arXiv subjects

Lin Gao

Publications and source records attributed to Lin Gao.

At least 19 recordsLinked to original sources

arXiv2026

GLIMPSE: Global-Local Indexing with Multi-Precision Spectral Encoding for Explorable Video World Models

A camera-controlled video world model must keep a scene consistent when the camera returns to it. Memories that render stored features into the target frame can turn depth and pose errors into ghosting or duplicated structure. Positional transport avoids that resampling by relocating a historical patch's address rather than its content, but assigns full positional precision even to inaccurate projections. We present GLIMPSE, a global-local memory interface with band-wise multi-precision coordinates and geometry-conditioned spectral attenuation to reduce sensitivity to inaccurate projected addresses. Local patch memory supplies view-specific appearance, while recurrent global memory provides scene context. Footprint-guided retrieval selects historical patches under a fixed local read budget. Under matched retrieval and read budgets, GLIMPSE improves geometric correspondence and photometric consistency over deterministic positional transport.

cs.CV↗

arXiv2026

GSHeadRelight: Fast Relightability for 3D Gaussian Head Synthesis

Relighting and novel view synthesis of human portraits are essential in applications such as portrait photography, virtual reality (VR), and augmented reality (AR). Despite recent progress, 3D-aware portrait relighting remains challenging due to the demands for photorealistic rendering, real-time performance, and generalization to unseen subjects. Existing works either rely on supervision from limited and expensive light stage captured data or produce suboptimal results. Moreover, many works are based on generative NeRFs, which suffer from poor 3D consistency and low real-time performance. We resort to recent progress on generative 3D Gaussians and design a lighting model based on a unified neural radiance transfer representation, which responds linearly to incident light. Using only in-the-wild images, our method achieves state-of-the-art relighting results and a significantly faster rendering speed ($\times12$) compared to previous 3D-aware portrait relighting research. Project page: http://www.geometrylearning.com/GSHeadRelight/.

cs.CV↗

arXiv2026

JointLight: Human-Centric Video Relighting with Asynchronous RGB-Irradiance Joint Modeling

Understanding and modeling light transport in human-centric scenes is a longstanding challenge. While recent progress in diffusion models has boosted research in relighting, precise illumination control still lags, primarily due to the scarcity of diverse, high-quality, precisely labeled video relighting datasets. Existing end-to-end relighting approaches model light transport implicitly through neural networks, which substantially complicates controllability. In contrast, some works resort to path-traced irradiance cues as an explicit blueprint for light transport. However, such constructed irradiance cues are prone to artifacts and hinder generative quality when directly used as control signals. With the key observation that noise corruption removes high-frequency artifacts and preserves low-frequency light transport, we introduce an asynchronous joint-modeling paradigm that facilitates fine-grained illumination control. During two-stage training, we force the model to jointly denoise RGB and ground-truth irradiance frames with separate timesteps scheduling. A domain matching strategy is applied to the second stage to bridge the domain gap between irradiance cues and ground-truth irradiance maps. At inference time, we introduce an asynchronous hijacked generation pipeline that facilitates flexible illumination control. For training data, we develop a multi-illumination high-quality synthetic video dataset with paired irradiance maps using a powerful game engine. Trained exclusively on this dataset, our model exhibits robust generalization to real- world scenarios and outperforms prior approaches in both lighting fidelity and generative quality. Project page: http://www.geometrylearning.com/JointLight/

cs.CV↗

arXiv2026

Inflak: A Unified Orchestration Architecture for Human-Agent Interaction

LLM-powered agents increasingly support open-ended activities, yet orchestrating when and how agents interact with humans remains challenging. Existing approaches often provide interaction mechanisms tied to particular tasks or execution patterns, lacking a unified architecture for diverse and evolving interaction needs. We introduce Inflak, a unified four-layer architecture for human-agent interaction orchestration that separates state reconstruction and routing, interaction and execution planning, interaction design, and interface rendering. Its layered design with progressive resolution enables general applicability, runtime flexibility, module composability, and orchestration maintainability. We implement Inflak as agent-native, skill-based plugins within a general-purpose agent framework and evaluate it through a gallery of various application scenarios and interaction paradigms, technical performance testing, and a comparative implementation case study. The evaluations show that Inflak can serve as a general and extensible foundation for orchestrating diverse human-agent interaction flows.

cs.HC↗

arXiv2026

Representation Dynamics Reveal Semantic Saliency and Similarity for Visual Token Pruning in MLLMs

Multimodal large language models (MLLMs) incur high inference latency from long visual token sequences. Existing pruning methods commonly use attention maps or output features to estimate token importance or redundancy. Several recent approaches also exploit representation changes, but when and how these changes reflect foreground saliency and semantic consistency remain insufficiently understood. We analyze visual token representation dynamics across encoder depth and uncover two findings. First, the relationship between token update magnitudes and foreground saliency is layer-dependent: large token updates concentrate on foreground regions in two depth intervals, separated by several sink-dominated layers at intermediate depths. Second, similarities between token update directions better distinguish same-class from different-class tokens than those between encoder output features. Building on these findings, we propose MSDG-Prune, a training-free method that uses update magnitudes and directions to preserve salient and diverse visual information. Specifically, we group tokens by update-direction similarity and use query-weighted saliency derived from update magnitudes across a chosen depth window for group-wise token pruning. Extensive experiments across four MLLMs demonstrate the effectiveness and generalizability of MSDG-Prune. On LLaVA-NeXT, it retains 91.9% of uncompressed performance on average with only 5.6% of visual tokens, while achieving a 7.8x prefilling speedup. Code is available at https://github.com/liweixuan-hitsz/MSDG-Prune.

cs.CV↗

arXiv2026

MoRA: MoE Pruning via Router Bias Learning and Expert Approximation

Mixture-of-Experts (MoE) models enable parameter scaling with limited per-token computation by activating only a small subset of experts for each token, but deploying them still requires loading the complete expert pool into memory. Structured expert pruning can effectively reduce the memory usage by removing experts. However, existing pruning methods either use expert ranking criteria that are not well aligned with model performance or rely on effective expert subset searching that is computationally expensive. Moreover, these methods typically overlook the routing-behavior redundancy among the retained experts. In this paper, we propose MoE Pruning via Router Bias Learning and Expert Approximation (MoRA), a framework for structured MoE expert pruning. We introduce a learnable router bias for each expert and optimize these biases by minimizing the language-modeling loss and a routing-diversity regularizer. The learned router biases sharpen the routing probability distributions to identify experts critical to model performance while encouraging the selection of experts with diverse routing preferences. In addition, we introduce an expert approximation mechanism as a post-pruning enhancement. It leverages the remaining experts to approximate the outputs of pruned experts by affine transformation, further improving the performance of the pruned model. We evaluate MoRA on Qwen3-30B-A3B, DeepSeek-V2-Lite, and Moonlight-16B-A3B, removing 25\% and 50\% of the routed experts in each MoE layer. Extensive experiments on nine zero-shot benchmarks show that MoRA outperforms state-of-the-art pruning algorithms. Our code will be released.

cs.LG↗

arXiv2026

Beyond Pixels: Visual Metaphor Transfer via Schema-Driven Agentic Reasoning

A visual metaphor constitutes a high-order form of human creativity, employing cross-domain semantic fusion to transform abstract concepts into impactful visual rhetoric. Despite the remarkable progress of generative AI, existing models remain largely confined to pixel-level instruction alignment and surface-level appearance preservation, failing to capture the underlying abstract logic necessary for genuine metaphorical generation. To bridge this gap, we introduce the task of Visual Metaphor Transfer (VMT), which challenges models to autonomously decouple the "creative essence" from a reference image and re-materialize that abstract logic onto a user-specified target subject. We propose a cognitive-inspired, multi-agent framework that operationalizes Conceptual Blending Theory (CBT) through a novel Schema Grammar ("G"). This structured representation decouples relational invariants from specific visual entities, providing a rigorous foundation for cross-domain logic re-instantiation. Our pipeline executes VMT through a collaborative system of specialized agents: a perception agent that distills the reference into a schema, a transfer agent that maintains generic space invariance to discover apt carriers, a generation agent for high-fidelity synthesis and a hierarchical diagnostic agent that mimics a professional critic, performing closed-loop backtracking to identify and rectify errors across abstract logic, component selection, and prompt encoding. Extensive experiments and human evaluations demonstrate that our method significantly outperforms SOTA baselines in metaphor consistency, analogy appropriateness, and visual creativity, paving the way for automated high-impact creative applications in advertising and media. Project page with source code and self-contained skills is at https://yuci-gpt.github.io/Beyond-Pixels/.

cs.CV↗

arXiv2026

Estimation of the pseudoscalar glueball mass based on a modified Transformer

A modified Transformer model is introduced for estimating the mass of pseudoscalar glueball in lattice QCD. The model takes as input a sequence of floating-point numbers and produces a two-dimensional vector output. It integrates floating-point embeddings and positional encoding, and is trained using binary cross-entropy loss. The paper provides a detailed description of the model's components and training methods, and compares the performance of the traditional least squares method, the previously used deep neural network, and the modified Transformer in mass estimation. The results show that the modified Transformer model achieves smaller statistical uncertainty in mass estimation than the traditional least squares method. Additionally, compared to the deep neural network, this model utilizes positional encoding and can handle input sequences of varying lengths, offering enhanced adaptability.

hep-lat↗

arXiv2026

Study of the mass of pseudoscalar glueball with a deep neural network

A deep neural network (DNN) is utilized to study the mass of the pseudoscalar glueball in lattice QCD based on Monte Carlo simulations. The DNN is constructed to extract the mass from the negative part of the topological charge density correlation function. The resulting mass estimates are compared with those obtained from conventional least-squares fitting. The DNN gives a pseudoscalar glueball mass of 2558(89)MeV, while the conventional fit gives 2612(112)MeV. The results suggest that the DNN provides a stable and complementary approach to conventional mass extraction from lattice correlation functions.

hep-lat↗

arXiv2026

IMMNet: Hybrid Fusion of Model-based and Data-driven Approaches for Maneuvering Target Tracking

Maneuvering target tracking in three-dimensional space remains a challenging problem due to complex motion dynamics and model mismatch. To address this, this paper proposes a hybrid model/data-driven algorithm named IMMNet, which integrates the interpretable structure of the interacting multiple model (IMM) algorithm with learnable neural components. Unlike end-to-end black-box methods, the proposed IMMNet algorithm not only can preserve the Bayesian inference mechanism that is essential for real-time radar applications, but also can adaptively learn motion patterns and noise characteristics from data. Extensive experiments demonstrate that the proposed IMMNet algorithm consistently outperforms the existing algorithms across various scenarios, validating it as a robust, interpretable, and practical solution for maneuvering target tracking.

cs.RO↗

arXiv2026

iFLYTEK-Embodied-Omni Technical Report

General-purpose embodied agents must understand multimodal instructions, anticipate how their environment will evolve, and produce precise control actions over extended horizons. Existing approaches typically specialize in visual-language reasoning, video-based world modeling, or action generation, while cascaded pipelines that first synthesize future observations and then infer actions can introduce interface bottlenecks and compound prediction errors. We present iFLYTEK-Embodied-Omni, a unified multimodal foundation model that jointly models vision(videos and images), language, and action within a single Omni framework. Its modality-specific visual-language, video-generation, and action-generation components communicate through shared multimodal self-attention. This design establishes brain-cerebellum collaboration: the vision-language modeland video generation model form a high-level brain for instruction understanding, task planning, progress tracking, and future visual-state prediction, whereas the action generation modelserves as a low-level cerebellum that directly converts planned subgoals and shared multimodal context into executable action chunks. To develop these capabilities, we combine action-annotated and action-free embodied videos from human demonstrations and robot interactions with embodied reasoning, embodied perception, and general-purpose image-text data to construct a comprehensive dataset. We further adopt a four-stage strategy that progressively trains the VLM, VGM, and AGM before jointly fine-tuning the complete model.

cs.AI↗

arXiv2026

Nijenhuis BiHom-Lie bialgebras and differential Lie bialgebras

In this paper, we first introduce the concept of Nijenhuis BiHom-Lie algebras. We then establish the equivalence relations between the Manin triples of Nijenhuis BiHom-Lie algebras, Nijenhuis BiHom-Lie bialgebras, and matched pairs of Nijenhuis BiHom-Lie algebras. Furthermore, we show that such an equivalence also holds for differential Lie bialgebras, together with their associated Manin triples and corresponding matched pairs.

math.RA↗

arXiv2026

ERNIE-Image Technical Report

We introduce ERNIE-Image, an open-source text-to-image generation model built upon an 8B single-stream DiT architecture. ERNIE-Image aims to bridge the gap between current open-source models and leading closed-source systems through more effective mining of large-scale pre-training data and improved supervision quality throughout training. During pre-training, we adopt a bottom-up data construction pipeline that combines fine-grained image categorization, rich caption annotation, aesthetic assessment, and hierarchical sampling. This strategy reduces data noise while preserving long-tail concepts and detailed real-world knowledge, providing a stronger foundation for complex generation tasks. In the post-training stage, we use a top-down data construction pipeline for high-demand scenarios, diversify prompt annotations to better match real user inputs, and apply a stabilized DPO strategy to align the model with human aesthetic preferences. We further train ERNIE-Image-Turbo for efficient 8-NFE generation and propose MT-DMD to mitigate capability drift during distillation. To make the model easier to use in practical scenarios, we equip it with a lightweight Prompt Enhancer that expands concise user intents into structured visual descriptions. In addition, we develop ERNIE-Image-Aes, an industrial-grade aesthetic model, together with ERNIE-Image-Aes-1K, a human-annotated benchmark for realistic aesthetic evaluation. Extensive qualitative and quantitative experiments show that ERNIE-Image achieves leading performance among open-source models and approaches top-tier commercial models in instruction following, text rendering, and aesthetic quality. We release the trained models and aesthetic resources to facilitate further academic research and technical progress in the AIGC community.

cs.CV↗

arXiv2026

Joint Communication and Computation Scheduling for MEC-enabled AIGC Services: A Game-Theoretic Stochastic Learning Approach

Artificial Intelligence Generated Content (AIGC) powered by Generative Diffusion Models (GDMs) has emerged as a transformative paradigm for automated content creation. To satisfy the stringent latency requirements of AIGC services in many edge intelligence scenarios (e.g., smart cities), Mobile Edge Computing (MEC) provides critical computational support by deploying GDMs at edge servers (ES) close to end users. This paper investigates an MEC-enabled AIGC network comprising multiple ES, wireless access points (APs), and mobile users (UEs) with heterogeneous latency and accuracy demands. We formulate a Joint Communication Association and Computation Offloading (JCACO) game, where each UE strategically selects its serving AP, ES, and inference steps to minimize the overall service completion time while meeting accuracy constraints. The problem is challenging due to the network dynamics and the incomplete information. We prove that the JCACO game is a potential game under both complete and stochastic information settings, ensuring the existence of Nash Equilibrium (NE) in both cases. To derive the NE efficiently, we develop a distributed Multi-Agent Stochastic Learning (MASL) algorithm that provably converges to the NE with strict performance guarantees. Unlike conventional best-response schemes, MASL requires neither the knowledge of other players' strategies nor global network information, making it fully distributed and adaptive to dynamic environments. We further provide a strict theoretical convergence analysis for MASL by using Ordinary Differential Equations (ODEs). Simulation results demonstrate that MASL significantly reduces service completion time compared with benchmark methods while satisfying accuracy constraints, confirming its effectiveness and practicality for real-world MEC-enabled AIGC networks.

cs.GT↗

arXiv2026

SCORP: Scene-Consistent Multi-agent Diffusion Planning with Stable Online Reinforcement Post-Training for Cooperative Driving

Cooperative driving is a safety- and efficiency-critical task that requires the coordination of diverse, interaction-realistic multi-agent trajectories. Although existing diffusion-based methods can capture multimodal behaviors from demonstrations, they often exhibit weak scene consistency and poor alignment with closed-loop cooperative objectives. This makes post-training necessary for further improvement, yet achieving stable online post-training in reactive multi-agent environments remains challenging. In this paper, we propose SCORP, a scene-consistent multi-agent diffusion planner with stable online reinforcement learning (RL) post-training for cooperative driving. For pre-training, we develop a scene-conditioned multi-agent denoising architecture that couples inter-agent self-attention with a dual-path conditioning mechanism: cross-attention provides direct scene-information injection, while AdaLN-Zero enables additional flexible and stable conditional modulation, thereby improving the scene consistency and road adherence of joint trajectories. For post-training, we formulate a two-layer Markov decision process (MDP) that explicitly integrates the reverse denoising chain with policy-environment interaction. We further co-design dense, well-shaped planning rewards and variance-gated group-relative policy optimization (VG-GRPO) to mitigate advantage collapse and gradient instability during closed-loop training. Extensive experiments show that SCORP outperforms strong open-source baselines on WOMD, with 10.47%-28.26% and 1.70%-7.22% improvements in core safety and efficiency metrics, respectively. Moreover, compared with alternative post-training methods, SCORP delivers significant and consistent gains in both driving safety and traffic efficiency, highlighting stable and sustained advances in closed-loop cooperative driving.

cs.RO↗

arXiv2026

Event-Triggered Distributed Target Tracking via PRIMEX

PRIMEX (prime-based graph encoding and extraction) is a recently proposed framework for scalable distributed fusion. In PRIMEX, the information pedigree of state estimates or probability density functions is encoded using the information codes, enabling lightweight arithmetic for redundancy removal and data integration. Building on PRIMEX and its memoryless fusion strategy based on a least-squares approximation, in this paper we present two efficient distributed tracking algorithms: a consensus-based PRIMEX method that fuses information from all neighbors, and a greedy gossip-based PRIMEX method that fuses with the most informative neighbor. To further increase communication efficiency, we incorporate an event-triggered mechanism, in which transmission decisions are driven by information novelty measured using differences between the information codes. The proposed methods are evaluated and compared with covariance intersection and centralized fusion in a distributed single target tracking scenario. Simulation results show that PRIMEX-based methods remain competitive in tracking accuracy while improving communication efficiency.

eess.SP↗

arXiv2026

SketchFaceGS: Real-Time Sketch-Driven Face Editing and Generation with Gaussian Splatting

3D Gaussian representations have emerged as a powerful paradigm for digital head modeling, achieving photorealistic quality with real-time rendering. However, intuitive and interactive creation or editing of 3D Gaussian head models remains challenging. Although 2D sketches provide an ideal interaction modality for fast, intuitive conceptual design, they are sparse, depth-ambiguous, and lack high-frequency appearance cues, making it difficult to infer dense, geometrically consistent 3D Gaussian structures from strokes - especially under real-time constraints. To address these challenges, we propose SketchFaceGS, the first sketch-driven framework for real-time generation and editing of photorealistic 3D Gaussian head models from 2D sketches. Our method uses a feed-forward, coarse-to-fine architecture. A Transformer-based UV feature-prediction module first reconstructs a coarse but geometrically consistent UV feature map from the input sketch, and then a 3D UV feature enhancement module refines it with high-frequency, photorealistic detail to produce a high-fidelity 3D head. For editing, we introduce a UV Mask Fusion technique combined with a layer-by-layer feature-fusion strategy, enabling precise, real-time, free-viewpoint modifications. Extensive experiments show that SketchFaceGS outperforms existing methods in both generation fidelity and editing flexibility, producing high-quality, editable 3D heads from sketches in a single forward pass.

cs.GR↗

arXiv2026

Averaging pre-Lie bialgebras

In this paper, we first introduce representations of averaging pre-Lie algebras and study their matched pairs, Manin triples, and bialgebra theories. We prove that these three notions are equivalent under certain conditions. Moreover, by introducing averaging operators on quadratic Rota-Baxter pre-Lie algebras, we show that such operators give rise to averaging pre-Lie bialgebras. Then we introduce the notion of admissible classical Yang-Baxter equations in averaging pre-Lie algebras, as well as the relative Rota-Baxter operators on averaging pre-Lie algebras, and show that the relative Rota-Baxter operators on averaging pre-Lie algebras yield symmetric solutions of admissible classical Yang-Baxter equations in averaging pre-Lie algebras. Finally, we show that every averaging pre-Lie bialgebra induces an averaging Lie bialgebra.

math.RA↗