovery. These findings complement existing evaluations of context compression and expose a breakdown of scaling laws when the objective shifts from plausible generation to faithful preservation.

cs.LG↗

arXiv2026

On the Borodin--Kostochka conjecture for graphs with large maximum degree

The Borodin--Kostochka conjecture states that every graph $G$ with maximum degree $Δ(G)\ge 9$ satisfies $χ(G)\le \max\{ω(G),Δ(G)-1\}$. In this paper, we verify this conjecture for graphs with sufficiently large maximum degree. More precisely, we prove that every graph $G$ with maximum degree $Δ\ge 5.3\times 10^6$ and clique number $ω(G)<Δ$ satisfies $χ(G)\le Δ-1$. This improves a longstanding result of Reed.

math.CO↗

arXiv2026

CapVector: Learning Transferable Capability Vectors in Parametric Space for Vision-Language-Action Models

This paper proposes a novel approach to address the challenge that pretrained VLA models often fail to effectively improve performance and reduce adaptation costs during standard supervised finetuning (SFT). Some advanced finetuning methods with auxiliary training objectives can improve performance and reduce the number of convergence steps. However, they typically incur significant computational overhead due to the additional losses from auxiliary objectives. To simultaneously achieve the enhanced capabilities of auxiliary training with the simplicity of standard SFT, we decouple the two objectives of auxiliary-objective SFT within the parameter space, namely, enhancing general capabilities and fitting task-specific action distributions. To deliver the goal, we only need to train the model to converge on a small-scale task set using two distinct training strategies, resulting in two finetuned models. The parameters' difference between the two models can then be interpreted as capability vectors provided by auxiliary objectives. These vectors are then merged with pretrained parameters to form a capability-enhanced meta model. Moreover, when standard SFT is augmented with a lightweight orthogonal regularization loss, the merged model attains performance comparable to auxiliary finetuned baselines with reduced computational overhead. Internal and external experiments demonstrate that our capability vectors (1) are effective and versatile across diverse models, (2) can generalize to novel environments and embodiments out of the box.

cs.CV↗

arXiv2026

RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark

Memory is a critical component of robotic intelligence, as robots must rely on past observations and actions to accomplish long-horizon tasks in partially observable environments. However, existing robotic memory benchmarks still lack multimodal annotations for memory formation, provide limited task coverage and structural complexity, and remain restricted to simulation without real-world evaluation. We address this gap with RoboMemArena, a large-scale benchmark of 26 tasks, with average trajectory lengths exceeding 1,000 steps per task and 68.9% of subtasks being memory-dependent. The generation pipeline leverages a vision-language model (VLM) to design and compose subtasks, generates full trajectories through atomic functions, and provides memory-related annotations, including subtask instructions and native keyframe annotations, while paired real-world memory tasks support physical evaluation. We further design PrediMem, a dual-system VLA in which a high-level VLM planner manages a memory bank with recent and keyframe buffers and uses a predictive coding head to improve sensitivity to task dynamics. Extensive experiments on RoboMemArena show that PrediMem outperforms all baselines and provides insights into memory management, model architecture, and scaling laws for complex memory systems.

cs.RO↗

arXiv2026

Cross-Sample Relational Fusion: Unifying Domain Generalization and Class-Incremental Learning

Class-Incremental Learning (CIL) requires a learning system to learn new classes while retaining previously learned knowledge. However, in real-world scenarios such as autonomous driving, a system trained on urban roads in sunny weather may later need to operate in rural or highway environments with different traffic patterns and weather conditions. This requires the model not only to overcome catastrophic forgetting, but also to effectively handle domain shifts. In this paper, we propose CrOss-sample Relational Fusion (CORF), a unified framework to address domain shift and catastrophic forgetting simultaneously. To enhance generalizability, we perform selective refinement of training samples by leveraging spatial contribution maps to highlight semantically informative regions. Furthermore, we incorporate predictive confidence to adaptively weigh samples, thereby facilitating the learning of domain-agnostic representations. To alleviate forgetting, we propose a cascaded distillation framework that captures cross-sample relational dependencies across multiple feature hierarchies, enabling multi-grained knowledge transfer from previous tasks. CORF can be seamlessly integrated into existing CIL algorithms to enhance their generalizability, achieving competitive performance across various benchmark datasets. Code is available at https://github.com/LAMDA-CL/TMM26-CORF .

cs.CV↗

arXiv2026

SoLAR: Error-Resilient Streamable Long-Horizon Free-Viewpoint Video Reconstruction with Anchor Activation and Latent Recalibration

Free-Viewpoint Video (FVV) has emerged as a cornerstone of next-generation immersive media systems and attracted widespread attention. Previous methods primarily focus on short video sequences and suffer from significant performance degradation when processing long-horizon free-viewpoint video (LFVV). Motivated by bit allocation theory, we analyze dynamic-anchor-based volumetric video representation within a rate-distortion optimization framework and propose \textbf{SoLAR}, which is the first error-resilient streamable FVV framework that maintains stable reconstruction quality on long sequences without requiring group-of-pictures partitioning. We propose the Anchor Activation Dynamics (AAD), which enables dynamic anchors to model non-rigid transformations by dynamically activating informative anchors and suppressing redundant ones. Furthermore, we introduce Latent Discrepancy Aware Recalibration (LaDAR), which is a mechanism to identify discrepancies between latent representations and recalibrate the correspondences encoded in the network, effectively mitigating error propagation in LFVV without compromising real-time performance or storage compactness. Extensive experiments demonstrate that \textbf{SoLAR} achieves state-of-the-art reconstruction performance while maintaining minimum storage overhead, which provides a new direction for LFVV reconstruction and advances the practical deployment of immersive systems. Demo free-viewpoint videos are provided in the supplementary material.

cs.CV↗

arXiv2026

Making Reconstruction FID Predictive of Diffusion Generation FID

It is well known that the reconstruction FID (rFID) of a VAE is poorly correlated with the generation FID (gFID) of a latent diffusion model. We propose interpolated FID (iFID), a simple variant of rFID that exhibits a strong correlation with gFID. Specifically, for each dataset element, we retrieve its nearest neighbor in latent space, interpolate between their latent representations, decode the interpolated latent, and compute the FID between the decoded samples and the original dataset. We provide an intuitive explanation for why iFID correlates well with gFID, and why reconstruction metrics can be negatively correlated with gFID, by connecting iFID to recent results on diffusion generalization and hallucination. Theoretically, we show that iFID evaluates decoded interpolations aligned with the ridge set around which diffusion samples concentrate, thereby measuring a quantity closely related to diffusion sample quality. Empirically, iFID is the first metric shown to strongly correlate with diffusion gFID across diverse VAEs, achieving Pearson and Spearman correlations of approximately $0.85$. The project page is available at https://tongdaxu.github.io/pages/ifid.html.

cs.CV↗

arXiv2026

ClawMark: A Living-World Benchmark for Multi-Turn, Multi-Day, Multimodal Coworker Agents

Language-model agents are increasingly used as persistent coworkers that assist users across multiple working days. During such workflows, the surrounding environment may change independently of the agent: new emails arrive, calendar entries shift, knowledge-base records are updated, and evidence appears across images, scanned PDFs, audio, video, and spreadsheets. Existing benchmarks do not adequately evaluate this setting because they typically run within a single static episode and remain largely text-centric. We introduce \bench{}, a benchmark for coworker agents built around multi-turn multi-day tasks, a stateful sandboxed service environment whose state evolves between turns, and rule-based verification. The current release contains 100 tasks across 13 professional scenarios, executed against five stateful sandboxed services (filesystem, email, calendar, knowledge base, spreadsheet) and scored by 1537 deterministic Python checkers over post-execution service state; no LLM-as-judge is invoked during scoring. We benchmark seven frontier agent systems. The strongest model reaches 75.8 weighted score, but the best strict Task Success is only 20.0\%, indicating that partial progress is common while complete end-to-end workflow completion remains rare. Turn-level analysis shows that performance drops after the first exogenous environment update, highlighting adaptation to changing state as a key open challenge. We release the benchmark, evaluation harness, and construction pipeline to support reproducible coworker-agent evaluation.

cs.CV↗

arXiv2026

Observation of Non-Markovian Evolution of Tripartite Quantum Steering

The memory effects in open quantum systems can induce information backflow and revive quantum correlations, thereby providing a powerful way to protect and recover useful quantum resources in realistic noisy environments. However, such dynamics remains experimentally unexplored in multipartite quantum steering. Here we observe different non-Markovian evolution of tripartite quantum steering using Greenberger-Horne-Zeilinger-type mixed states, covering both death and revival processes. In particular, we experimentally demonstrate the more intricate asymmetric steering structure of tripartite quantum steering through different bipartitions, which do not arise in bipartite systems. Our results provide foundational insights into the hierarchical and directional structures in multipartite quantum steering, and highlight its potential as a useful resource for asymmetric quantum information processing.

quant-ph↗

arXiv2026

SynMotion: Semantic-Visual Adaptation for Motion Customized Video Generation

Diffusion-based video motion customization facilitates the acquisition of human motion representations from a few video samples, while achieving arbitrary subjects transfer through precise textual conditioning. Existing approaches often rely on semantic-level alignment, expecting the model to learn new motion concepts and combine them with other entities (e.g., ''cats'' or ''dogs'') to produce visually appealing results. However, video data involve complex spatio-temporal patterns, and focusing solely on semantics cause the model to overlook the visual complexity of motion. Conversely, tuning only the visual representation leads to semantic confusion in representing the intended action. To address these limitations, we propose SynMotion, a new motion-customized video generation model that jointly leverages semantic guidance and visual adaptation. At the semantic level, we introduce the dual-embedding semantic comprehension mechanism which disentangles subject and motion representations, allowing the model to learn customized motion features while preserving its generative capabilities for diverse subjects. At the visual level, we integrate parameter-efficient motion adapters into a pre-trained video generation model to enhance motion fidelity and temporal coherence. Furthermore, we introduce a new embedding-specific training strategy which \textbf{alternately optimizes} subject and motion embeddings, supported by the manually constructed Subject Prior Video (SPV) training dataset. This strategy promotes motion specificity while preserving generalization across diverse subjects. Lastly, we introduce MotionBench, a newly curated benchmark with diverse motion patterns. Experimental results across both T2V and I2V settings demonstrate that \method outperforms existing baselines. Project page: https://lucaria-academy.github.io/SynMotion/

cs.CV↗

arXiv2026

Lifting Unlabeled Internet-level Data for 3D Scene Understanding

Annotated 3D scene data is scarce and expensive to acquire, while abundant unlabeled videos are readily available on the internet. In this paper, we demonstrate that carefully designed data engines can leverage web-curated, unlabeled videos to automatically generate training data, to facilitate end-to-end models in 3D scene understanding alongside human-annotated datasets. We identify and analyze bottlenecks in automated data generation, revealing critical factors that determine the efficiency and effectiveness of learning from unlabeled data. To validate our approach across different perception granularities, we evaluate on three tasks spanning low-level perception, i.e., 3D object detection and instance segmentation, to high-evel reasoning, i.e., 3D spatial Visual Question Answering (VQA) and Vision-Lanugage Navigation (VLN). Models trained on our generated data demonstrate strong zero-shot performance and show further improvement after finetuning. This demonstrates the viability of leveraging readily available web data as a path toward more capable scene understanding systems.

cs.CV↗

arXiv2026

Final states of two-dimensional turbulence above large-scale topography: stationary vortex solutions and barotropic stability

The final states of freely decaying two-dimensional (2D) topographic turbulence consist of a background flow and localized vortices. While the background flow satisfies a linear potential vorticity (PV)-streamfunction relation, the vortex structures remain poorly understood. To address this gap and ensure oceanic relevance, we examine quasi-stationary final states of 2D turbulence over a sinusoidal topography featuring a bump and a dip, where two oppositely signed vortices are locked to the topographic extrema. After subtracting the background flow, the vortices exhibit a "sinh"-like PV-streamfunction relation, as observed in flat-bottom turbulence. Motivated by Gaussian vortex profiles in flat-bottom turbulence, we propose an empirical model combining the background flow with Gaussian vortices centered at the topographic extrema. This model accurately reproduces quasi-stationary states and yields locally stationary solutions to the inviscid governing equation. We further test the model under complex topography and high-energy conditions, confirming that the "sinh"-like trend and Gaussian profiles are robust features of localized vortices. Linear stability analyses of these stationary vortex solutions reveal background flow-dependent stability: cyclone/elevation and anticyclone/depression configurations are stable at low background energy, while anticyclone/elevation and cyclone/depression configurations are stable at high background energy. These findings align with vortex-topography correlations observed in simulations across energy regimes. Our results provide explicit vortex solutions for quasi-stationary final states of 2D topographic turbulence and elucidate the mechanism underlying vortex-topography correlations through stability analyses of vortices embedded in topographic background flows.

physics.flu-dyn↗

arXiv2026

Marginal Girsanov Reweighting: Stable Variance Reduction for Long-Timescale Dynamics from Biased Simulation

Recovering unbiased kinetic and thermodynamic observables from the enhanced sampling simulations is a central challenge in rare-event sampling. Classical Girsanov Reweighting (GR) offers a principled solution by yielding exact pathwise probability ratios between biased and unbiased processes. However, the variance of GR weights grows rapidly with time, rendering it impractical for long-horizon reweighting. We introduce Marginal Girsanov Reweighting (MGR), which mitigates variance explosion by marginalizing over intermediate paths, producing stable and scalable weights for long-timescale dynamics. Experiments on various molecular dynamics systems demonstrate that MGR accurately recovers unbiased kinetic properties from trajectories generated under both umbrella sampling and metadynamics biases.

q-bio.QM↗

arXiv2026

Compressing then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding

Multimodal Large Language Models advance multimodal representation learning by acquiring transferable semantic embeddings, thereby substantially enhancing performance across a range of vision-language tasks, including cross-modal retrieval, clustering, and classification. An effective embedding is expected to comprehensively preserve the semantic content of the input while simultaneously emphasizing features that are discriminative for downstream tasks. Recent approaches demonstrate that MLLMs can be adapted into competitive embedding models via large-scale contrastive learning, enabling the simultaneous optimization of two complementary objectives. We argue that the two aforementioned objectives can be decoupled: a comprehensive understanding of the input enables the embedding model to achieve superior performance on downstream tasks via contrastive learning. In this paper, we propose CoMa, a compressed pre-training phase, which serves as a warm-up stage for contrastive learning. Experiments demonstrate that with only a small amount of pre-training data, we can transform an MLLM into a competitive embedding model. CoMa achieves new state-of-the-art results among MLLMs of comparable size on the MMEB, realizing optimization in both efficiency and effectiveness. Our project is available at https://github.com/Trustworthy-Information-Access/CoMa.

cs.CV↗