arXiv subjects

Zhenzhong Chen

Publications and source records attributed to Zhenzhong Chen.

At least 19 recordsLinked to original sources

arXiv2026

DiffVC-ONE: Diffusion-based Generative Video Compression with One-Step Video Diffusion Transformer

Generative video compression can recover rich visual details at low bitrates, but simultaneously achieving high temporal consistency and low inference cost remains challenging. To address this issue, we propose DiffVC-ONE, a diffusion-based generative video compression framework built on a one-step Video Diffusion Transformer. First, we introduce a Unified Unidirectional Latent Compressor that uses a shared model to efficiently and uniformly compress compact latent slices. We then develop a Video DiT-based One-Step Diffusion Enhancer that uses the reconstructed latent slices as content anchors and performs single-step spatio-temporal perceptual enhancement over an entire group of pictures. Finally, a Hybrid Condition Generator extracts structural, strength, and semantic conditions from the reconstructed content and quantization information. These conditions preserve faithful regions, control the degree of generative enhancement, and supplement content-aware perceptual details during one-step diffusion enhancement. Extensive experiments on multiple standard benchmarks demonstrate that DiffVC-ONE achieves state-of-the-art perceptual quality and temporal consistency with low inference cost.

cs.CV↗

arXiv2026

ETC: Extreme Token Compression via Task-aware Visual Information Distillation in VLMs

In Vision-Language Models (VLMs), high-resolution images produce a large number of visual tokens, resulting in high computational costs and KV-cache overhead during inference. To address this problem, we propose an Extreme Token Compression (ETC) framework that minimizes task loss when reducing the number of input tokens based on the principle of variational information distillation. Specifically, from an information-theoretic perspective, we show that minimizing task loss requires the compact representation to preserve the instruction-aware sufficient statistic of the task-relevant visual information for prediction. In practice, ETC leverages text-to-image cross-attention to weight the original visual features to approximate the latent instruction-aware predictive statistic. Moreover, ETC introduces a variational information distillation, enabling the compact representation to preserve the essential information to recover this predictive statistic. Experiments on LLaVA-1.5-7B and Qwen3-VL-2B show that ETC remains effective even under single-token compression, substantially reducing KV-cache overhead while retaining strong task performance.

cs.CV↗

arXiv2026

XL-DocBench: Benchmarking Evidence-Grounded Extra-Long Document Understanding

Real-world document tasks often ask professionals to answer questions from annual reports, regulations, clinical guidelines, and technical manuals that span hundreds or thousands of pages. Some questions also require comparing related reports. Reliable long-document understanding is therefore a prerequisite for using LLMs in compliance, clinical, financial, and engineering workflows, where decisions must be traceable to specific evidence pages and the cost of an unsupported answer is high -- yet most existing benchmarks still measure short-context or single-page QA. We introduce XL-DocBench, a fully human-verified benchmark for extra-long document understanding, with 1,519 retained questions from six professional domains and contexts up to 2,303 pages. XL-DocBench goes beyond page-level lookup. 1,103 examples (72.6\%) use multiple evidence pages. The final set also includes 556 questions (36.6\%) that use tables, charts, or figures, and 165 questions (10.9\%) that require evidence from multiple documents. Each question has one of twelve reasoning labels, expert-annotated evidence pages, a typed verification rule, and an answer format, including 218 None-answer cases. We build the benchmark with a tree-guided synthesis pipeline followed by artifact filters and full verification by 194 human experts. By coupling extra-long professional contexts with page-level evidence and typed rules, XL-DocBench fills a gap left by prior single-page, short multi-page, or text-only long-context benchmarks, and lets future work attribute system failures to retrieval, evidence use, or rule following rather than to a single leaderboard score. The results show that current systems still struggle with long contexts, multi-page evidence, and structured reasoning over professional documents.

cs.CL↗

arXiv2026

DocAtlas: Long-Document Understanding as Mutable-State Interaction

Long-document understanding requires models to find and combine evidence across many pages, layouts, tables, figures, and charts. Existing retrieval-augmented systems usually select evidence from a static index before generation, while recent agentic systems add multi-turn tool use but often rely on frozen proprietary backbones whose behavior is set by prompts. We present DocAtlas, a system that treats long-document understanding as a mutable-state information-seeking process. We instantiate DocAtlas as a mutable document harness: an external environment that determines what document information is searched, read, stored, reviewed, and shown to the model at each step. Given a document and question, the harness exposes search, reading, note-taking, and review tools, maintains a hierarchical tree and note store, and updates both as the agent records evidence. DocAtlas combines self-improving retrieval, selective evidence access, and active working memory under a fixed context budget. The same harness supports inference-time use with large VLMs and end-to-end reinforcement learning for compact VLM agents. With GPT-5.4, DocAtlas reaches 71.4\% on MMLongBench-Doc, exceeding the human-expert reference of 65.8\%. A Qwen3.5-4B VLM trained with end-to-end RL in the DocAtlas environment reaches 63.7\%, compared with a 54.4\% direct-input baseline, showing that mutable document-harness design can improve compact document agents by a large margin.

cs.CL↗

arXiv2026

DiffCVE: Diffusion-based Compressed Video Enhancement

Perceptual quality enhancement of severely compressed videos remains challenging due to complex artifact patterns and substantial information loss. Recent diffusion models have demonstrated strong generative capability for visual restoration, but directly applying them to compressed video often ignores compression degradation characteristics and may introduce structure-inconsistent hallucinations. To address this issue, this paper presents a diffusion-based compressed video enhancement method, named DiffCVE. Coding Prior-enhanced Dual Conditioning (CPDC) branches are designed to jointly model compressed video and coding prior conditions, where coding priors including residuals and motion vectors provide complementary structural and motion guidance during the diffusion denoising process. To make the diffusion process aware of compression severity, a Compression Degradation Semantic Prompting (CDSP) mechanism is introduced to leverage QP-conditioned textual prompts together with LoRA fine-tuning. In addition, a Coding Prior-guided Weighted Fusion (CPWF) module is incorporated into the VAE decoder to fuse VAE encoder and coding prior encoder features with QP-predicted weights. Extensive experiments demonstrate the effectiveness of the proposed method in improving perceptual quality, especially under severe compression settings. The project page with enhanced video demonstrations is available at https://wqmaker.github.io/projects/DiffCVE/.

cs.CV↗

arXiv2026

The First Challenge on Mobile Real-World Image Super-Resolution at NTIRE 2026: Benchmark Results and Method Overview

This paper provides a review of the NTIRE 2026 challenge on mobile real-world image super-resolution, highlighting the proposed solutions and the resulting outcomes. The challenge aims to recover high-resolution (HR) images from low-resolution (LR) counterparts generated through unknown degradations with a x4 scaling factor while ensuring the models remain executable on mobile devices. The objective is to develop effective and efficient network designs or solutions that achieve state-of-the-art real-world image super-resolution performance. The track of the challenge evaluates performance using a weighted combination of image quality assessment (IQA) score and speedup ratios. The competition attracted 108 registrants, with 16 teams achieving a valid score in the final ranking. This collaborative effort advances the performance of mobile real-world image super-resolution while offering an in-depth overview of the latest trends in the field.

cs.CV↗

arXiv2026

Robotic Scene Cloning:Advancing Zero-Shot Robotic Scene Adaptation in Manipulation via Visual Prompt Editing

Modern robots can perform a wide range of simple tasks and adapt to diverse scenarios in the well-trained environment. However, deploying pre-trained robot models in real-world user scenarios remains challenging due to their limited zero-shot capabilities, often necessitating extensive on-site data collection. To address this issue, we propose Robotic Scene Cloning (RSC), a novel method designed for scene-specific adaptation by editing existing robot operation trajectories. RSC achieves accurate and scene-consistent sample generation by leveraging a visual prompting mechanism and a carefully tuned condition injection module. Not only transferring textures but also performing moderate shape adaptations in response to the visual prompts, RSC demonstrates reliable task performance across a variety of object types. Experiments across various simulated and real-world environments demonstrate that RSC significantly enhances policy generalization in target environments.

cs.RO↗

arXiv2026

DiffVC-RT: Towards Practical Real-Time Diffusion-based Perceptual Neural Video Compression

The practical deployment of diffusion-based Neural Video Compression (NVC) faces critical challenges, including severe information loss, prohibitive inference latency, and poor temporal consistency. To bridge this gap, we propose DiffVC-RT, the first framework designed to achieve real-time diffusion-based perceptual NVC. First, we introduce an Efficient and Informative Model Architecture. Through strategic module replacements and pruning, this architecture significantly reduces computational complexity while mitigating structural information loss. Second, to address generative flickering artifacts, we propose Explicit and Implicit Consistency Modeling. We enhance temporal consistency by explicitly incorporating a zero-cost Online Temporal Shift Module within the U-Net, complemented by hybrid implicit consistency constraints. Finally, we present an Asynchronous and Parallel Decoding Pipeline incorporating Mixed Half Precision, which enables asynchronous latent decoding and parallel frame reconstruction via a Batch-dimension Temporal Shift design. Experiments show that DiffVC-RT achieves 80.1% bitrate savings in terms of LPIPS over VTM-17.0 on HEVC dataset with real-time encoding and decoding speeds of 206 / 30 fps for 720p videos on an NVIDIA H800 GPU, marking a significant milestone in diffusion-based video compression.

cs.CV↗

arXiv2025

SemCo: Toward Semantic Coherent Visual Relationship Forecasting

Visual Relationship Forecasting (VRF) aims to anticipate relations among objects without observing future visual content. The task relies on capturing and modeling the semantic coherence in object interactions, as it underpins the evolution of events and scenes in videos. However, existing VRF datasets offer limited support for learning such coherence due to noisy annotations in the datasets and weak correlations between different actions and relationship transitions in subject-object pair. Furthermore, existing methods struggle to distinguish similar relationships and overfit to unchanging relationships in consecutive frames. To address these challenges, we present SemCoBench, a benchmark that emphasizes semantic coherence for visual relationship forecasting. Based on action labels and short-term subject-object pairs, SemCoBench decomposes relationship categories and dynamics by cleaning and reorganizing video datasets to ensure predicting semantic coherence in object interactions. In addition, we also present Semantic Coherent Transformer method (SemCoFormer) to model the semantic coherence with a Relationship Augmented Module (RAM) and a Coherence Reasoning Module (CRM). RAM is designed to distinguish similar relationships, and CRM facilitates the model's focus on the dynamics in relationships. The experimental results on SemCoBench demonstrate that modeling the semantic coherence is a key step toward reasonable, fine-grained, and diverse visual relationship forecasting, contributing to a more comprehensive understanding of video scenes.

cs.CV↗

arXiv2025

SemPT: Semantic Prompt Tuning for Vision-Language Models

Visual transfer learning for unseen categories presents an active research topic yet a challenging task, due to the inherent conflict between preserving category-specific representations and acquiring transferable knowledge. Vision-Language Models (VLMs) pre-trained on large amounts of image-text pairs offer a promising solution. However, existing prompt tuning methods rely on sparse category labels or disparate LLM-generated descriptions, which fragment knowledge representation and hinder transferability. To address this limitation, we introduce Semantic Prompt Tuning (SemPT), a novel framework that tackles the generalization challenge by leveraging shared attribute-level knowledge across categories. Specifically, SemPT adopts a two-step prompting strategy to guide LLM in extracting shared visual attributes and generating attribute-level descriptions, capturing transferable semantic cues beyond labels while ensuring coherent structure. Then, visually guided weighting is applied to the embeddings of attribute-level descriptions to reduce noise from irrelevant attributes and enhance the text embeddings. Additionally, image embeddings are jointly aligned with both label and attribute-enhanced text embeddings, balancing discrimination for seen categories and transferability to unseen ones. Considering the availability of category exposure, our inference dynamically selects between standard label embeddings for seen categories and attribute-enhanced embeddings for unseen ones to ensure effective adaptation. Extensive experiments on 15 benchmark datasets demonstrate that SemPT achieves state-of-the-art performance across various settings, including base-to-novel generalization, cross-dataset transfer, cross-domain transfer, and few-shot learning.

cs.CV↗

arXiv2025

DiffVC-OSD: One-Step Diffusion-based Perceptual Neural Video Compression Framework

In this work, we first propose DiffVC-OSD, a One-Step Diffusion-based Perceptual Neural Video Compression framework. Unlike conventional multi-step diffusion-based methods, DiffVC-OSD feeds the reconstructed latent representation directly into a One-Step Diffusion Model, enhancing perceptual quality through a single diffusion step guided by both temporal context and the latent itself. To better leverage temporal dependencies, we design a Temporal Context Adapter that encodes conditional inputs into multi-level features, offering more fine-grained guidance for the Denoising Unet. Additionally, we employ an End-to-End Finetuning strategy to improve overall compression performance. Extensive experiments demonstrate that DiffVC-OSD achieves state-of-the-art perceptual compression performance, offers about 20$\times$ faster decoding and a 86.92\% bitrate reduction compared to the corresponding multi-step diffusion-based variant.

eess.IV↗

arXiv2025

LOP: Learning Optimal Pruning for Efficient On-Demand MLLMs Scaling

Structural pruning techniques are essential for deploying multimodal large language models (MLLMs) across various hardware platforms, from edge devices to cloud servers. However, current pruning methods typically determine optimal strategies through iterative search processes, resulting in substantial computational overhead for on-demand MLLMs adaptation. To address this challenge, we propose LOP, an efficient neural pruning framework that learns optimal pruning strategies from the target pruning constraint, eliminating the need for computationally expensive search-based methods. LOP approach trains autoregressive neural networks (NNs) to directly predict layer-wise pruning strategies adaptive to the target pruning constraint, eliminating the time-consuming iterative searches. Experimental results across multiple tasks show that LOP outperforms state-of-the-art pruning methods in various metrics while achieving up to three orders of magnitude speedup.

cs.CV↗

arXiv2025

RSFAKE-1M: A Large-Scale Dataset for Detecting Diffusion-Generated Remote Sensing Forgeries

Detecting forged remote sensing images is becoming increasingly critical, as such imagery plays a vital role in environmental monitoring, urban planning, and national security. While diffusion models have emerged as the dominant paradigm for image generation, their impact on remote sensing forgery detection remains underexplored. Existing benchmarks primarily target GAN-based forgeries or focus on natural images, limiting progress in this critical domain. To address this gap, we introduce RSFAKE-1M, a large-scale dataset of 500K forged and 500K real remote sensing images. The fake images are generated by ten diffusion models fine-tuned on remote sensing data, covering six generation conditions such as text prompts, structural guidance, and inpainting. This paper presents the construction of RSFAKE-1M along with a comprehensive experimental evaluation using both existing detectors and unified baselines. The results reveal that diffusion-based remote sensing forgeries remain challenging for current methods, and that models trained on RSFAKE-1M exhibit notably improved generalization and robustness. Our findings underscore the importance of RSFAKE-1M as a foundation for developing and evaluating next-generation forgery detection approaches in the remote sensing domain. The dataset and other supplementary materials are available at https://huggingface.co/datasets/TZHSW/RSFAKE/.

cs.CV↗

arXiv2025

Training-Free Reasoning and Reflection in MLLMs

Recent advances in Reasoning LLMs (e.g., DeepSeek-R1 and OpenAI-o1) have showcased impressive reasoning capabilities via reinforcement learning. However, extending these capabilities to Multimodal LLMs (MLLMs) is hampered by the prohibitive costs of retraining and the scarcity of high-quality, verifiable multimodal reasoning datasets. This paper introduces FRANK Model, a training-FRee ANd r1-liKe MLLM that imbues off-the-shelf MLLMs with reasoning and reflection abilities, without any gradient updates or extra supervision. Our key insight is to decouple perception and reasoning across MLLM decoder layers. Specifically, we observe that compared to the deeper decoder layers, the shallow decoder layers allocate more attention to visual tokens, while the deeper decoder layers concentrate on textual semantics. This observation motivates a hierarchical weight merging approach that combines a visual-pretrained MLLM with a reasoning-specialized LLM. To this end, we propose a layer-wise, Taylor-derived closed-form fusion mechanism that integrates reasoning capacity into deep decoder layers while preserving visual grounding in shallow decoder layers. Extensive experiments on challenging multimodal reasoning benchmarks demonstrate the effectiveness of our approach. On the MMMU benchmark, our model FRANK-38B achieves an accuracy of 69.2, outperforming the strongest baseline InternVL2.5-38B by +5.3, and even surpasses the proprietary GPT-4o model. Our project homepage is at: http://iip.whu.edu.cn/frank/index.html

cs.CV↗

arXiv2025

LaMD: Latent Motion Diffusion for Image-Conditional Video Generation

The video generation field has witnessed rapid improvements with the introduction of recent diffusion models. While these models have successfully enhanced appearance quality, they still face challenges in generating coherent and natural movements while efficiently sampling videos. In this paper, we propose to condense video generation into a problem of motion generation, to improve the expressiveness of motion and make video generation more manageable. This can be achieved by breaking down the video generation process into latent motion generation and video reconstruction. Specifically, we present a latent motion diffusion (LaMD) framework, which consists of a motion-decomposed video autoencoder and a diffusion-based motion generator, to implement this idea. Through careful design, the motion-decomposed video autoencoder can compress patterns in movement into a concise latent motion representation. Consequently, the diffusion-based motion generator is able to efficiently generate realistic motion on a continuous latent space under multi-modal conditions, at a cost that is similar to that of image diffusion models. Results show that LaMD generates high-quality videos on various benchmark datasets, including BAIR, Landscape, NATOPS, MUG and CATER-GEN, that encompass a variety of stochastic dynamics and highly controllable movements on multiple image-conditional video generation tasks, while significantly decreases sampling time.

cs.CV↗

arXiv2025

Frequency Enhancement for Image Demosaicking

Recovering high-frequency textures in image demosaicking remains a challenging issue. While existing methods introduced elaborate spatial learning methods, they still exhibit limited performance. To address this issue, a frequency enhancement approach is proposed. Based on the frequency analysis of color filter array (CFA)/demosaicked/ground truth images, we propose Dual-path Frequency Enhancement Network (DFENet), which reconstructs RGB images in a divide-and-conquer manner through fourier-domain frequency selection. In DFENet, two frequency selectors are employed, each selecting a set of frequency components for processing along separate paths. One path focuses on generating missing information through detail refinement in spatial domain, while the other aims at suppressing undesirable frequencies with the guidance of CFA images in frequency domain. Multi-level frequency supervision with a stagewise training strategy is employed to further improve the reconstruction performance. With these designs, the proposed DFENet outperforms other state-of-the-art algorithms on different datasets and demonstrates significant advantages on hard cases. Moreover, to better assess algorithms' ability to reconstruct high-frequency textures, a new dataset, LineSet37, is contributed, which consists of 37 artificially designed and generated images. These images feature complex line patterns and are prone to severe visual artifacts like color moiré after demosaicking. Experiments on LineSet37 offer a more targeted evaluation of performance on challenging cases. The code and dataset are available at https://github.com/VelvetReverie/DFENet-demosaicking.

cs.CV↗

arXiv2025

Timestep-Aware Diffusion Model for Extreme Image Rescaling

Image rescaling aims to learn the optimal low-resolution (LR) image that can be accurately reconstructed to its original high-resolution (HR) counterpart, providing an efficient image processing and storage method for ultra-high definition media. However, extreme downscaling factors pose significant challenges to the upscaling process due to its highly ill-posed nature, causing existing image rescaling methods to struggle in generating semantically correct structures and perceptual friendly textures. In this work, we propose a novel framework called Timestep-Aware Diffusion Model (TADM) for extreme image rescaling, which performs rescaling operations in the latent space of a pre-trained autoencoder and effectively leverages powerful natural image priors learned by a pre-trained text-to-image diffusion model. Specifically, TADM adopts a pseudo-invertible module to establish the bidirectional mapping between the latent features of the HR image and the target-sized LR image. Then, the rescaled latent features are enhanced by a pre-trained diffusion model to generate more faithful details. Considering the spatially non-uniform degradation caused by the rescaling operation, we propose a novel time-step alignment strategy, which can adaptively allocate the generative capacity of the diffusion model based on the quality of the reconstructed latent features. Extensive experiments demonstrate the superiority of TADM over previous methods in both quantitative and qualitative evaluations.

cs.CV↗

arXiv2025

LongCaptioning: Unlocking the Power of Long Video Caption Generation in Large Multimodal Models

Large Multimodal Models (LMMs) have demonstrated exceptional performance in video captioning tasks, particularly for short videos. However, as the length of the video increases, generating long, detailed captions becomes a significant challenge. In this paper, we investigate the limitations of LMMs in generating long captions for long videos. Our analysis reveals that open-source LMMs struggle to consistently produce outputs exceeding 300 words, leading to incomplete or overly concise descriptions of the visual content. This limitation hinders the ability of LMMs to provide comprehensive and detailed captions for long videos, ultimately missing important visual information. Through controlled experiments, we find that the scarcity of paired examples with long-captions during training is the primary factor limiting the model's output length. However, manually annotating long-caption examples for long-form videos is time-consuming and expensive. To overcome the annotation bottleneck, we propose the LongCaption-Agent, a framework that synthesizes long caption data by hierarchical semantic aggregation. % aggregating multi-level descriptions. Using LongCaption-Agent, we curated a new long-caption dataset, LongCaption-10K. We also develop LongCaption-Bench, a benchmark designed to comprehensively evaluate the quality of long captions generated by LMMs. By incorporating LongCaption-10K into training, we enable LMMs to generate captions exceeding 1,000 words for long-form videos, while maintaining high output quality. In LongCaption-Bench, our model achieved State-of-The-Art performance, even surpassing larger proprietary models like GPT4o.

cs.CV↗