arXiv subjects

Yiwei Hu

Publications and source records attributed to Yiwei Hu.

At least 19 recordsLinked to original sources

arXiv2026

AgiWorld: An Agentic Video World Model for Building Persistent Worlds

Recent interactive video world models generate visually compelling videos from a single image, following user-provided camera motion and text prompts. However, such low-level control demands substantial user effort to realize a high-level intent, and the generated world struggles to remain persistent as the exploration proceeds. In this work, we present AgiWorld, an agentic video world model that explores and builds a persistent world from a single image and a free-form instruction. For high-level planning, our agent AgiWorld-Agent designs the world and the journey through it in a world state and directs the exploration. For low-level realization, our video world model AgiWorld-WM generates the video chunk by chunk and stores it with its geometry in a spatial memory, from which the agent updates its world state. This closed loop keeps the world faithful to the user's intent and keeps the agent's world state consistent with the generated content throughout the exploration. We co-design the two levels around the shared spatial memory and an interface of metric camera trajectories and structured prompts. In AgiWorld-WM, to handle occlusion and accumulated geometric errors in the spatial memory over long rollouts, we propose occlusion-aware history selection, which retrieves past views by surface visibility, and reliability-based memory maintenance, which prefers reliable geometry and earlier observations. AgiWorld-WM achieves state-of-the-art overall performance on world revisiting and exploration benchmarks, and qualitative results and ablations show that AgiWorld builds persistent worlds from free-form instructions over minute-long journeys. Persistence thus arises not from the video model alone but from its closed loop with the agent.

cs.CV↗

arXiv2026

GUMPIM: Unitary and Malleable Memory for Processing-in-Memory with Guaranteed PIM Pages

DRAM-based Processing-in-Memory (PIM) addresses the "memory wall" by executing computations directly inside main memory. However, memory interleaving and virtual memory limit contiguous data size visible to PIM units, constraining PIM task granularity. To mitigate this, existing PIM systems drastically isolate PIM memory or disable memory interleaving. These design choices, however, decrease the CPU memory bandwidth and introduce extra data transfer, leading to an additional "system memory wall" that must be resolved to realize PIM's full potential. In this work, we propose GUMPIM, a PIM system that allows interleaved CPU pages and non-interleaved PIM pages to coexist in a Unitary and Malleable memory space with Guaranteed PIM page allocation. GUMPIM enables zero-copy during PIM task offloading and maintains CPU memory bandwidth while ensuring low PIM offloading overhead. First, we propose a dual-track memory management mechanism consisting of independent page allocation and address translation for CPU and PIM pages. Second, we design GUMPIM interface hardware on PIM-enabled DRAMs to provide a dynamic address mapping for the different data layouts of CPU and PIM pages. Third, we propose a PIM-assisted page migration mechanism that transparently migrates pages while preserving CPU access bandwidth, thereby enabling guaranteed and accelerated PIM page allocation. GUMPIM requires no changes to commodity DRAM standards; all hardware modifications are limited to the DRAM side, ensuring full compatibility with existing CPUs and enabling immediate deployment on current HBMx- and LPDDRx-based PIM platforms. Our results show only <0.1% performance degradation for CPU workloads on GUMPIM. For PIM workloads, GUMPIM reduces memory allocation and CPU-part computation times by 2.7x and 4.93x, respectively, yielding an end-to-end 2.3x speedup over a state-of-the-art baseline system.

cs.AR↗

arXiv2026

RealMat: Realistic Materials with Diffusion and Reinforcement Learning

Generative models for high-quality materials are particularly desirable to make 3D content authoring more accessible. However, the majority of material generation methods are trained on synthetic data. Synthetic data provides precise supervision for material maps, which is convenient but also tends to create a significant visual gap with real-world materials. Alternatively, recent work used a small dataset of real flash photographs to guarantee realism, however such data is limited in scale and diversity. To address these limitations, we propose RealMat, a diffusion-based material generator that leverages realistic priors, including a text-to-image model and a dataset of realistic material photos under natural lighting. In RealMat, we first finetune a pretrained Stable Diffusion XL (SDXL) with synthetic material maps arranged in $2 \times 2$ grids. This way, our model inherits some realism of SDXL while learning the data distribution of the synthetic material grids. Still, this creates a realism gap, with some generated materials appearing synthetic. We propose to further finetune our model through reinforcement learning (RL), encouraging the generation of realistic materials. We develop a realism reward function for any material image under natural lighting, by collecting a large-scale dataset of realistic material images. We show that this approach increases generated materials' realism compared to our base model and related work.

cs.GR↗

arXiv2026

HiMat: DiT-based Ultra-High Resolution SVBRDF Generation

Creating ultra-high-resolution spatially varying bidirectional reflectance functions (SVBRDFs) is critical for photorealistic 3D content creation, to faithfully represent fine-scale surface details required for close-up rendering. However, achieving 4K generation faces two key challenges: (1) the need to synthesize multiple reflectance maps at full resolution, which multiplies the pixel budget and imposes prohibitive memory and computational cost, and (2) the requirement to maintain strong pixel-level alignment across maps at 4K, which is particularly difficult when adapting pretrained models designed for the RGB image domain. We introduce HiMat, a diffusion-based framework tailored for efficient and diverse 4K SVBRDF generation. To address the first challenge, HiMat performs generation in a high-compression latent space via DC-AE, and employs a pretrained diffusion transformer with linear attention to improve per-map efficiency. To address the second challenge, we propose CrossStitch, a lightweight convolutional module that enforces cross-map consistency without incurring the cost of global attention. Our experiments show that HiMat achieves high-fidelity 4K SVBRDF generation with superior efficiency, structural consistency, and diversity compared to prior methods. Beyond materials, our framework also generalizes to related applications such as intrinsic decomposition.

cs.CV↗

arXiv2026

OmniRoam: World Wandering via Long-Horizon Panoramic Video Generation

Modeling scenes using video generation models has garnered growing research interest in recent years. However, most existing approaches rely on perspective video models that synthesize only limited observations of a scene, leading to issues of completeness and global consistency. We propose OmniRoam, a controllable panoramic video generation framework that exploits the rich per-frame scene coverage and inherent long-term spatial and temporal consistency of panoramic representation, enabling long-horizon scene wandering. Our framework begins with a preview stage, where a trajectory-controlled video generation model creates a quick overview of the scene from a given input image or video. Then, in the refine stage, this video is temporally extended and spatially upsampled to produce long-range, high-resolution videos, thus enabling high-fidelity world wandering. To train our model, we introduce two panoramic video datasets that incorporate both synthetic and real-world captured videos. Experiments show that our framework consistently outperforms state-of-the-art methods in terms of visual quality, controllability, and long-term scene consistency, both qualitatively and quantitatively. We further showcase several extensions of this framework, including real-time video generation and 3D reconstruction. Code is available at https://github.com/yuhengliu02/OmniRoam.

cs.CV↗

arXiv2026

SDD-YOLO: A Small-Target Detection Framework for Ground-to-Air Anti-UAV Surveillance with Edge-Efficient Deployment

Detecting small unmanned aerial vehicles (UAVs) from a ground-to-air (G2A) perspective presents significant challenges, including extremely low pixel occupancy, cluttered aerial backgrounds, and strict real-time constraints. Existing YOLO-based detectors are primarily optimized for general object detection and often lack adequate feature resolution for sub-pixel targets, while introducing complexities during deployment. In this paper, we propose SDD-YOLO, a small-target detection framework tailored for G2A anti-UAV surveillance. To capture fine-grained spatial details critical for micro-targets, SDD-YOLO introduces a P2 high-resolution detection head operating at 4 times downsampling. Furthermore, we integrate the recent architectural advancements from YOLO26, including a DFL-free, NMS-free architecture for streamlined inference, and the MuSGD hybrid training strategy with ProgLoss and STAL, which substantially mitigates gradient oscillation on sparse small-target signals. To support our evaluation, we construct DroneSOD-30K, a large-scale G2A dataset comprising approximately 30,000 annotated images covering diverse meteorological conditions. Experiments demonstrate that SDD-YOLO-n achieves a mAP@0.5 of 86.0% on DroneSOD-30K, surpassing the YOLOv5n baseline by 7.8 percentage points. Extensive inference analysis shows our model attains 226 FPS on an NVIDIA RTX 5090 and 35 FPS on an Intel Xeon CPU, demonstrating exceptional efficiency for future edge deployment.

cs.CV↗

arXiv2026

tttLRM: Test-Time Training for Long Context and Autoregressive 3D Reconstruction

We propose tttLRM, a novel large 3D reconstruction model that leverages a Test-Time Training (TTT) layer to enable long-context, autoregressive 3D reconstruction with linear computational complexity, further scaling the model's capability. Our framework efficiently compresses multiple image observations into the fast weights of the TTT layer, forming an implicit 3D representation in the latent space that can be decoded into various explicit formats, such as Gaussian Splats (GS) for downstream applications. The online learning variant of our model supports progressive 3D reconstruction and refinement from streaming observations. We demonstrate that pretraining on novel view synthesis tasks effectively transfers to explicit 3D modeling, resulting in improved reconstruction quality and faster convergence. Extensive experiments show that our method achieves superior performance in feedforward 3D Gaussian reconstruction compared to state-of-the-art approaches on both objects and scenes.

cs.CV↗

arXiv2025

OmniVCus: Feedforward Subject-driven Video Customization with Multimodal Control Conditions

Existing feedforward subject-driven video customization methods mainly study single-subject scenarios due to the difficulty of constructing multi-subject training data pairs. Another challenging problem that how to use the signals such as depth, mask, camera, and text prompts to control and edit the subject in the customized video is still less explored. In this paper, we first propose a data construction pipeline, VideoCus-Factory, to produce training data pairs for multi-subject customization from raw videos without labels and control signals such as depth-to-video and mask-to-video pairs. Based on our constructed data, we develop an Image-Video Transfer Mixed (IVTM) training with image editing data to enable instructive editing for the subject in the customized video. Then we propose a diffusion Transformer framework, OmniVCus, with two embedding mechanisms, Lottery Embedding (LE) and Temporally Aligned Embedding (TAE). LE enables inference with more subjects by using the training subjects to activate more frame embeddings. TAE encourages the generation process to extract guidance from temporally aligned control signals by assigning the same frame embeddings to the control and noise tokens. Experiments demonstrate that our method significantly surpasses state-of-the-art methods in both quantitative and qualitative evaluations. Video demos are at our project page: https://caiyuanhao1998.github.io/project/OmniVCus/. Our code, models, data are released at https://github.com/caiyuanhao1998/Open-OmniVCus

cs.CV↗

arXiv2025

V-RGBX: Video Editing with Accurate Controls over Intrinsic Properties

Large-scale video generation models have shown remarkable potential in modeling photorealistic appearance and lighting interactions in real-world scenes. However, a closed-loop framework that jointly understands intrinsic scene properties (e.g., albedo, normal, material, and irradiance), leverages them for video synthesis, and supports editable intrinsic representations remains unexplored. We present V-RGBX, the first end-to-end framework for intrinsic-aware video editing. V-RGBX unifies three key capabilities: (1) video inverse rendering into intrinsic channels, (2) photorealistic video synthesis from these intrinsic representations, and (3) keyframe-based video editing conditioned on intrinsic channels. At the core of V-RGBX is an interleaved conditioning mechanism that enables intuitive, physically grounded video editing through user-selected keyframes, supporting flexible manipulation of any intrinsic modality. Extensive qualitative and quantitative results show that V-RGBX produces temporally consistent, photorealistic videos while propagating keyframe edits across sequences in a physically plausible manner. We demonstrate its effectiveness in diverse applications, including object appearance editing and scene-level relighting, surpassing the performance of prior methods.

cs.CV↗

arXiv2025

Generating 360° Video is What You Need For a 3D Scene

Generating 3D scenes is still a challenging task due to the lack of readily available scene data. Most existing methods only produce partial scenes and provide limited navigational freedom. We introduce a practical and scalable solution that uses 360° video as an intermediate scene representation, capturing the full-scene context and ensuring consistent visual content throughout the generation. We propose WorldPrompter, a generative pipeline that synthesizes traversable 3D scenes from text prompts. WorldPrompter incorporates a conditional 360° panoramic video generator, capable of producing a 128-frame video that simulates a person walking through and capturing a virtual environment. The resulting video is then reconstructed as Gaussian splats by a fast feedforward 3D reconstructor, enabling a true walkable experience within the 3D scene. Experiments demonstrate that our panoramic video generation model, trained with a mix of image and video data, achieves convincing spatial and temporal consistency for static scenes. This is validated by an average COLMAP matching rate of 94.6\%, allowing for high-quality panoramic Gaussian splat reconstruction and improved navigation throughout the scene. Qualitative and quantitative results also show it outperforms the state-of-the-art 360° video generators and 3D scene generation models.

cs.GR↗

arXiv2025

MaterialPicker: Multi-Modal DiT-Based Material Generation

High-quality material generation is key for virtual environment authoring and inverse rendering. We propose MaterialPicker, a multi-modal material generator leveraging a Diffusion Transformer (DiT) architecture, improving and simplifying the creation of high-quality materials from text prompts and/or photographs. Our method can generate a material based on an image crop of a material sample, even if the captured surface is distorted, viewed at an angle or partially occluded, as is often the case in photographs of natural scenes. We further allow the user to specify a text prompt to provide additional guidance for the generation. We finetune a pre-trained DiT-based video generator into a material generator, where each material map is treated as a frame in a video sequence. We evaluate our approach both quantitatively and qualitatively and show that it enables more diverse material generation and better distortion correction than previous work.

cs.CV↗

arXiv2025

SoK: Automated Vulnerability Repair: Methods, Tools, and Assessments

The increasing complexity of software has led to the steady growth of vulnerabilities. Vulnerability repair investigates how to fix software vulnerabilities. Manual vulnerability repair is labor-intensive and time-consuming because it relies on human experts, highlighting the importance of Automated Vulnerability Repair (AVR). In this SoK, we present the systematization of AVR methods through the three steps of AVR workflow: vulnerability analysis, patch generation, and patch validation. We assess AVR tools for C/C++ and Java programs as they have been widely studied by the community. Since existing AVR tools for C/C++ programs are evaluated with different datasets, which often consist of a few vulnerabilities, we construct the first C/C++ vulnerability repair benchmark dataset, dubbed Vul4C, which contains 144 vulnerabilities as well as their exploits and patches. We use Vul4C to evaluate seven AVR tools for C/C++ programs and use the third-party Vul4J dataset to evaluate two AVR tools for Java programs. We also discuss future research directions.

cs.SE↗

arXiv2025

MILLION: Mastering Long-Context LLM Inference Via Outlier-Immunized KV Product Quantization

Large language models (LLMs) are increasingly utilized for complex tasks requiring longer context lengths, with some models supporting up to 128K or 1M tokens. This trend, however, presents significant challenges in inference speed and memory management. Quantization emerges as a promising approach to address the widening gap between LLM size and memory capacity. However, traditional quantization schemes often yield suboptimal compression results for KV caches due to two key factors: i) On-the-fly quantization and de-quantization, causing significant performance overhead; ii) Prevalence of outliers in KV values, challenging low-bitwidth uniform quantization. To this end, we propose MILLION, a novel quantization framework achieving low-bitwidth KV cache through product quantization. First, we conduct a thorough analysis of KV cache distribution, revealing the limitations of existing quantization schemes. Second, we introduce a non-uniform quantization algorithm based on product quantization, which efficiently compresses data while preserving accuracy. Third, we develop a high-performance GPU inference framework with efficient attention kernel and pipeline design for MILLION that leverages sparse computation and asynchronous quantization, significantly enhancing inference speed. Comprehensive evaluation results demonstrate that MILLION can achieve 4 bits quantization with trivial perplexity and accuracy loss, and achieve 2.09x end-to-end performance gains at 32K context length. Code is released at https://github.com/ZongwuWang/MILLION.

cs.DC↗

arXiv2024

Turbo3D: Ultra-fast Text-to-3D Generation

We present Turbo3D, an ultra-fast text-to-3D system capable of generating high-quality Gaussian splatting assets in under one second. Turbo3D employs a rapid 4-step, 4-view diffusion generator and an efficient feed-forward Gaussian reconstructor, both operating in latent space. The 4-step, 4-view generator is a student model distilled through a novel Dual-Teacher approach, which encourages the student to learn view consistency from a multi-view teacher and photo-realism from a single-view teacher. By shifting the Gaussian reconstructor's inputs from pixel space to latent space, we eliminate the extra image decoding time and halve the transformer sequence length for maximum efficiency. Our method demonstrates superior 3D generation results compared to previous baselines, while operating in a fraction of their runtime.

cs.CV↗

arXiv2024

Buffer Anytime: Zero-Shot Video Depth and Normal from Image Priors

We present Buffer Anytime, a framework for estimation of depth and normal maps (which we call geometric buffers) from video that eliminates the need for paired video--depth and video--normal training data. Instead of relying on large-scale annotated video datasets, we demonstrate high-quality video buffer estimation by leveraging single-image priors with temporal consistency constraints. Our zero-shot training strategy combines state-of-the-art image estimation models based on optical flow smoothness through a hybrid loss function, implemented via a lightweight temporal attention architecture. Applied to leading image models like Depth Anything V2 and Marigold-E2E-FT, our approach significantly improves temporal consistency while maintaining accuracy. Experiments show that our method not only outperforms image-based approaches but also achieves results comparable to state-of-the-art video models trained on large-scale paired video datasets, despite using no such paired video data.

cs.CV↗

arXiv2024

RelitLRM: Generative Relightable Radiance for Large Reconstruction Models

We propose RelitLRM, a Large Reconstruction Model (LRM) for generating high-quality Gaussian splatting representations of 3D objects under novel illuminations from sparse (4-8) posed images captured under unknown static lighting. Unlike prior inverse rendering methods requiring dense captures and slow optimization, often causing artifacts like incorrect highlights or shadow baking, RelitLRM adopts a feed-forward transformer-based model with a novel combination of a geometry reconstructor and a relightable appearance generator based on diffusion. The model is trained end-to-end on synthetic multi-view renderings of objects under varying known illuminations. This architecture design enables to effectively decompose geometry and appearance, resolve the ambiguity between material and lighting, and capture the multi-modal distribution of shadows and specularity in the relit appearance. We show our sparse-view feed-forward RelitLRM offers competitive relighting results to state-of-the-art dense-view optimization-based baselines while being significantly faster. Our project page is available at: https://relit-lrm.github.io/.

cs.CV↗

arXiv2024

TexSliders: Diffusion-Based Texture Editing in CLIP Space

Generative models have enabled intuitive image creation and manipulation using natural language. In particular, diffusion models have recently shown remarkable results for natural image editing. In this work, we propose to apply diffusion techniques to edit textures, a specific class of images that are an essential part of 3D content creation pipelines. We analyze existing editing methods and show that they are not directly applicable to textures, since their common underlying approach, manipulating attention maps, is unsuitable for the texture domain. To address this, we propose a novel approach that instead manipulates CLIP image embeddings to condition the diffusion generation. We define editing directions using simple text prompts (e.g., "aged wood" to "new wood") and map these to CLIP image embedding space using a texture prior, with a sampling-based approach that gives us identity-preserving directions in CLIP space. To further improve identity preservation, we project these directions to a CLIP subspace that minimizes identity variations resulting from entangled texture attributes. Our editing pipeline facilitates the creation of arbitrary sliders using natural language prompts only, with no ground-truth annotated data necessary.

cs.GR↗

arXiv2024

RGB$\leftrightarrow$X: Image decomposition and synthesis using material- and lighting-aware diffusion models

The three areas of realistic forward rendering, per-pixel inverse rendering, and generative image synthesis may seem like separate and unrelated sub-fields of graphics and vision. However, recent work has demonstrated improved estimation of per-pixel intrinsic channels (albedo, roughness, metallicity) based on a diffusion architecture; we call this the RGB$\rightarrow$X problem. We further show that the reverse problem of synthesizing realistic images given intrinsic channels, X$\rightarrow$RGB, can also be addressed in a diffusion framework. Focusing on the image domain of interior scenes, we introduce an improved diffusion model for RGB$\rightarrow$X, which also estimates lighting, as well as the first diffusion X$\rightarrow$RGB model capable of synthesizing realistic images from (full or partial) intrinsic channels. Our X$\rightarrow$RGB model explores a middle ground between traditional rendering and generative models: we can specify only certain appearance properties that should be followed, and give freedom to the model to hallucinate a plausible version of the rest. This flexibility makes it possible to use a mix of heterogeneous training datasets, which differ in the available channels. We use multiple existing datasets and extend them with our own synthetic and real data, resulting in a model capable of extracting scene properties better than previous work and of generating highly realistic images of interior scenes.

cs.CV↗