Feed - arxlens

0

FeatDistill: A Feature Distillation Enhanced Multi-Expert Ensemble Framework for Robust AI-generated Image Detection

cs.CV cs.MM Zhilin Tu, Kemou Li, Fengpeng Li et al. · Mar 23, 2026

FeatDistill tackles robust detection of AI-generated images under real-world degradations via a multi-expert ensemble of CLIP and SigLIP backbones. The framework combines extensive data expansion with a two-stage training paradigm featuring feature-level self-distillation. It aims to balance strong generalization across unseen generators with practical inference efficiency.

The rapid iteration and widespread dissemination of deepfake technology have posed severe challenges to information security, making robust and generalizable detection of AI-generated forged images increasingly important. In this paper, we propose FeatDistill, an AI-generated image detection framework that integrates feature distillation with a multi-expert ensemble, developed for the NTIRE Challenge on Robust AI-Generated Image Detection in the Wild. The framework explicitly targets three practical bottlenecks in real-world forensics: degradation interference, insufficient feature representation, and limited generalization. Concretely, we build a four-backbone Vision Transformer (ViT) ensemble composed of CLIP and SigLIP variants to capture complementary forensic cues. To improve data coverage, we expand the training set and introduce comprehensive degradation modeling, which exposes the detector to diverse quality variations and synthesis artifacts commonly encountered in unconstrained scenarios. We further adopt a two-stage training paradigm: the model is first optimized with a standard binary classification objective, then refined by dense feature-level self-distillation for representation alignment. This design effectively mitigates overfitting and enhances semantic consistency of learned features. At inference time, the final prediction is obtained by averaging the probabilities from four independently trained experts, yielding stable and reliable decisions across unseen generators and complex degradations. Despite the ensemble design, the framework remains efficient, requiring only about 10 GB peak GPU memory. Extensive evaluations in the NTIRE challenge setting demonstrate that FeatDistill achieves strong robustness and generalization under diverse ``in-the-wild'' conditions, offering an effective and practical solution for real-world deepfake image detection.

Read abstractHide abstract

0

StreamingEval: A Unified Evaluation Protocol towards Realistic Streaming Video Understanding

cs.CV cs.MM Guowei Tang, Tianwen Qian, Huanran Zheng et al. · Mar 23, 2026

StreamingEval introduces a unified evaluation framework for Video-LLMs under realistic streaming constraints, moving beyond offline benchmarks to assess continuous, real-time video understanding with limited memory. The protocol enforces a fixed-capacity memory bank and jointly measures encoding throughput (MaxFPS), decoding latency (TTFT), memory usage, and task accuracy via a composite StreamingScore. Experiments reveal that current "online" models often fail under strict streaming constraints, while offline models adapted with FIFO memory banks frequently outperform specialized streaming architectures at the cost of higher resource consumption.

Real-time, continuous understanding of visual signals is essential for real-world interactive AI applications, and poses a fundamental system-level challenge. Existing research on streaming video understanding, however, typically focuses on isolated aspects such as question-answering accuracy under limited visual context or improvements in encoding efficiency, while largely overlooking practical deployability under realistic resource constraints. To bridge this gap, we introduce StreamingEval, a unified evaluation framework for assessing the streaming video understanding capabilities of Video-LLMs under realistic constraints. StreamingEval benchmarks both mainstream offline models and recent online video models under a standardized protocol, explicitly characterizing the trade-off between efficiency, storage and accuracy. Specifically, we adopt a fixed-capacity memory bank to normalize accessible historical visual context, and jointly evaluate visual encoding efficiency, text decoding latency, and task performance to quantify overall system deployability. Extensive experiments across multiple datasets reveal substantial gaps between current Video-LLMs and the requirements of realistic streaming applications, providing a systematic basis for future research in this direction. Codes will be released at https://github.com/wwgTang-111/StreamingEval1.

Read abstractHide abstract

0

DSCSNet: A Dynamic Sparse Compression Sensing Network for Closely-Spaced Infrared Small Target Unmixing

cs.CV cs.MM Zhiyang Tang, Yiming Zhu, Ruimin Huang et al. · Mar 22, 2026

This paper tackles the Close Small Object Unmixing (CSOU) problem for infrared imagery, where distant clustered targets appear as overlapping mixed spots due to optical diffraction limits. The authors propose DSCSNet, a deep-unfolded network that unrolls the ADMM algorithm with learnable parameters to recover target count, sub-pixel positions, and radiant intensities from mixed spots. The core idea is to replace the traditional ℓ2-norm smoothness terms with strict ℓ1-norm sparsity constraints and add a dynamic thresholding mechanism for scene-adaptive reconstruction.

Due to the limitations of optical lens focal length and detector resolution, distant clustered infrared small targets often appear as mixed spots. The Close Small Object Unmixing (CSOU) task aims to recover the number, sub-pixel positions, and radiant intensities of individual targets from these spots, which is a highly ill-posed inverse problem. Existing methods struggle to balance the rigorous sparsity guarantees of model-driven approaches and the dynamic scene adaptability of data-driven methods. To address this dilemma, this paper proposes a Dynamic Sparse Compressed Sensing Network (DSCSNet), a deep-unfolded network that couples the Alternating Direction Method of Multipliers (ADMM) with learnable parameters. Specifically, we embed a strict $\ell_1$-norm sparsity constraint into the auxiliary variable update step of ADMM to replace the traditional $\ell_2$-norm smoothness-promoting terms, which effectively preserves the discrete energy peaks of small targets. We also integrate a self-attention-based dynamic thresholding mechanism into the reconstruction stage, which adaptively adjusts the sparsification intensity using the sparsity-enhanced information from the iterative process. These modules are jointly optimized end-to-end across the three iterative steps of ADMM. Retaining the physical logic of compressed sensing, DSCSNet achieves robust sparsity induction and scene adaptability, thus enhancing the unmixing accuracy and generalization in complex infrared scenarios. Extensive experiments on the synthetic infrared dataset CSIST-100K demonstrate that DSCSNet outperforms state-of-the-art methods in key metrics such as CSO-mAP and sub-pixel localization error.

Read abstractHide abstract

0

OrbitStream: Training-Free Adaptive 360-degree Video Streaming via Semantic Potential Fields

cs.NI cs.CV cs.MM Aizierjiang Aiersilan, Zhangfei Yang · Mar 22, 2026

OrbitStream addresses adaptive 360° video streaming for teleoperation by proposing a training-free framework that combines semantic scene understanding with robust control theory. It formulates viewport prediction as a Gravitational Viewport Prediction (GVP) problem where semantic objects (pedestrians, vehicles) generate potential fields that "attract" user gaze with task-relevant mass, while a Saturation-Based Proportional-Derivative (PD) Controller handles bitrate adaptation. This offers an interpretable, zero-shot alternative to black-box Deep Reinforcement Learning methods for safety-critical systems where deployment constraints prohibit lengthy training.

Adaptive 360{\deg} video streaming for teleoperation faces dual challenges: viewport prediction under uncertain gaze patterns and bitrate adaptation over volatile wireless channels. While data-driven and Deep Reinforcement Learning (DRL) methods achieve high Quality of Experience (QoE), their "black-box" nature and reliance on training data can limit deployment in safety-critical systems. To address this, we propose OrbitStream, a training-free framework that combines semantic scene understanding with robust control theory. We formulate viewport prediction as a Gravitational Viewport Prediction (GVP) problem, where semantic objects generate potential fields that attract user gaze. Furthermore, we employ a Saturation-Based Proportional-Derivative (PD) Controller for buffer regulation. On object-rich teleoperation traces, OrbitStream achieves a 94.7\% zero-shot viewport prediction accuracy without user-specific profiling, approaching trajectory-extrapolation baselines ($\sim$98.5\%). Across 3,600 Monte Carlo simulations on diverse network traces, OrbitStream yields a mean QoE of 2.71. It ranks second among 12 evaluated algorithms, close to the top-performing BOLA-E (2.80) while outperforming FastMPC (1.84). The system exhibits an average decision latency of 1.01 ms with minimal rebuffering events. By providing competitive QoE with interpretability and zero training overhead, OrbitStream demonstrates that physics-based control, combined with semantic modeling, offers a practical solution for 360{\deg} streaming in teleoperation.

Read abstractHide abstract

0

Structured Visual Narratives Undermine Safety Alignment in Multimodal Large Language Models

cs.CR cs.AI cs.MM Rui Yang Tan, Yujia Hu, Roy Ka-Wei Lee · Mar 23, 2026

This paper exposes a critical vulnerability in Multimodal Large Language Models (MLLMs): safety alignment fails when harmful intent is embedded in structured visual narratives. The authors introduce ComicJailbreak, a benchmark of 1,167 three-panel comics where panels 1–2 establish narrative context and panel 3 contains a blank speech bubble filled with a paraphrased harmful goal. The model is prompted to "complete the comic" by generating the fourth panel. Across 15 state-of-the-art MLLMs, comic-based attacks achieve ensemble success rates exceeding 90% on Gemini-family models and 85%+ on most open-source models—substantially outperforming plain-text and random-image baselines. The work also reveals that existing defenses (AdaShield, Attack as Defense) trigger severe over-refusal on benign prompts, and that automated safety judges are unreliable on sensitive-but-benign content.

Multimodal Large Language Models (MLLMs) extend text-only LLMs with visual reasoning, but also introduce new safety failure modes under visually grounded instructions. We study comic-template jailbreaks that embed harmful goals inside simple three-panel visual narratives and prompt the model to role-play and "complete the comic." Building on JailbreakBench and JailbreakV, we introduce ComicJailbreak, a comic-based jailbreak benchmark with 1,167 attack instances spanning 10 harm categories and 5 task setups. Across 15 state-of-the-art MLLMs (six commercial and nine open-source), comic-based attacks achieve success rates comparable to strong rule-based jailbreaks and substantially outperform plain-text and random-image baselines, with ensemble success rates exceeding 90% on several commercial models. Then, with the existing defense methodologies, we show that these methods are effective against the harmful comics, they will induce a high refusal rate when prompted with benign prompts. Finally, using automatic judging and targeted human evaluation, we show that current safety evaluators can be unreliable on sensitive but non-harmful content. Our findings highlight the need for safety alignment robust to narrative-driven multimodal jailbreaks.

Read abstractHide abstract

0

Cross-Scenario Deraining Adaptation with Unpaired Data: Superpixel Structural Priors and Multi-Stage Pseudo-Rain Synthesis

cs.CV cs.AI cs.GR Kangbo Zhao, Miaoxin Guan, Xiang Chen et al. · Mar 23, 2026

This paper tackles the domain generalization problem in image deraining, where models trained on synthetic data fail catastrophically on out-of-distribution (OOD) real-world scenarios. The authors propose a three-stage pipeline—Superpixel Generation, Resolution-adaptive Fusion, and Pseudo-label Re-Synthesis—that adapts source-domain models to target domains using only unpaired rain-free images, eliminating the need for costly paired rainy data collection.

Image deraining plays a pivotal role in low-level computer vision, serving as a prerequisite for robust outdoor surveillance and autonomous driving systems. While deep learning paradigms have achieved remarkable success in firmly aligned settings, they often suffer from severe performance degradation when generalized to unseen Out-of-Distribution (OOD) scenarios. This failure stems primarily from the significant domain discrepancy between synthetic training datasets and the complex physical dynamics of real-world rain. To address these challenges, this paper proposes a pioneering cross-scenario deraining adaptation framework. Diverging from conventional approaches, our method obviates the requirements for paired rainy observations in the target domain, leveraging exclusively rain-free background images. We design a Superpixel Generation (Sup-Gen) module to extract stable structural priors from the source domain using Simple Linear Iterative Clustering. Subsequently, a Resolution-adaptive Fusion strategy is introduced to align these source structures with target backgrounds through texture similarity, ensuring the synthesis of diverse and realistic pseudo-data. Finally, we implement a pseudo-label re-Synthesize mechanism that employs multi-stage noise generation to simulate realistic rain streaks. This framework functions as a versatile plug-and-play module capable of seamless integration into arbitrary deraining architectures. Extensive experiments on state-of-the-art models demonstrate that our approach yields remarkable PSNR gains of up to 32% to 59% in OOD domains while significantly accelerating training convergence.

Read abstractHide abstract

0

Harmful Visual Content Manipulation Matters in Misinformation Detection Under Multimedia Scenarios

cs.LG cs.AI cs.MM Bing Wang, Ximing Li, Changchun Li et al. · Mar 22, 2026

This paper tackles multimodal misinformation detection by distinguishing between harmful and harmless visual content manipulation—a nuance often overlooked by existing methods. The authors propose Havc-m4d, a framework that extracts manipulation and intention features using weakly-supervised positive-unlabeled (PU) learning to overcome the lack of ground-truth manipulation labels. By treating real articles with manipulated visuals as likely harmless and fake articles as potentially harmful, the method introduces intention-aware cues that consistently improve detection across four benchmark datasets.

Nowadays, the widespread dissemination of misinformation across numerous social media platforms has led to severe negative effects on society. To address this challenge, the automatic detection of misinformation, particularly under multimedia scenarios, has gained significant attention from both academic and industrial communities, leading to the emergence of a research task known as Multimodal Misinformation Detection (MMD). Typically, current MMD approaches focus on capturing the semantic relationships and inconsistency between various modalities but often overlook certain critical indicators within multimodal content. Recent research has shown that manipulated features within visual content in social media articles serve as valuable clues for MMD. Meanwhile, we argue that the potential intentions behind the manipulation, e.g., harmful and harmless, also matter in MMD. Therefore, in this study, we aim to identify such multimodal misinformation by capturing two types of features: manipulation features, which represent if visual content has been manipulated, and intention features, which assess the nature of these manipulations, distinguishing between harmful and harmless intentions. Unfortunately, the manipulation and intention labels that supervise these features to be discriminative are unknown. To address this, we introduce two weakly supervised indicators as substitutes by incorporating supplementary datasets focused on image manipulation detection and framing two different classification tasks as positive and unlabeled learning issues. With this framework, we introduce an innovative MMD approach, titled Harmful Visual Content Manipulation Matters in MMD (HAVC-M4 D). Comprehensive experiments conducted on four prevalent MMD datasets indicate that HAVC-M4 D significantly and consistently enhances the performance of existing MMD methods.

Read abstractHide abstract

Nothing here yet