arXiv subjects

Zhensong Zhang

Publications and source records attributed to Zhensong Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Caption-once, Frames-on-Demand: Visual-Need Routing for Budget-Aware Agentic Long Video Understanding

Long-video understanding on edge devices must reason over hours of content under tight compute and bandwidth budgets. Subsampling visual tokens loses temporal structure, while text-only video memories lose fine-grained visual attributes. We observe a visual-textual duality: language memories carry long-range temporal structure better than dense frames, while pixels remain decisive for attribute-level perception. Building on this insight, we propose Caption-once, Frames-onDemand (CFD), a budget-aware edge-cloud agentic framework. The edge runs a single offline captioning pass that builds a dual-track narrative index, an event-level story skeleton plus a clip-level micro-log, cached and reused across queries without re-captioning. At query time, a cloud-side MLLM reasons over the index in a story-first loop centered on a lightweight Visual-Need Router: a per-query gating module that triggers bounded keyframe retrieval only for perceptual questions (appearance, on-screen text, attribute disambiguation) and keeps temporal-structural questions in language space. The router turns visual access into a first-class, query-conditioned cost, capping per-query frame consumption regardless of video length. Experiments on long-video benchmarks demonstrate strong accuracy-efficiency trade-offs while substantially reducing online visual processing.

cs.CV↗

arXiv2026

Video Generative Models as Geometry Learner

Recent generative approaches to geometry estimation adapt pretrained image diffusion models and treat the task as image-conditioned generation. Leveraging off-the-shelf image diffusion models, they either (i) train task-specific geometry models (for depth and surface normal estimation) independently, losing the opportunity of exploring the intrinsic correlation of these geometric targets, or (ii) jointly fine-tune modified image diffusion backbones (e.g., altered self-attention), which typically demands substantial labeled data. To overcome these limitations in a principled fashion, we repurpose pretrained video generative models as a unified and data-efficient framework for geometry estimation, formulated innovatively as a next-frames prediction task. Our method, GeoNeXt, inherits naturally structured knowledge and richer priors from the video model, while further adapting them for joint modeling of images and geometry targets (image <-> geometry), enabling more data efficient and effective learning of geometry. Extensive experiments validate our method for zero-shot monocular depth and surface normal estimation across diverse datasets, outperforming both previous task-specific and unified generative competitors while using substantially less training data. Notably, our method rivals discriminative state-of-the-art approaches trained on over 100x more data and even standouts on several benchmarks.

cs.CV↗

arXiv2026

LatSearch: Latent Reward-Guided Search for Faster Inference-Time Scaling in Video Diffusion

The recent success of inference-time scaling in large language models has inspired similar explorations in video diffusion. In particular, motivated by the existence of "golden noise" that enhances video quality, prior work has attempted to improve inference by optimising or searching for better initial noise. However, these approaches have notable limitations: they either rely on priors imposed at the beginning of noise sampling or on rewards evaluated only on the denoised and decoded videos. This leads to error accumulation, delayed and sparse reward signals, and prohibitive computational cost, which prevents the use of stronger search algorithms. Crucially, stronger search algorithms are precisely what could unlock substantial gains in controllability, sample efficiency and generation quality for video diffusion, provided their computational cost can be reduced. To fill in this gap, we enable efficient inference-time scaling for video diffusion through latent reward guidance, which provides intermediate, informative and efficient feedback along the denoising trajectory. We introduce a latent reward model that scores partially denoised latents at arbitrary timesteps with respect to visual quality, motion quality, and text alignment. Building on this model, we propose LatSearch, a novel inference-time search mechanism that performs Reward-Guided Resampling and Pruning (RGRP). In the resampling stage, candidates are sampled according to reward-normalised probabilities to reduce over-reliance on the reward model. In the pruning stage, applied at the final scheduled step, only the candidate with the highest cumulative reward is retained, improving both quality and efficiency. We evaluate LatSearch on the VBench-2.0 benchmark and demonstrate that it consistently improves video generation across multiple evaluation dimensions compared to the baseline Wan2.1 model.

cs.CV↗

arXiv2026

Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation

Maintaining physical consistency in video generators and world models increasingly relies on vision-language models (VLMs) as automated judges that provide reward signals, ranking decisions, and data-filtering criteria. Yet VLMs differ substantially in training data and architecture, encoding physical phenomena through distinct internal representations. A single global evaluation schema therefore gives every VLM the same axes of competence, regardless of what each can actually perceive. We propose JudgeFit, an iterative refinement procedure that discovers a per-VLM evaluation taxonomy. An initial taxonomy is constructed by prompting the target VLM to enumerate physics errors on a small set of videos and clustering the resulting descriptions. The taxonomy is then refined through a diagnostic step: we calibrate the VLM's per-dimension scores to human physical-commonsense ratings, diagnose which dimensions it scores unreliably or redundantly, and prompt an LLM to repair them, iterating until convergence. We further instantiate this procedure as a benchmark and apply it to 16 VLMs spanning eight model families. The refined taxonomy outperforms the global-schema baseline on held-out videos for every VLM tested, with a mean relative improvement of approximately 32%. Beyond aggregate accuracy, the per-VLM profiles expose model-specific blind spots that overall rankings cannot anticipate, with reliability patterns differing markedly across model families.

cs.CV↗

arXiv2026

LiteVSR: Lightweight Adaptation of Frozen Diffusion Transformers for Video Super-Resolution

Adapting large-scale pre-trained video generators for Video Super-Resolution (VSR) in novel domains remains computationally prohibitive. Methods that reformulate generation as direct Low-Quality to High-Quality mappings deviate from the original generative formulation, demanding extensive fine-tuning. ControlNet-style adapters lose their efficiency under modern Diffusion Transformers since the absence of encoder-decoder hierarchy forces duplication of the entire backbone. We observe that flow matching offers a principled alternative for cross-domain VSR adaptation. By predicting a constant velocity field across all timesteps, the adaptation task reduces to learning a fixed injection pattern rather than time-varying transformations. Building on this insight, we propose LiteVSR, a minimalist framework that performs VSR using a completely frozen Diffusion Transformer with a lightweight State-Aware Adapter. The adapter employs a dual-stream architecture that extracts static structural cues from the LQ input and dynamic cues from intermediate denoising states, aligning them through time-dependent cross-attention to enable adaptive transition from structural alignment to texture refinement as denoising proceeds. LiteVSR achieves competitive restoration quality with only 11.25% trainable parameters and 12 GPU-hours of training on a single A100, while maintaining fast sampling (down to a single step) compatibility.

cs.CV↗

arXiv2026

GRVS: a Generalizable and Recurrent Approach to Monocular Dynamic View Synthesis

Synthesizing novel views from monocular videos of dynamic scenes remains a challenging problem. Scene-specific methods that optimize 4D representations with explicit motion priors often break down in highly dynamic regions where multi-view information is hard to exploit. Diffusion-based approaches that integrate camera control into large pre-trained models can produce visually plausible videos but frequently suffer from geometric inconsistencies across both static and dynamic areas. Both families of methods also require substantial computational resources. Building on the success of generalizable models for static novel view synthesis, we adapt the framework to dynamic inputs and propose a new model with two key components: (1) a recurrent loop that enables unbounded and asynchronous mapping between input and target videos and (2) an efficient use of plane sweeps over dynamic inputs to disentangle camera and scene motion, and achieve fine-grained, six-degrees-of-freedom camera controls. We train and evaluate our model on the UCSD dataset and on Kubric-4D-dyn, a new monocular dynamic dataset featuring longer, higher resolution sequences with more complex scene dynamics than existing alternatives. Our model outperforms four Gaussian Splatting-based scene-specific approaches, as well as two diffusion-based approaches in reconstructing fine-grained geometric details across both static and dynamic regions.

cs.CV↗

arXiv2026

Off The Grid: Detection of Primitives for Feed-Forward 3D Gaussian Splatting

Feed-forward 3D Gaussian Splatting (3DGS) models enable real-time scene generation but are hindered by suboptimal pixel-aligned primitive placement, which relies on a dense, rigid grid that limits both quality and efficiency. We introduce a new feed-forward architecture that detects 3D Gaussian primitives at a sub-pixel level, replacing the pixel grid with an adaptive, ``Off-The-Grid" distribution. Inspired by keypoint detection, our decoder learns to locally distribute primitives across image patches. We also provide an Adaptive Density mechanism by assigning varying number of primitives per patch based on Shannon entropy. We combine the proposed decoder with a pre-trained 3D reconstruction backbone and train them end-to-end using photometric supervision without any 3D annotation. The resulting pose-free model generates photorealistic 3DGS scenes in seconds, achieving state-of-the-art novel view synthesis for feed-forward models. It outperforms competitors while using far fewer primitives, demonstrating a more accurate and efficient allocation that captures fine details and reduces artifacts. Project page: https://arthurmoreau.github.io/OffTheGrid/.

cs.CV↗

arXiv2026

Diffusion-Based Makeup Transfer with Facial Region-Aware Makeup Features

Current diffusion-based makeup transfer methods commonly use the makeup information encoded by off-the-shelf foundation models (e.g., CLIP) as condition to preserve the makeup style of reference image in the generation. Although effective, these works mainly have two limitations: (1) foundation models pre-trained for generic tasks struggle to capture makeup styles; (2) the makeup features of reference image are injected to the diffusion denoising model as a whole for global makeup transfer, overlooking the facial region-aware makeup features (i.e., eyes, mouth, etc) and limiting the regional controllability for region-specific makeup transfer. To address these, in this work, we propose Facial Region-Aware Makeup features (FRAM), which has two stages: (1) makeup CLIP fine-tuning; (2) identity and facial region-aware makeup injection. For makeup CLIP fine-tuning, unlike prior works using off-the-shelf CLIP, we synthesize annotated makeup style data using GPT-o3 and text-driven image editing model, and then use the data to train a makeup CLIP encoder through self-supervised and image-text contrastive learning. For identity and facial region-aware makeup injection, we construct before-and-after makeup image pairs from the edited images in stage 1 and then use them to learn to inject identity of source image and makeup of reference image to the diffusion denoising model for makeup transfer. Specifically, we use learnable tokens to query the makeup CLIP encoder to extract facial region-aware makeup features for makeup injection, which is learned via an attention loss to enable regional control. As for identity injection, we use a ControlNet Union to encode source image and its 3D mesh simultaneously. The experimental results verify the superiority of our regional controllability and our makeup transfer performance. Code is available at https://github.com/zaczgao/Facial_Region-Aware_Makeup.

cs.CV↗

arXiv2026

Color When It Counts: Grayscale-Guided Online Triggering for Always-On Streaming Video Sensing

Always-on sensing is essential for next-generation edge/wearable AI systems, yet continuous high-fidelity RGB video capture remains prohibitively expensive for resource-constrained mobile and edge platforms. We present a new paradigm for efficient streaming video understanding: grayscale-always, color-on-demand. Through preliminary studies, we discover that color is not always necessary. Sparse RGB frames suffice for comparable performance when temporal structure is preserved via continuous grayscale streams. Building on this insight, we propose ColorTrigger, an online training-free trigger that selectively activates color capture based on windowed grayscale affinity analysis. Designed for real-time edge deployment, ColorTrigger uses lightweight quadratic programming to detect chromatic redundancy causally, coupled with credit-budgeted control and dynamic token routing to jointly reduce sensing and inference costs. On streaming video understanding benchmarks, ColorTrigger achieves 91.6% of full-color baseline performance while using only 8.1% RGB frames, demonstrating substantial color redundancy in natural videos and enabling practical always-on video sensing on resource-constrained devices.

cs.CV↗

arXiv2026

Egocentric Co-Pilot: Web-Native Smart-Glasses Agents for Assistive Egocentric AI

What if accessing the web did not require a screen, a stable desk, or even free hands? For people navigating crowded cities, living with low vision, or experiencing cognitive overload, smart glasses coupled with AI agents could turn the web into an always-on assistive layer over daily life. We present Egocentric Co-Pilot, a web-native neuro-symbolic framework that runs on smart glasses and uses a Large Language Model (LLM) to orchestrate a toolbox of perception, reasoning, and web tools. An egocentric reasoning core combines Temporal Chain-of-Thought with Hierarchical Context Compression to support long-horizon question answering and decision support over continuous first-person video, far beyond a single model's context window. Additionally, a lightweight multimodal intent layer maps noisy speech and gaze into structured commands. We further implement and evaluate a cloud-native WebRTC pipeline integrating streaming speech, video, and control messages into a unified channel for smart glasses and browsers. In parallel, we deploy an on-premise WebSocket baseline, exposing concrete trade-offs between local inference and cloud offloading in terms of latency, mobility, and resource use. Experiments on Egolife and HD-EPIC demonstrate competitive or state-of-the-art egocentric QA performance, and a human-in-the-loop study on smart glasses shows higher task completion and user satisfaction than leading commercial baselines. Taken together, these results indicate that web-connected egocentric co-pilots can be a practical path toward more accessible, context-aware assistance in everyday life. By grounding operation in web-native communication primitives and modular, auditable tool use, Egocentric Co-Pilot offers a concrete blueprint for assistive, always-on web agents that support education, accessibility, and social inclusion for people who may benefit most from contextual, egocentric AI.

cs.HC↗

arXiv2026

ICo3D: An Interactive Conversational 3D Virtual Human

This work presents Interactive Conversational 3D Virtual Human (ICo3D), a method for generating an interactive, conversational, and photorealistic 3D human avatar. Based on multi-view captures of a subject, we create an animatable 3D face model and a dynamic 3D body model, both rendered by splatting Gaussian primitives. Once merged together, they represent a lifelike virtual human avatar suitable for real-time user interactions. We equip our avatar with an LLM for conversational ability. During conversation, the audio speech of the avatar is used as a driving signal to animate the face model, enabling precise synchronization. We describe improvements to our dynamic Gaussian models that enhance photorealism: SWinGS++ for body reconstruction and HeadGaS++ for face reconstruction, and provide as well a solution to merge the separate face and body models without artifacts. We also present a demo of the complete system, showcasing several use cases of real-time conversation with the 3D avatar. Our approach offers a fully integrated virtual avatar experience, supporting both oral and written form interactions in immersive environments. ICo3D is applicable to a wide range of fields, including gaming, virtual assistance, and personalized education, among others. Project page: https://ico3d.github.io/

cs.CV↗

arXiv2026

Map2Thought: Explicit 3D Spatial Reasoning via Metric Cognitive Maps

We propose Map2Thought, a framework that enables explicit and interpretable spatial reasoning for 3D VLMs. The framework is grounded in two key components: Metric Cognitive Map (Metric-CogMap) and Cognitive Chain-of-Thought (Cog-CoT). Metric-CogMap provides a unified spatial representation by integrating a discrete grid for relational reasoning with a continuous, metric-scale representation for precise geometric understanding. Building upon the Metric-CogMap, Cog-CoT performs explicit geometric reasoning through deterministic operations, including vector operations, bounding-box distances, and occlusion-aware appearance order cues, producing interpretable inference traces grounded in 3D structure. Experimental results show that Map2Thought enables explainable 3D understanding, achieving 59.9% accuracy using only half the supervision, closely matching the 60.9% baseline trained with the full dataset. It consistently outperforms state-of-the-art methods by 5.3%, 4.8%, and 4.0% under 10%, 25%, and 50% training subsets, respectively, on the VSI-Bench.

cs.CV↗

arXiv2026

Optimizing Multimodal LLMs for Egocentric Video Understanding: A Solution for the HD-EPIC VQA Challenge

Multimodal Large Language Models (MLLMs) struggle with complex video QA benchmarks like HD-EPIC VQA due to ambiguous queries/options, poor long-range temporal reasoning, and non-standardized outputs. We propose a framework integrating query/choice pre-processing, domain-specific Qwen2.5-VL fine-tuning, a novel Temporal Chain-of-Thought (T-CoT) prompting for multi-step reasoning, and robust post-processing. This system achieves 41.6% accuracy on HD-EPIC VQA, highlighting the need for holistic pipeline optimization in demanding video understanding. Our code, fine-tuned models are available at https://github.com/YoungSeng/Egocentric-Co-Pilot.

cs.CV↗

arXiv2025

GASPACHO: Gaussian Splatting for Controllable Humans and Objects

We present GASPACHO, a method for generating photorealistic, controllable renderings of human-object interactions from multi-view RGB video. Unlike prior work that reconstructs only the human and treats objects as background, GASPACHO simultaneously recovers animatable templates for both the human and the interacting object as distinct sets of Gaussians, thereby allowing for controllable renderings of novel human object interactions in different poses from novel-camera viewpoints. We introduce a novel formulation that learns object Gaussians on an underlying 2D surface manifold rather than in 3D volume, yielding sharper, fine-grained object details for dynamic object reconstruction. We further propose a contact constraint in Gaussian space that regularizes human-object relations and enables natural, physically plausible animation. Across three benchmarks - BEHAVE, NeuralDome, and DNA-Rendering - GASPACHO achieves high-quality reconstructions under heavy occlusion and supports controllable synthesis of novel human-object interactions. We also demonstrate that our method allows for composition of humans and objects in 3D scenes and for the first time showcase that neural rendering can be used for the controllable generation of photoreal humans interacting with dynamic objects in diverse scenes. Our results are available at: https://miraymen.github.io/gaspacho/

cs.CV↗

arXiv2025

Charge: A Comprehensive Novel View Synthesis Benchmark and Dataset to Bind Them All

This paper presents a new dataset for Novel View Synthesis, generated from a high-quality, animated film with stunning realism and intricate detail. Our dataset captures a variety of dynamic scenes, complete with detailed textures, lighting, and motion, making it ideal for training and evaluating cutting-edge 4D scene reconstruction and novel view generation models. In addition to high-fidelity RGB images, we provide multiple complementary modalities, including depth, surface normals, object segmentation and optical flow, enabling a deeper understanding of scene geometry and motion. The dataset is organised into three distinct benchmarking scenarios: a dense multi-view camera setup, a sparse camera arrangement, and monocular video sequences, enabling a wide range of experimentation and comparison across varying levels of data sparsity. With its combination of visual richness, high-quality annotations, and diverse experimental setups, this dataset offers a unique resource for pushing the boundaries of view synthesis and 3D vision.

cs.CV↗

arXiv2025

Plug-and-Play Clarifier: A Zero-Shot Multimodal Framework for Egocentric Intent Disambiguation

The performance of egocentric AI agents is fundamentally limited by multimodal intent ambiguity. This challenge arises from a combination of underspecified language, imperfect visual data, and deictic gestures, which frequently leads to task failure. Existing monolithic Vision-Language Models (VLMs) struggle to resolve these multimodal ambiguous inputs, often failing silently or hallucinating responses. To address these ambiguities, we introduce the Plug-and-Play Clarifier, a zero-shot and modular framework that decomposes the problem into discrete, solvable sub-tasks. Specifically, our framework consists of three synergistic modules: (1) a text clarifier that uses dialogue-driven reasoning to interactively disambiguate linguistic intent, (2) a vision clarifier that delivers real-time guidance feedback, instructing users to adjust their positioning for improved capture quality, and (3) a cross-modal clarifier with grounding mechanism that robustly interprets 3D pointing gestures and identifies the specific objects users are pointing to. Extensive experiments demonstrate that our framework improves the intent clarification performance of small language models (4--8B) by approximately 30%, making them competitive with significantly larger counterparts. We also observe consistent gains when applying our framework to these larger models. Furthermore, our vision clarifier increases corrective guidance accuracy by over 20%, and our cross-modal clarifier improves semantic answer accuracy for referential grounding by 5%. Overall, our method provides a plug-and-play framework that effectively resolves multimodal ambiguity and significantly enhances user experience in egocentric interaction.

cs.HC↗

arXiv2025

CHROMA: Consistent Harmonization of Multi-View Appearance via Bilateral Grid Prediction

Modern camera pipelines apply extensive on-device processing, such as exposure adjustment, white balance, and color correction, which, while beneficial individually, often introduce photometric inconsistencies across views. These appearance variations violate multi-view consistency and degrade novel view synthesis. Joint optimization of scene-specific representations and per-image appearance embeddings has been proposed to address this issue, but with increased computational complexity and slower training. In this work, we propose a generalizable, feed-forward approach that predicts spatially adaptive bilateral grids to correct photometric variations in a multi-view consistent manner. Our model processes hundreds of frames in a single step, enabling efficient large-scale harmonization, and seamlessly integrates into downstream 3D reconstruction models, providing cross-scene generalization without requiring scene-specific retraining. To overcome the lack of paired data, we employ a hybrid self-supervised rendering loss leveraging 3D foundation models, improving generalization to real-world variations. Extensive experiments show that our approach outperforms or matches the reconstruction quality of existing scene-specific optimization methods with appearance modeling, without significantly affecting the training time of baseline 3D models.

cs.CV↗

arXiv2025

Human Motion Video Generation: A Survey

Human motion video generation has garnered significant research interest due to its broad applications, enabling innovations such as photorealistic singing heads or dynamic avatars that seamlessly dance to music. However, existing surveys in this field focus on individual methods, lacking a comprehensive overview of the entire generative process. This paper addresses this gap by providing an in-depth survey of human motion video generation, encompassing over ten sub-tasks, and detailing the five key phases of the generation process: input, motion planning, motion video generation, refinement, and output. Notably, this is the first survey that discusses the potential of large language models in enhancing human motion video generation. Our survey reviews the latest developments and technological trends in human motion video generation across three primary modalities: vision, text, and audio. By covering over two hundred papers, we offer a thorough overview of the field and highlight milestone works that have driven significant technological breakthroughs. Our goal for this survey is to unveil the prospects of human motion video generation and serve as a valuable resource for advancing the comprehensive applications of digital humans. A complete list of the models examined in this survey is available in Our Repository https://github.com/Winn1y/Awesome-Human-Motion-Video-Generation.

cs.CV↗