Xiaoming Wei

Chuyi Li, Lulu Li, Hongliang Jiang et al.

For years, the YOLO series has been the de facto industry-level standard for efficient object detection. The YOLO community has prospered overwhelmingly to enrich its use in a multitude of hardware platforms and abundant scenarios. In this technical report, we strive to push its limits to the next level, stepping forward with an unwavering mindset for industry application. Considering the diverse requirements for speed and accuracy in the real environment, we extensively examine the up-to-date object detection advancements either from industry or academia. Specifically, we heavily assimilate ideas from recent network design, training strategies, testing techniques, quantization, and optimization methods. On top of this, we integrate our thoughts and practice to build a suite of deployment-ready networks at various scales to accommodate diversified use cases. With the generous permission of YOLO authors, we name it YOLOv6. We also express our warm welcome to users and contributors for further enhancement. For a glimpse of performance, our YOLOv6-N hits 35.9% AP on the COCO dataset at a throughput of 1234 FPS on an NVIDIA Tesla T4 GPU. YOLOv6-S strikes 43.5% AP at 495 FPS, outperforming other mainstream detectors at the same scale~(YOLOv5-S, YOLOX-S, and PPYOLOE-S). Our quantized version of YOLOv6-S even brings a new state-of-the-art 43.3% AP at 869 FPS. Furthermore, YOLOv6-M/L also achieves better accuracy performance (i.e., 49.5%/52.3%) than other detectors with a similar inference speed. We carefully conducted experiments to validate the effectiveness of each component. Our code is made available at https://github.com/meituan/YOLOv6.

Zhengcong Fei, Junshi Huang, Xiaoming Wei et al.

Existing approaches to image captioning usually generate the sentence word-by-word from left to right, with the constraint of conditioned on local context including the given image and history generated words. There have been many studies target to make use of global information during decoding, e.g., iterative refinement. However, it is still under-explored how to effectively and efficiently incorporate the future context. To respond to this issue, inspired by that Non-Autoregressive Image Captioning (NAIC) can leverage two-side relation with modified mask operation, we aim to graft this advance to the conventional Autoregressive Image Captioning (AIC) model while maintaining the inference efficiency without extra time cost. Specifically, AIC and NAIC models are first trained combined with shared visual encoders, forcing the visual encoder to contain sufficient and valid future context; then the AIC model is encouraged to capture the causal dynamics of cross-layer interchanging from NAIC model on its unconfident words, which follows a teacher-student paradigm and optimized with the distribution calibration training objective. Empirical evidences demonstrate that our proposed approach clearly surpass the state-of-the-art baselines in both automatic metrics and human evaluations on the MS COCO benchmark. The source code is available at: https://github.com/feizc/Future-Caption.

Meituan LongCat Team, Xunliang Cai, Meng Cheng et al.

Despite advances in audio-driven video generation, achieving commercial-grade stability remains challenging. We present LongCat-Video-Avatar 1.5, an upgraded open-source framework prioritizing systematic engineering and production-readiness over architectural novelty. By upgrading the audio encoder to Whisper Large and meticulously scaling our training recipes, v1.5 achieves accurate lip-synchronization, full-body temporal stability, and robust long-video generation with strict identity consistency. Through rigorous data curation and RLHF Training, the model readily generalizes to stylized domains such as anime and animals, and natively handles complex real-world conditions, such as multi-person interactions and object handling. Furthermore, addressing the practical demands of industrial deployment, we employ advanced step distillation to accelerate inference to an optimal 8 NFE, achieving a favorable trade-off between serving efficiency and visual fidelity. The superiority of our approach is validated through extensive quantitative metrics and a rigorous human evaluation conducted on a comprehensive benchmark of over 500 diverse test cases. Results show that v1.5 achieves competitive or superior performance compared to leading closed-source systems (e.g., HeyGen, OmniHuman 1.5, Kling Avatar 2.0) across human-likeness ratings and expert-level quality assessments on our benchmark. With its open-source release, LongCat-Video-Avatar 1.5 narrows the gap between academic research prototypes and commercial-grade deployment.

Shaofei Huang, Rui Ling, Hongyu Li et al.

In this paper, we propose an Audio-Language-Referenced SAM 2 (AL-Ref-SAM 2) pipeline to explore the training-free paradigm for audio and language-referenced video object segmentation, namely AVS and RVOS tasks. The intuitive solution leverages GroundingDINO to identify the target object from a single frame and SAM 2 to segment the identified object throughout the video, which is less robust to spatiotemporal variations due to a lack of video context exploration. Thus, in our AL-Ref-SAM 2 pipeline, we propose a novel GPT-assisted Pivot Selection (GPT-PS) module to instruct GPT-4 to perform two-step temporal-spatial reasoning for sequentially selecting pivot frames and pivot boxes, thereby providing SAM 2 with a high-quality initial object prompt. Within GPT-PS, two task-specific Chain-of-Thought prompts are designed to unleash GPT's temporal-spatial reasoning capacity by guiding GPT to make selections based on a comprehensive understanding of video and reference information. Furthermore, we propose a Language-Binded Reference Unification (LBRU) module to convert audio signals into language-formatted references, thereby unifying the formats of AVS and RVOS tasks in the same pipeline. Extensive experiments on both tasks show that our training-free AL-Ref-SAM 2 pipeline achieves performances comparable to or even better than fully-supervised fine-tuning methods. The code is available at: https://github.com/appletea233/AL-Ref-SAM2.

Dengsheng Chen, Jie Hu, Wenwen Qiang et al.

Over past few years afterward the birth of ResNet, skip connection has become the defacto standard for the design of modern architectures due to its widespread adoption, easy optimization and proven performance. Prior work has explained the effectiveness of the skip connection mechanism from different perspectives. In this work, we deep dive into the model's behaviors with skip connections which can be formulated as a learnable Markov chain. An efficient Markov chain is preferred as it always maps the input data to the target domain in a better way. However, while a model is explained as a Markov chain, it is not guaranteed to be optimized following an efficient Markov chain by existing SGD-based optimizers which are prone to get trapped in local optimal points. In order to towards a more efficient Markov chain, we propose a simple routine of penal connection to make any residual-like model become a learnable Markov chain. Aside from that, the penal connection can also be viewed as a particular model regularization and can be easily implemented with one line of code in the most popular deep learning frameworks~\footnote{Source code: \url{https://github.com/densechen/penal-connection}}. The encouraging experimental results in multi-modal translation and image recognition empirically confirm our conjecture of the learnable Markov chain view and demonstrate the superiority of the proposed penal connection.

Kaiheng Weng, Xiangxiang Chu, Xiaoming Xu et al.

We present a hardware-efficient architecture of convolutional neural network, which has a repvgg-like architecture. Flops or parameters are traditional metrics to evaluate the efficiency of networks which are not sensitive to hardware including computing ability and memory bandwidth. Thus, how to design a neural network to efficiently use the computing ability and memory bandwidth of hardware is a critical problem. This paper proposes a method how to design hardware-aware neural network. Based on this method, we designed EfficientRep series convolutional networks, which are high-computation hardware(e.g. GPU) friendly and applied in YOLOv6 object detection framework. YOLOv6 has published YOLOv6N/YOLOv6S/YOLOv6M/YOLOv6L models in v1 and v2 versions.

Zihan Ding, Tianrui Hui, Junshi Huang et al.

Referring video object segmentation aims to predict foreground labels for objects referred by natural language expressions in videos. Previous methods either depend on 3D ConvNets or incorporate additional 2D ConvNets as encoders to extract mixed spatial-temporal features. However, these methods suffer from spatial misalignment or false distractors due to delayed and implicit spatial-temporal interaction occurring in the decoding phase. To tackle these limitations, we propose a Language-Bridged Duplex Transfer (LBDT) module which utilizes language as an intermediary bridge to accomplish explicit and adaptive spatial-temporal interaction earlier in the encoding phase. Concretely, cross-modal attention is performed among the temporal encoder, referring words and the spatial encoder to aggregate and transfer language-relevant motion and appearance information. In addition, we also propose a Bilateral Channel Activation (BCA) module in the decoding phase for further denoising and highlighting the spatial-temporal consistent features via channel-wise activation. Extensive experiments show our method achieves new state-of-the-art performances on four popular benchmarks with 6.8% and 6.9% absolute AP gains on A2D Sentences and J-HMDB Sentences respectively, while consuming around 7x less computational overhead.

Zhengcong Fei, Mingyuan Fan, Li Zhu et al.

It is well believed that the higher uncertainty in a word of the caption, the more inter-correlated context information is required to determine it. However, current image captioning methods usually consider the generation of all words in a sentence sequentially and equally. In this paper, we propose an uncertainty-aware image captioning framework, which parallelly and iteratively operates insertion of discontinuous candidate words between existing words from easy to difficult until converged. We hypothesize that high-uncertainty words in a sentence need more prior information to make a correct decision and should be produced at a later stage. The resulting non-autoregressive hierarchy makes the caption generation explainable and intuitive. Specifically, we utilize an image-conditioned bag-of-word model to measure the word uncertainty and apply a dynamic programming algorithm to construct the training pairs. During inference, we devise an uncertainty-adaptive parallel beam search technique that yields an empirically logarithmic time complexity. Extensive experiments on the MS COCO benchmark reveal that our approach outperforms the strong baseline and related methods on both captioning quality as well as decoding speed.

Yan Zhu, Junbao Zhuo, Bin Ma et al.

Zero-shot video recognition (ZSVR) is a task that aims to recognize video categories that have not been seen during the model training process. Recently, vision-language models (VLMs) pre-trained on large-scale image-text pairs have demonstrated impressive transferability for ZSVR. To make VLMs applicable to the video domain, existing methods often use an additional temporal learning module after the image-level encoder to learn the temporal relationships among video frames. Unfortunately, for video from unseen categories, we observe an abnormal phenomenon where the model that uses spatial-temporal feature performs much worse than the model that removes temporal learning module and uses only spatial feature. We conjecture that improper temporal modeling on video disrupts the spatial feature of the video. To verify our hypothesis, we propose Feature Factorization to retain the orthogonal temporal feature of the video and use interpolation to construct refined spatial-temporal feature. The model using appropriately refined spatial-temporal feature performs better than the one using only spatial feature, which verifies the effectiveness of the orthogonal temporal feature for the ZSVR task. Therefore, an Orthogonal Temporal Interpolation module is designed to learn a better refined spatial-temporal video feature during training. Additionally, a Matching Loss is introduced to improve the quality of the orthogonal temporal feature. We propose a model called OTI for ZSVR by employing orthogonal temporal interpolation and the matching loss based on VLMs. The ZSVR accuracies on popular video datasets (i.e., Kinetics-600, UCF101 and HMDB51) show that OTI outperforms the previous state-of-the-art method by a clear margin.

Zihan Ding, Zi-han Ding, Tianrui Hui et al.

Panoptic Narrative Grounding (PNG) is an emerging task whose goal is to segment visual objects of things and stuff categories described by dense narrative captions of a still image. The previous two-stage approach first extracts segmentation region proposals by an off-the-shelf panoptic segmentation model, then conducts coarse region-phrase matching to ground the candidate regions for each noun phrase. However, the two-stage pipeline usually suffers from the performance limitation of low-quality proposals in the first stage and the loss of spatial details caused by region feature pooling, as well as complicated strategies designed for things and stuff categories separately. To alleviate these drawbacks, we propose a one-stage end-to-end Pixel-Phrase Matching Network (PPMN), which directly matches each phrase to its corresponding pixels instead of region proposals and outputs panoptic segmentation by simple combination. Thus, our model can exploit sufficient and finer cross-modal semantic correspondence from the supervision of densely annotated pixel-phrase pairs rather than sparse region-phrase pairs. In addition, we also propose a Language-Compatible Pixel Aggregation (LCPA) module to further enhance the discriminative ability of phrase features through multi-round refinement, which selects the most compatible pixels for each phrase to adaptively aggregate the corresponding visual context. Extensive experiments show that our method achieves new state-of-the-art performance on the PNG benchmark with 4.0 absolute Average Recall gains.

Sixiang Chen, Jianyu Lai, Jialin Gao et al.

Image-to-poster generation is a high-demand task requiring not only local adjustments but also high-level design understanding. Models must generate text, layout, style, and visual elements while preserving semantic fidelity and aesthetic coherence. The process spans two regimes: local editing, where ID-driven generation, rescaling, filling, and extending must preserve concrete visual entities; and global creation, where layout- and style-driven tasks rely on understanding abstract design concepts. These intertwined demands make image-to-poster a multi-dimensional process coupling entity-preserving editing with concept-driven creation under image-prompt control. To address these challenges, we propose PosterOmni, a generalized artistic poster creation framework that unlocks the potential of a base edit model for multi-task image-to-poster generation. PosterOmni integrates the two regimes, namely local editing and global creation, within a single system through an efficient data-distillation-reward pipeline: (i) constructing multi-scenario image-to-poster datasets covering six task types across entity-based and concept-based creation; (ii) distilling knowledge between local and global experts for supervised fine-tuning; and (iii) applying unified PosterOmni Reward Feedback to jointly align visual entity-preserving and aesthetic preference across all tasks. Additionally, we establish PosterOmni-Bench, a unified benchmark for evaluating both local editing and global creation. Extensive experiments show that PosterOmni significantly enhances reference adherence, global composition quality, and aesthetic harmony, outperforming all open-source baselines and even surpassing several proprietary systems.

Tianrui Hui, Zihan Ding, Junshi Huang et al.

Panoptic narrative grounding (PNG) aims to segment things and stuff objects in an image described by noun phrases of a narrative caption. As a multimodal task, an essential aspect of PNG is the visual-linguistic interaction between image and caption. The previous two-stage method aggregates visual contexts from offline-generated mask proposals to phrase features, which tend to be noisy and fragmentary. The recent one-stage method aggregates only pixel contexts from image features to phrase features, which may incur semantic misalignment due to lacking object priors. To realize more comprehensive visual-linguistic interaction, we propose to enrich phrases with coupled pixel and object contexts by designing a Phrase-Pixel-Object Transformer Decoder (PPO-TD), where both fine-grained part details and coarse-grained entity clues are aggregated to phrase features. In addition, we also propose a PhraseObject Contrastive Loss (POCL) to pull closer the matched phrase-object pairs and push away unmatched ones for aggregating more precise object contexts from more phrase-relevant object tokens. Extensive experiments on the PNG benchmark show our method achieves new state-of-the-art performance with large margins.

Runze He, Kai Ma, Linjiang Huang et al.

Introducing user-specified visual concepts in image editing is highly practical as these concepts convey the user's intent more precisely than text-based descriptions. We propose FreeEdit, a novel approach for achieving such reference-based image editing, which can accurately reproduce the visual concept from the reference image based on user-friendly language instructions. Our approach leverages the multi-modal instruction encoder to encode language instructions to guide the editing process. This implicit way of locating the editing area eliminates the need for manual editing masks. To enhance the reconstruction of reference details, we introduce the Decoupled Residual ReferAttention (DRRA) module. This module is designed to integrate fine-grained reference features extracted by a detail extractor into the image editing process in a residual way without interfering with the original self-attention. Given that existing datasets are unsuitable for reference-based image editing tasks, particularly due to the difficulty in constructing image triplets that include a reference image, we curate a high-quality dataset, FreeBench, using a newly developed twice-repainting scheme. FreeBench comprises the images before and after editing, detailed editing instructions, as well as a reference image that maintains the identity of the edited object, encompassing tasks such as object addition, replacement, and deletion. By conducting phased training on FreeBench followed by quality tuning, FreeEdit achieves high-quality zero-shot editing through convenient language instructions. We conduct extensive experiments to evaluate the effectiveness of FreeEdit across multiple task types, demonstrating its superiority over existing methods. The code will be available at: https://freeedit.github.io/.

Jiajun Liu, Yibing Wang, Hanghang Ma et al.

Rapid advancements have been made in extending Large Language Models (LLMs) to Large Multi-modal Models (LMMs). However, extending input modality of LLMs to video data remains a challenging endeavor, especially for long videos. Due to insufficient access to large-scale high-quality video data and the excessive compression of visual features, current methods exhibit limitations in effectively processing long videos. In this paper, we introduce Kangaroo, a powerful Video LMM aimed at addressing these challenges. Confronted with issue of inadequate training data, we develop a data curation system to build a large-scale dataset with high-quality annotations for vision-language pre-training and instruction tuning. In addition, we design a curriculum training pipeline with gradually increasing resolution and number of input frames to accommodate long videos. Evaluation results demonstrate that, with 8B parameters, Kangaroo achieves state-of-the-art performance across a variety of video understanding benchmarks while exhibiting competitive results on others. Particularly, on benchmarks specialized for long videos, Kangaroo excels some larger models with over 10B parameters and proprietary models.

Yanbing Zeng, Jia Wang, Hanghang Ma et al.

Integrating image generation and understanding into a single framework has become a pivotal goal in the multimodal domain. However, how understanding can effectively assist generation has not been fully explored. Unlike previous works that focus on leveraging reasoning abilities and world knowledge from understanding models, this paper introduces a novel perspective: leveraging understanding to enhance the fidelity and detail richness of generated images. To this end, we propose Forge-and-Quench, a new unified framework that puts this principle into practice. In the generation process of our framework, an MLLM first reasons over the entire conversational context, including text instructions, to produce an enhanced text instruction. This refined instruction is then mapped to a virtual visual representation, termed the Bridge Feature, via a novel Bridge Adapter. This feature acts as a crucial link, forging insights from the understanding model to quench and refine the generation process. It is subsequently injected into the T2I backbone as a visual guidance signal, alongside the enhanced text instruction that replaces the original input. To validate this paradigm, we conduct comprehensive studies on the design of the Bridge Feature and Bridge Adapter. Our framework demonstrates exceptional extensibility and flexibility, enabling efficient migration across different MLLM and T2I models with significant savings in training overhead, all without compromising the MLLM's inherent multimodal understanding capabilities. Experiments show that Forge-and-Quench significantly improves image fidelity and detail across multiple models, while also maintaining instruction-following accuracy and enhancing world knowledge application. Models and codes are available at https://github.com/YanbingZeng/Forge-and-Quench.

Meituan LongCat Team, Hanghang Ma, Haoxian Tan et al.

We introduce LongCat-Image, a pioneering open-source and bilingual (Chinese-English) foundation model for image generation, designed to address core challenges in multilingual text rendering, photorealism, deployment efficiency, and developer accessibility prevalent in current leading models. 1) We achieve this through rigorous data curation strategies across the pre-training, mid-training, and SFT stages, complemented by the coordinated use of curated reward models during the RL phase. This strategy establishes the model as a new state-of-the-art (SOTA), delivering superior text-rendering capabilities and remarkable photorealism, and significantly enhancing aesthetic quality. 2) Notably, it sets a new industry standard for Chinese character rendering. By supporting even complex and rare characters, it outperforms both major open-source and commercial solutions in coverage, while also achieving superior accuracy. 3) The model achieves remarkable efficiency through its compact design. With a core diffusion model of only 6B parameters, it is significantly smaller than the nearly 20B or larger Mixture-of-Experts (MoE) architectures common in the field. This ensures minimal VRAM usage and rapid inference, significantly reducing deployment costs. Beyond generation, LongCat-Image also excels in image editing, achieving SOTA results on standard benchmarks with superior editing consistency compared to other open-source works. 4) To fully empower the community, we have established the most comprehensive open-source ecosystem to date. We are releasing not only multiple model versions for text-to-image and image editing, including checkpoints after mid-training and post-training stages, but also the entire toolchain of training procedure. We believe that the openness of LongCat-Image will provide robust support for developers and researchers, pushing the frontiers of visual content creation.

Xiangyu Kong, Xiaoyu Jin, Yihan Pan et al.

In natural face-to-face interaction, participants seamlessly alternate between speaking and listening, producing facial behaviors (FBs) that are finely informed by long-range context and naturally exhibit contextual appropriateness and emotional rationality. Interactive Head Generation (IHG) aims to synthesize lifelike avatar head video emulating such capabilities. Existing IHG methods typically condition on dual-track signals (i.e., human user's behaviors and pre-defined audio for avatar) within a short temporal window, jointly driving generation of avatar's audio-aligned lip articulation and non-verbal FBs. However, two main challenges persist in these methods: (i) the reliance on short-clip behavioral cues without long-range contextual modeling leads them to produce facial behaviors lacking contextual appropriateness; and (ii) the entangled, role-agnostic fusion of dual-track signals empirically introduces cross-signal interference, potentially compromising lip-region synchronization during speaking. To this end, we propose ECHO, a novel IHG framework comprising two key components: a Long-range Contextual Understanding (LCU) component that facilitates contextual understanding of both behavior-grounded dynamics and linguistic-driven affective semantics to promote contextual appropriateness and emotional rationality of synthesized avatar FBs; and a block-wise Spatial-aware Decoupled Cross-attention Modulation (SDCM) module, that preserves self-audio-driven lip articulation while adaptively integrating user contextual behavioral cues for non-lip facial regions, complemented by our designed two-stage training paradigm, to jointly enhance lip synchronization and visual fidelity. Extensive experiments demonstrate the effectiveness of proposed components and ECHO's superior IHG performance.

Zhengcong Fei, Shuman Tian, Junshi Huang et al.

Ensemble of machine learning models yields improved performance as well as robustness. However, their memory requirements and inference costs can be prohibitively high. Knowledge distillation is an approach that allows a single model to efficiently capture the approximate performance of an ensemble while showing poor scalability as demand for re-training when introducing new teacher models. In this paper, we study if we can utilize the meta-learning strategy to directly predict the parameters of a single model with comparable performance of an ensemble. Hereto, we introduce WeightFormer, a Transformer-based model that can predict student network weights layer by layer in a forward pass, according to the teacher model parameters. The proprieties of WeightFormer are investigated on the CIFAR-10, CIFAR-100, and ImageNet datasets for model structures of VGGNet-11, ResNet-50, and ViT-B/32, where it demonstrates that our method can achieve approximate classification performance of an ensemble and outperforms both the single network and standard knowledge distillation. More encouragingly, we show that WeightFormer results can further exceeds average ensemble with minor fine-tuning. Importantly, our task along with the model and results can potentially lead to a new, more efficient, and scalable paradigm of ensemble networks parameter learning.

Hongyu Li, Tianrui Hui, Zihan Ding et al.

Panoptic narrative grounding (PNG), whose core target is fine-grained image-text alignment, requires a panoptic segmentation of referred objects given a narrative caption. Previous discriminative methods achieve only weak or coarse-grained alignment by panoptic segmentation pretraining or CLIP model adaptation. Given the recent progress of text-to-image Diffusion models, several works have shown their capability to achieve fine-grained image-text alignment through cross-attention maps and improved general segmentation performance. However, the direct use of phrase features as static prompts to apply frozen Diffusion models to the PNG task still suffers from a large task gap and insufficient vision-language interaction, yielding inferior performance. Therefore, we propose an Extractive-Injective Phrase Adapter (EIPA) bypass within the Diffusion UNet to dynamically update phrase prompts with image features and inject the multimodal cues back, which leverages the fine-grained image-text alignment capability of Diffusion models more sufficiently. In addition, we also design a Multi-Level Mutual Aggregation (MLMA) module to reciprocally fuse multi-level image and phrase features for segmentation refinement. Extensive experiments on the PNG benchmark show that our method achieves new state-of-the-art performance.

Xiangyu Liu, Feng Gao, Xiaomei Zhang et al.

Existing audio-driven video digital human generation models rely on multi-step denoising, resulting in substantial computational overhead that severely limits their deployment in real-world settings. While one-step distillation approaches can significantly accelerate inference, they often suffer from training instability. To address this challenge, we propose TurboTalk, a two-stage progressive distillation framework that effectively compresses a multi-step audio-driven video diffusion model into a single-step generator. We first adopt Distribution Matching Distillation to obtain a strong and stable 4-step student, and then progressively reduce the denoising steps from 4 to 1 through adversarial distillation. To ensure stable training under extreme step reduction, we introduce a progressive timestep sampling strategy and a self-compare adversarial objective that provides an intermediate adversarial reference that stabilizes progressive distillation. Our method achieve single-step generation of video talking avatar, boosting inference speed by 120 times while maintaining high generation quality.

Hongyu Li, Jinyu Chen, Ziyu Wei et al.

Recent advancements in multimodal large language models (MLLMs) have shown promising results, yet existing approaches struggle to effectively handle both temporal and spatial localization simultaneously. This challenge stems from two key issues: first, incorporating spatial-temporal localization introduces a vast number of coordinate combinations, complicating the alignment of linguistic and visual coordinate representations; second, encoding fine-grained temporal and spatial information during video feature compression is inherently difficult. To address these issues, we propose LLaVA-ST, a MLLM for fine-grained spatial-temporal multimodal understanding. In LLaVA-ST, we propose Language-Aligned Positional Embedding, which embeds the textual coordinate special token into the visual space, simplifying the alignment of fine-grained spatial-temporal correspondences. Additionally, we design the Spatial-Temporal Packer, which decouples the feature compression of temporal and spatial resolutions into two distinct point-to-region attention processing streams. Furthermore, we propose ST-Align dataset with 4.3M training samples for fine-grained spatial-temporal multimodal understanding. With ST-align, we present a progressive training pipeline that aligns the visual and textual feature through sequential coarse-to-fine stages.Additionally, we introduce an ST-Align benchmark to evaluate spatial-temporal interleaved fine-grained understanding tasks, which include Spatial-Temporal Video Grounding (STVG) , Event Localization and Captioning (ELC) and Spatial Video Grounding (SVG). LLaVA-ST achieves outstanding performance on 11 benchmarks requiring fine-grained temporal, spatial, or spatial-temporal interleaving multimodal understanding. Our code, data and benchmark will be released at Our code, data and benchmark will be released at https://github.com/appletea233/LLaVA-ST .

Chen Duan, Pei Fu, Shan Guo et al.

In recent years, text-image joint pre-training techniques have shown promising results in various tasks. However, in Optical Character Recognition (OCR) tasks, aligning text instances with their corresponding text regions in images poses a challenge, as it requires effective alignment between text and OCR-Text (referring to the text in images as OCR-Text to distinguish from the text in natural language) rather than a holistic understanding of the overall image content. In this paper, we propose a new pre-training method called OCR-Text Destylization Modeling (ODM) that transfers diverse styles of text found in images to a uniform style based on the text prompt. With ODM, we achieve better alignment between text and OCR-Text and enable pre-trained models to adapt to the complex and diverse styles of scene text detection and spotting tasks. Additionally, we have designed a new labeling generation method specifically for ODM and combined it with our proposed Text-Controller module to address the challenge of annotation costs in OCR tasks, allowing a larger amount of unlabeled data to participate in pre-training. Extensive experiments on multiple public datasets demonstrate that our method significantly improves performance and outperforms current pre-training methods in scene text detection and spotting tasks. Code is available at https://github.com/PriNing/ODM.

Hewen Pan, Cong Wei, Dashuang Liang et al.

With the advancement of multi-modal Large Language Models (LLMs), Video LLMs have been further developed to perform on holistic and specialized video understanding. However, existing works are limited to specialized video understanding tasks, failing to achieve a comprehensive and multi-grained video perception. To bridge this gap, we introduce UFVideo, the first Video LLM with unified multi-grained cooperative understanding capabilities. Specifically, we design unified visual-language guided alignment to flexibly handle video understanding across global, pixel and temporal scales within a single model. UFVideo dynamically encodes the visual and text inputs of different tasks and generates the textual response, temporal localization, or grounded mask. Additionally, to evaluate challenging multi-grained video understanding tasks, we construct the UFVideo-Bench consisting of three distinct collaborative tasks within the scales, which demonstrates UFVideo's flexibility and advantages over GPT-4o. Furthermore, we validate the effectiveness of our model across 9 public benchmarks covering various common video understanding tasks, providing valuable insights for future Video LLMs.

Ruiqi Wu, Xuanhua He, Meng Cheng et al.

We propose Infinite-World, a robust interactive world model capable of maintaining coherent visual memory over 1000+ frames in complex real-world environments. While existing world models can be efficiently optimized on synthetic data with perfect ground-truth, they lack an effective training paradigm for real-world videos due to noisy pose estimations and the scarcity of viewpoint revisits. To bridge this gap, we first introduce a Hierarchical Pose-free Memory Compressor (HPMC) that recursively distills historical latents into a fixed-budget representation. By jointly optimizing the compressor with the generative backbone, HPMC enables the model to autonomously anchor generations in the distant past with bounded computational cost, eliminating the need for explicit geometric priors. Second, we propose an Uncertainty-aware Action Labeling module that discretizes continuous motion into a tri-state logic. This strategy maximizes the utilization of raw video data while shielding the deterministic action space from being corrupted by noisy trajectories, ensuring robust action-response learning. Furthermore, guided by insights from a pilot toy study, we employ a Revisit-Dense Finetuning Strategy using a compact, 30-minute dataset to efficiently activate the model's long-range loop-closure capabilities. Extensive experiments, including objective metrics and user studies, demonstrate that Infinite-World achieves superior performance in visual quality, action controllability, and spatial consistency.

SiXiang Chen, Jianyu Lai, Jialin Gao et al.

Generating aesthetic posters is more challenging than simple design images: it requires not only precise text rendering but also the seamless integration of abstract artistic content, striking layouts, and overall stylistic harmony. To address this, we propose PosterCraft, a unified framework that abandons prior modular pipelines and rigid, predefined layouts, allowing the model to freely explore coherent, visually compelling compositions. PosterCraft employs a carefully designed, cascaded workflow to optimize the generation of high-aesthetic posters: (i) large-scale text-rendering optimization on our newly introduced Text-Render-2M dataset; (ii) region-aware supervised fine-tuning on HQ-Poster100K; (iii) aesthetic-text-reinforcement learning via best-of-n preference optimization; and (iv) joint vision-language feedback refinement. Each stage is supported by a fully automated data-construction pipeline tailored to its specific needs, enabling robust training without complex architectural modifications. Evaluated on multiple experiments, PosterCraft significantly outperforms open-source baselines in rendering accuracy, layout coherence, and overall visual appeal-approaching the quality of SOTA commercial systems. Our code, models, and datasets can be found in the Project page: https://ephemeral182.github.io/PosterCraft

Xuanhua He, Tianyu Yang, Ke Cao et al.

Current video avatar generation methods excel at identity preservation and motion alignment but lack genuine agency, they cannot autonomously pursue long-term goals through adaptive environmental interaction. We address this by introducing L-IVA (Long-horizon Interactive Visual Avatar), a task and benchmark for evaluating goal-directed planning in stochastic generative environments, and ORCA (Online Reasoning and Cognitive Architecture), the first framework enabling active intelligence in video avatars. ORCA embodies Internal World Model (IWM) capabilities through two key innovations: (1) a closed-loop OTAR cycle (Observe-Think-Act-Reflect) that maintains robust state tracking under generative uncertainty by continuously verifying predicted outcomes against actual generations, and (2) a hierarchical dual-system architecture where System 2 performs strategic reasoning with state prediction while System 1 translates abstract plans into precise, model-specific action captions. By formulating avatar control as a POMDP and implementing continuous belief updating with outcome verification, ORCA enables autonomous multi-step task completion in open-domain scenarios. Extensive experiments demonstrate that ORCA significantly outperforms open-loop and non-reflective baselines in task success rate and behavioral coherence, validating our IWM-inspired design for advancing video avatar intelligence from passive animation to active, goal-oriented behavior.

Haonan Han, Jiancheng Huang, Xiaopeng Sun et al.

Beneath the stunning visual fidelity of modern AIGC models lies a "logical desert", where systems fail tasks that require physical, causal, or complex spatial reasoning. Current evaluations largely rely on superficial metrics or fragmented benchmarks, creating a ``performance mirage'' that overlooks the generative process. To address this, we introduce ViGoR Vision-G}nerative Reasoning-centric Benchmark), a unified framework designed to dismantle this mirage. ViGoR distinguishes itself through four key innovations: 1) holistic cross-modal coverage bridging Image-to-Image and Video tasks; 2) a dual-track mechanism evaluating both intermediate processes and final results; 3) an evidence-grounded automated judge ensuring high human alignment; and 4) granular diagnostic analysis that decomposes performance into fine-grained cognitive dimensions. Experiments on over 20 leading models reveal that even state-of-the-art systems harbor significant reasoning deficits, establishing ViGoR as a critical ``stress test'' for the next generation of intelligent vision models. The demo have been available at https://vincenthancoder.github.io/ViGoR-Bench/

Xiaoping Wu, Jie Hu, Xiaoming Wei

Diffusion Probabilistic Models (DPMs) have emerged as the de facto approach for high-fidelity image synthesis, operating diffusion processes on continuous VAE latent, which significantly differ from the text generation methods employed by Large Language Models (LLMs). In this paper, we introduce a novel generative framework, the Recurrent Diffusion Probabilistic Model (RDPM), which enhances the diffusion process through a recurrent token prediction mechanism, thereby pioneering the field of Discrete Diffusion. By progressively introducing Gaussian noise into the latent representations of images and encoding them into vector-quantized tokens in a recurrent manner, RDPM facilitates a unique diffusion process on discrete-value domains. This process iteratively predicts the token codes for subsequent timesteps, transforming the initial standard Gaussian noise into the source data distribution, aligning with GPT-style models in terms of the loss function. RDPM demonstrates superior performance while benefiting from the speed advantage of requiring only a few inference steps. This model not only leverages the diffusion process to ensure high-quality generation but also converts continuous signals into a series of high-fidelity discrete tokens, thereby maintaining a unified optimization strategy with other discrete tokens, such as text. We anticipate that this work will contribute to the development of a unified model for multimodal generation, specifically by integrating continuous signal domains such as images, videos, and audio with text. We will release the code and model weights to the open-source community.

Chen Zhao, Jiawei Chen, Hongyu Li et al.

Recent advances in video diffusion models have significantly improved visual quality, yet ultra-high-resolution (UHR) video generation remains a formidable challenge due to the compounded difficulties of motion modeling, semantic planning, and detail synthesis. To address these limitations, we propose \textbf{LUVE}, a \textbf{L}atent-cascaded \textbf{U}HR \textbf{V}ideo generation framework built upon dual frequency \textbf{E}xperts. LUVE employs a three-stage architecture comprising low-resolution motion generation for motion-consistent latent synthesis, video latent upsampling that performs resolution upsampling directly in the latent space to mitigate memory and computational overhead, and high-resolution content refinement that integrates low-frequency and high-frequency experts to jointly enhance semantic coherence and fine-grained detail generation. Extensive experiments demonstrate that our LUVE achieves superior photorealism and content fidelity in UHR video generation, and comprehensive ablation studies further validate the effectiveness of each component. The project is available at \href{https://unicornanrocinu.github.io/LUVE_web/}{https://github.io/LUVE/}.

Dengsheng Chen, Jie Hu, Xiaoming Wei et al.

Incorporating a temporal dimension into pretrained image diffusion models for video generation is a prevalent approach. However, this method is computationally demanding and necessitates large-scale video datasets. More critically, the heterogeneity between image and video datasets often results in catastrophic forgetting of the image expertise. Recent attempts to directly extract video snippets from image diffusion models have somewhat mitigated these problems. Nevertheless, these methods can only generate brief video clips with simple movements and fail to capture fine-grained motion or non-grid deformation. In this paper, we propose a novel Zero-Shot video Sampling algorithm, denoted as $\mathcal{ZS}^2$, capable of directly sampling high-quality video clips from existing image synthesis methods, such as Stable Diffusion, without any training or optimization. Specifically, $\mathcal{ZS}^2$ utilizes the dependency noise model and temporal momentum attention to ensure content consistency and animation coherence, respectively. This ability enables it to excel in related tasks, such as conditional and context-specialized video generation and instruction-guided video editing. Experimental results demonstrate that $\mathcal{ZS}^2$ achieves state-of-the-art performance in zero-shot video generation, occasionally outperforming recent supervised methods. Homepage: \url{https://densechen.github.io/zss/}.

Dengsheng Chen, Jie Hu, Xiaoming Wei et al.

The Gaussian diffusion model, initially designed for image generation, has recently been adapted for 3D point cloud generation. However, these adaptations have not fully considered the intrinsic geometric characteristics of 3D shapes, thereby constraining the diffusion model's potential for 3D shape manipulation. To address this limitation, we introduce a novel deformable 3D shape diffusion model that facilitates comprehensive 3D shape manipulation, including point cloud generation, mesh deformation, and facial animation. Our approach innovatively incorporates a differential deformation kernel, which deconstructs the generation of geometric structures into successive non-rigid deformation stages. By leveraging a probabilistic diffusion model to simulate this step-by-step process, our method provides a versatile and efficient solution for a wide range of applications, spanning from graphics rendering to facial expression animation. Empirical evidence highlights the effectiveness of our approach, demonstrating state-of-the-art performance in point cloud generation and competitive results in mesh deformation. Additionally, extensive visual demonstrations reveal the significant potential of our approach for practical applications. Our method presents a unique pathway for advancing 3D shape manipulation and unlocking new opportunities in the realm of virtual reality.

Youxin Pang, Jiajun Liu, Lingfeng Tan et al.

We propose MAViD, a novel Multimodal framework for Audio-Visual Dialogue understanding and generation. Existing approaches primarily focus on non-interactive systems and are limited to producing constrained and unnatural human speech.The primary challenge of this task lies in effectively integrating understanding and generation capabilities, as well as achieving seamless multimodal audio-video fusion. To solve these problems, we propose a Conductor-Creator architecture that divides the dialogue system into two primary components.The Conductor is tasked with understanding, reasoning, and generating instructions by breaking them down into motion and speech components, thereby enabling fine-grained control over interactions. The Creator then delivers interactive responses based on these instructions.Furthermore, to address the difficulty of generating long videos with consistent identity, timbre, and tone using dual DiT structures, the Creator adopts a structure that combines autoregressive (AR) and diffusion models. The AR model is responsible for audio generation, while the diffusion model ensures high-quality video generation.Additionally, we propose a novel fusion module to enhance connections between contextually consecutive clips and modalities, enabling synchronized long-duration audio-visual content generation.Extensive experiments demonstrate that our framework can generate vivid and contextually coherent long-duration dialogue interactions and accurately interpret users' multimodal queries.

Youxin Pang, Yong Zhang, Ruizhi Shao et al.

We propose UniMo, an innovative autoregressive model for joint modeling of 2D human videos and 3D human motions within a unified framework, enabling simultaneous generation and understanding of these two modalities for the first time. Current methods predominantly focus on generating one modality given another as the condition or integrating either of them with other modalities such as text and audio. Unifying 2D videos and 3D motions for simultaneous optimization and generation remains largely unexplored, presenting significant challenges due to their substantial structural and distributional differences. Inspired by the LLM's ability to unify different modalities, our method models videos and 3D motions as a unified tokens sequence, utilizing separate embedding layers to mitigate distribution gaps. Additionally, we devise a sequence modeling strategy that integrates two distinct tasks within a single framework, proving the effectiveness of unified modeling. Moreover, to efficiently align with visual tokens and preserve 3D spatial information, we design a novel 3D motion tokenizer with a temporal expansion strategy, using a single VQ-VAE to produce quantized motion tokens. It features multiple expert decoders that handle body shapes, translation, global orientation, and body poses for reliable 3D motion reconstruction. Extensive experiments demonstrate that our method simultaneously generates corresponding videos and motions while performing accurate motion capture. This work taps into the capacity of LLMs to fuse diverse data types, paving the way for integrating human-centric information into existing models and potentially enabling multimodal, controllable joint modeling of humans, objects, and scenes.

Meituan LongCat Team, Xunliang Cai, Qilong Huang et al.

Video generation is a critical pathway toward world models, with efficient long video inference as a key capability. Toward this end, we introduce LongCat-Video, a foundational video generation model with 13.6B parameters, delivering strong performance across multiple video generation tasks. It particularly excels in efficient and high-quality long video generation, representing our first step toward world models. Key features include: Unified architecture for multiple tasks: Built on the Diffusion Transformer (DiT) framework, LongCat-Video supports Text-to-Video, Image-to-Video, and Video-Continuation tasks with a single model; Long video generation: Pretraining on Video-Continuation tasks enables LongCat-Video to maintain high quality and temporal coherence in the generation of minutes-long videos; Efficient inference: LongCat-Video generates 720p, 30fps videos within minutes by employing a coarse-to-fine generation strategy along both the temporal and spatial axes. Block Sparse Attention further enhances efficiency, particularly at high resolutions; Strong performance with multi-reward RLHF: Multi-reward RLHF training enables LongCat-Video to achieve performance on par with the latest closed-source and leading open-source models. Code and model weights are publicly available to accelerate progress in the field.

Zhuo Li, Weiqing Min, Jiajun Song et al.

Optimizing the approximation of Average Precision (AP) has been widely studied for image retrieval. Limited by the definition of AP, such methods consider both negative and positive instances ranking before each positive instance. However, we claim that only penalizing negative instances before positive ones is enough, because the loss only comes from these negative instances. To this end, we propose a novel loss, namely Penalizing Negative instances before Positive ones (PNP), which can directly minimize the number of negative instances before each positive one. In addition, AP-based methods adopt a fixed and sub-optimal gradient assignment strategy. Therefore, we systematically investigate different gradient assignment solutions via constructing derivative functions of the loss, resulting in PNP-I with increasing derivative functions and PNP-D with decreasing ones. PNP-I focuses more on the hard positive instances by assigning larger gradients to them and tries to make all relevant instances closer. In contrast, PNP-D pays less attention to such instances and slowly corrects them. For most real-world data, one class usually contains several local clusters. PNP-I blindly gathers these clusters while PNP-D keeps them as they were. Therefore, PNP-D is more superior. Experiments on three standard retrieval datasets show consistent results with the above analysis. Extensive evaluations demonstrate that PNP-D achieves the state-of-the-art performance. Code is available at https://github.com/interestingzhuo/PNPloss

Zhe Kong, Feng Gao, Yong Zhang et al.

Audio-driven human animation methods, such as talking head and talking body generation, have made remarkable progress in generating synchronized facial movements and appealing visual quality videos. However, existing methods primarily focus on single human animation and struggle with multi-stream audio inputs, facing incorrect binding problems between audio and persons. Additionally, they exhibit limitations in instruction-following capabilities. To solve this problem, in this paper, we propose a novel task: Multi-Person Conversational Video Generation, and introduce a new framework, MultiTalk, to address the challenges during multi-person generation. Specifically, for audio injection, we investigate several schemes and propose the Label Rotary Position Embedding (L-RoPE) method to resolve the audio and person binding problem. Furthermore, during training, we observe that partial parameter training and multi-task training are crucial for preserving the instruction-following ability of the base model. MultiTalk achieves superior performance compared to other methods on several datasets, including talking head, talking body, and multi-person datasets, demonstrating the powerful generation capabilities of our approach.

Hongwei Zheng, Linyuan Zhou, Han Li et al.

Data mixing methods play a crucial role in semi-supervised learning (SSL), but their application is unexplored in long-tailed semi-supervised learning (LTSSL). The primary reason is that the in-batch mixing manner fails to address class imbalance. Furthermore, existing LTSSL methods mainly focus on re-balancing data quantity but ignore class-wise uncertainty, which is also vital for class balance. For instance, some classes with sufficient samples might still exhibit high uncertainty due to indistinguishable features. To this end, this paper introduces the Balanced and Entropy-based Mix (BEM), a pioneering mixing approach to re-balance the class distribution of both data quantity and uncertainty. Specifically, we first propose a class balanced mix bank to store data of each class for mixing. This bank samples data based on the estimated quantity distribution, thus re-balancing data quantity. Then, we present an entropy-based learning approach to re-balance class-wise uncertainty, including entropy-based sampling strategy, entropy-based selection module, and entropy-based class balanced loss. Our BEM first leverages data mixing for improving LTSSL, and it can also serve as a complement to the existing re-balancing methods. Experimental results show that BEM significantly enhances various LTSSL frameworks and achieves state-of-the-art performances across multiple benchmarks.

Lichen Ma, Tiezhu Yue, Pei Fu et al.

Recently, significant advancements have been made in diffusion-based visual text generation models. Although the effectiveness of these methods in visual text rendering is rapidly improving, they still encounter challenges such as inaccurate characters and strokes when rendering complex visual text. In this paper, we propose CharGen, a highly accurate character-level visual text generation and editing model. Specifically, CharGen employs a character-level multimodal encoder that not only extracts character-level text embeddings but also encodes glyph images character by character. This enables it to capture fine-grained cross-modality features more effectively. Additionally, we introduce a new perceptual loss in CharGen to enhance character shape supervision and address the issue of inaccurate strokes in generated text. It is worth mentioning that CharGen can be integrated into existing diffusion models to generate visual text with high accuracy. CharGen significantly improves text rendering accuracy, outperforming recent methods in public benchmarks such as AnyText-benchmark and MARIO-Eval, with improvements of more than 8% and 6%, respectively. Notably, CharGen achieved a 5.5% increase in accuracy on Chinese test sets.

Huijie Liu, Jingyun Wang, Shuai Ma et al.

Motion customization aims to adapt the diffusion model (DM) to generate videos with the motion specified by a set of video clips with the same motion concept. To realize this goal, the adaptation of DM should be possible to model the specified motion concept, without compromising the ability to generate diverse appearances. Thus, the key to solving this problem lies in how to separate the motion concept from the appearance in the adaptation process of DM. Typical previous works explore different ways to represent and insert a motion concept into large-scale pretrained text-to-video diffusion models, e.g., learning a motion LoRA, using latent noise residuals, etc. While those methods can encode the motion concept, they also inevitably encode the appearance in the reference videos, resulting in weakened appearance generation capability. In this paper, we follow the typical way to learn a motion LoRA to encode the motion concept, but propose two novel strategies to enhance motion-appearance separation, including temporal attention purification (TAP) and appearance highway (AH). Specifically, we assume that in the temporal attention module, the pretrained Value embeddings are sufficient to serve as basic components needed by producing a new motion. Thus, in TAP, we choose only to reshape the temporal attention with motion LoRAs so that Value embeddings can be reorganized to produce a new motion. Further, in AH, we alter the starting point of each skip connection in U-Net from the output of each temporal attention module to the output of each spatial attention module. Extensive experiments demonstrate that compared to previous works, our method can generate videos with appearance more aligned with the text descriptions and motion more consistent with the reference videos.

Xiuyuan Guo, Chengqi Xu, Guinan Guo et al.

Currently, training large-scale deep learning models is typically achieved through parallel training across multiple GPUs. However, due to the inherent communication overhead and synchronization delays in traditional model parallelism methods, seamless parallel training cannot be achieved, which, to some extent, affects overall training efficiency. To address this issue, we present PPLL (Pipeline Parallelism based on Local Learning), a novel framework that leverages local learning algorithms to enable effective parallel training across multiple GPUs. PPLL divides the model into several distinct blocks, each allocated to a separate GPU. By utilizing queues to manage data transfers between GPUs, PPLL ensures seamless cross-GPU communication, allowing multiple blocks to execute forward and backward passes in a pipelined manner. This design minimizes idle times and prevents bottlenecks typically caused by sequential gradient updates, thereby accelerating the overall training process. We validate PPLL through extensive experiments using ResNet and Vision Transformer (ViT) architectures on CIFAR-10, SVHN, and STL-10 datasets. Our results demonstrate that PPLL significantly enhances the training speed of the local learning method while achieving comparable or even superior training speed to traditional pipeline parallelism (PP) without sacrificing model performance. In a 4-GPU training setup, PPLL accelerated local learning training on ViT and ResNet by 162% and 33%, respectively, achieving 1.25x and 0.85x the speed of traditional pipeline parallelism.

Junhao Su, Feiyu Zhu, Hengyu Shi et al.

Deep learning typically relies on end-to-end backpropagation for training, a method that inherently suffers from issues such as update locking during parameter optimization, high GPU memory consumption, and a lack of biological plausibility. In contrast, supervised local learning seeks to mitigate these challenges by partitioning the network into multiple local blocks and designing independent auxiliary networks to update each block separately. However, because gradients are propagated solely within individual local blocks, performance degradation occurs, preventing supervised local learning from supplanting end-to-end backpropagation. To address these limitations and facilitate inter-block information flow, we propose the Momentum Auxiliary Network++ (MAN++). MAN++ introduces a dynamic interaction mechanism by employing the Exponential Moving Average (EMA) of parameters from adjacent blocks to enhance communication across the network. The auxiliary network, updated via EMA, effectively bridges the information gap between blocks. Notably, we observed that directly applying EMA parameters can be suboptimal due to feature discrepancies between local blocks. To resolve this issue, we introduce a learnable scaling bias that balances feature differences, thereby further improving performance. We validate MAN++ through extensive experiments on tasks that include image classification, object detection, and image segmentation, utilizing multiple network architectures. The experimental results demonstrate that MAN++ achieves performance comparable to end-to-end training while significantly reducing GPU memory usage. Consequently, MAN++ offers a novel perspective for supervised local learning and presents a viable alternative to conventional training methods.

Junjie Hu, Tianyang Han, Kai Ma et al.

Recent subject-driven image customization has achieved significant advancements in fidelity, yet fine-grained instance-level spatial control remains elusive, hindering broader real-world application. This limitation is mainly attributed to the absence of scalable datasets that bind identity with precise positional cues. To this end, we introduce PositionIC, a unified framework that enforces position and identity consistency for multi-subject customization. We construct a scalable synthesis pipeline that employs a bidirectional generation paradigm to eliminate subject drift and maintain semantic coherence. On top of these data, we design a lightweight positional modulation operation that decouples spatial embeddings among subjects, enabling independent, accurate placement while preserving visual fidelity. Extensive experiments demonstrate that our approach can achieve precise spatial control while maintaining high consistency in image customization tasks. PositionIC paves the way for controllable, high-fidelity image customization in open-world, multi-entity scenarios and will be released to foster further research.

Zhe Kong, Le Li, Yong Zhang et al.

Real-world video super-resolution (VSR) presents significant challenges due to complex and unpredictable degradations. Although some recent methods utilize image diffusion models for VSR and have shown improved detail generation capabilities, they still struggle to produce temporally consistent frames. We attempt to use Stable Video Diffusion (SVD) combined with ControlNet to address this issue. However, due to the intrinsic image-animation characteristics of SVD, it is challenging to generate fine details using only low-quality videos. To tackle this problem, we propose DAM-VSR, an appearance and motion disentanglement framework for VSR. This framework disentangles VSR into appearance enhancement and motion control problems. Specifically, appearance enhancement is achieved through reference image super-resolution, while motion control is achieved through video ControlNet. This disentanglement fully leverages the generative prior of video diffusion models and the detail generation capabilities of image super-resolution models. Furthermore, equipped with the proposed motion-aligned bidirectional sampling strategy, DAM-VSR can conduct VSR on longer input videos. DAM-VSR achieves state-of-the-art performance on real-world data and AIGC data, demonstrating its powerful detail generation capabilities.

Huijie Liu, Bingcan Wang, Jie Hu et al.

Dish images play a crucial role in the digital era, with the demand for culturally distinctive dish images continuously increasing due to the digitization of the food industry and e-commerce. In general cases, existing text-to-image generation models excel in producing high-quality images; however, they struggle to capture diverse characteristics and faithful details of specific domains, particularly Chinese dishes. To address this limitation, we propose Omni-Dish, the first text-to-image generation model specifically tailored for Chinese dishes. We develop a comprehensive dish curation pipeline, building the largest dish dataset to date. Additionally, we introduce a recaption strategy and employ a coarse-to-fine training scheme to help the model better learn fine-grained culinary nuances. During inference, we enhance the user's textual input using a pre-constructed high-quality caption library and a large language model, enabling more photorealistic and faithful image generation. Furthermore, to extend our model's capability for dish editing tasks, we propose Concept-Enhanced P2P. Based on this approach, we build a dish editing dataset and train a specialized editing model. Extensive experiments demonstrate the superiority of our methods.

Dengsheng Chen, Jie Hu, Tiezhu Yue et al.

Recently, autoregressive models have demonstrated remarkable performance in class-conditional image generation. However, the application of next-token prediction to high-resolution text-to-image generation remains largely unexplored. In this paper, we introduce \textbf{D-JEPA$\cdot$T2I}, an autoregressive model based on continuous tokens that incorporates innovations in both architecture and training strategy to generate high-quality, photorealistic images at arbitrary resolutions, up to 4K. Architecturally, we adopt the denoising joint embedding predictive architecture (D-JEPA) while leveraging a multimodal visual transformer to effectively integrate textual and visual features. Additionally, we introduce flow matching loss alongside the proposed Visual Rotary Positional Embedding (VoPE) to enable continuous resolution learning. In terms of training strategy, we propose a data feedback mechanism that dynamically adjusts the sampling procedure based on statistical analysis and an online learning critic model. This encourages the model to move beyond its comfort zone, reducing redundant training on well-mastered scenarios and compelling it to address more challenging cases with suboptimal generation quality. For the first time, we achieve state-of-the-art high-resolution image synthesis via next-token prediction.

Jianyu Lai, Sixiang Chen, Jialin Gao et al.

Recent advancements in the text-rendering capabilities of image generation models have made the end-to-end creation of graphic design content, such as posters, increasingly feasible. However, existing reward models fall short of accurately assessing design quality, as they primarily focus on global image aesthetics while overlooking the critical dimensions of typography and layout. Furthermore, the scarcity of domain-specific preference data remains a significant bottleneck, which limits the further development of graphic design evaluation and generation. To bridge this gap, we introduce an automated pipeline to construct a high-quality dataset of 70k poster preferences by leveraging the consensus of multiple Multi-modal Large Language Models (MLLMs) to simulate human-like judgment. Utilizing this dataset, we develop PosterReward, a reward model specifically designed for high-precision poster assessment through a cascaded, multi-stage training strategy. We also provide multiple variants of the model to cater to different application scenarios. Finally, we introduce PosterRewardBench and PosterBench to evaluate the performance of existing reward models in poster assessment and the generation capabilities of current text-to-image models in poster creation, respectively.

Jia Wang, Jie Hu, Xiaoqi Ma et al.

Text-to-image (T2I) generation has achieved remarkable progress in instruction following and aesthetics. However, a persistent challenge is the prevalence of physical artifacts, such as anatomical and structural flaws, which severely degrade perceptual quality and limit application. Given the diversity and complexity of these artifacts, a systematic and fine-grained evaluation framework is required, which is lacking in current benchmarks. To fill this gap, we introduce MagicMirror, a comprehensive framework for artifacts assessment. We first establish a detailed taxonomy of generated image artifacts. Guided by this taxonomy, we manually annotate MagicData340K, the first human-annotated large-scale dataset of 340K generated images with fine-grained artifact labels. Building on this dataset, we train MagicAssessor, a Vision-Language Model (VLM) that provides detailed assessments and corresponding labels. To overcome challenges like class imbalance and reward hacking, we design a novel data sampling strategy and a multi-level reward system for Group Relative Policy Optimization (GRPO). Finally, we leverage MagicAssessor to construct MagicBench, an automated benchmark for evaluating the image artifacts of current T2I models. Our evaluation with MagicBench reveals that despite their widespread adoption, even top-tier models like GPT-image-1 are consistently plagued by significant artifacts, highlighting artifact reduction as a critical frontier for future T2I development. Project page: https://wj-inf.github.io/MagicMirror-page/.

Shaoshu Yang, Zhe Kong, Feng Gao et al.

Recent breakthroughs in video AIGC have ushered in a transformative era for audio-driven human animation. However, conventional video dubbing techniques remain constrained to mouth region editing, resulting in discordant facial expressions and body gestures that compromise viewer immersion. To overcome this limitation, we introduce sparse-frame video dubbing, a novel paradigm that strategically preserves reference keyframes to maintain identity, iconic gestures, and camera trajectories while enabling holistic, audio-synchronized full-body motion editing. Through critical analysis, we identify why naive image-to-video models fail in this task, particularly their inability to achieve adaptive conditioning. Addressing this, we propose InfiniteTalk, a streaming audio-driven generator designed for infinite-length long sequence dubbing. This architecture leverages temporal context frames for seamless inter-chunk transitions and incorporates a simple yet effective sampling strategy that optimizes control strength via fine-grained reference frame positioning. Comprehensive evaluations on HDTF, CelebV-HQ, and EMTD datasets demonstrate state-of-the-art performance. Quantitative metrics confirm superior visual realism, emotional coherence, and full-body motion synchronization.

Ying Guo, Xi Liu, Cheng Zhen et al.

Face-to-face communication, as a common human activity, motivates the research on interactive head generation. A virtual agent can generate motion responses with both listening and speaking capabilities based on the audio or motion signals of the other user and itself. However, previous clip-wise generation paradigm or explicit listener/speaker generator-switching methods have limitations in future signal acquisition, contextual behavioral understanding, and switching smoothness, making it challenging to be real-time and realistic. In this paper, we propose an autoregressive (AR) based frame-wise framework called ARIG to realize the real-time generation with better interaction realism. To achieve real-time generation, we model motion prediction as a non-vector-quantized AR process. Unlike discrete codebook-index prediction, we represent motion distribution using diffusion procedure, achieving more accurate predictions in continuous space. To improve interaction realism, we emphasize interactive behavior understanding (IBU) and detailed conversational state understanding (CSU). In IBU, based on dual-track dual-modal signals, we summarize short-range behaviors through bidirectional-integrated learning and perform contextual understanding over long ranges. In CSU, we use voice activity signals and context features of IBU to understand the various states (interruption, feedback, pause, etc.) that exist in actual conversations. These serve as conditions for the final progressive motion prediction. Extensive experiments have verified the effectiveness of our model.

Jinming Su, Chao Chen, Ke Zhang et al.

Recently, lane detection has made great progress with the rapid development of deep neural networks and autonomous driving. However, there exist three mainly problems including characterizing lanes, modeling the structural relationship between scenes and lanes, and supporting more attributes (e.g., instance and type) of lanes. In this paper, we propose a novel structure guided framework to solve these problems simultaneously. In the framework, we first introduce a new lane representation to characterize each instance. Then a topdown vanishing point guided anchoring mechanism is proposed to produce intensive anchors, which efficiently capture various lanes. Next, multi-level structural constraints are used to improve the perception of lanes. In the process, pixel-level perception with binary segmentation is introduced to promote features around anchors and restore lane details from bottom up, a lane-level relation is put forward to model structures (i.e., parallel) around lanes, and an image-level attention is used to adaptively attend different regions of the image from the perspective of scenes. With the help of structural guidance, anchors are effectively classified and regressed to obtain precise locations and shapes. Extensive experiments on public benchmark datasets show that the proposed approach outperforms state-of-the-art methods with 117 FPS on a single GPU.