Min Cao

Min Cao, Shiping Li, Juntao Li et al.

In the past few years, cross-modal image-text retrieval (ITR) has experienced increased interest in the research community due to its excellent research value and broad real-world application. It is designed for the scenarios where the queries are from one modality and the retrieval galleries from another modality. This paper presents a comprehensive and up-to-date survey on the ITR approaches from four perspectives. By dissecting an ITR system into two processes: feature extraction and feature alignment, we summarize the recent advance of the ITR approaches from these two perspectives. On top of this, the efficiency-focused study on the ITR system is introduced as the third perspective. To keep pace with the times, we also provide a pioneering overview of the cross-modal pre-training ITR approaches as the fourth perspective. Finally, we outline the common benchmark datasets and valuation metric for ITR, and conduct the accuracy comparison among the representative ITR approaches. Some critical yet less studied issues are discussed at the end of the paper.

Min Cao, Yang Bai, Ziyin Zeng et al.

Text-based Person Search (TBPS) aims to retrieve the person images using natural language descriptions. Recently, Contrastive Language Image Pretraining (CLIP), a universal large cross-modal vision-language pre-training model, has remarkably performed over various cross-modal downstream tasks due to its powerful cross-modal semantic learning capacity. TPBS, as a fine-grained cross-modal retrieval task, is also facing the rise of research on the CLIP-based TBPS. In order to explore the potential of the visual-language pre-training model for downstream TBPS tasks, this paper makes the first attempt to conduct a comprehensive empirical study of CLIP for TBPS and thus contribute a straightforward, incremental, yet strong TBPS-CLIP baseline to the TBPS community. We revisit critical design considerations under CLIP, including data augmentation and loss function. The model, with the aforementioned designs and practical training tricks, can attain satisfactory performance without any sophisticated modules. Also, we conduct the probing experiments of TBPS-CLIP in model generalization and model compression, demonstrating the effectiveness of TBPS-CLIP from various aspects. This work is expected to provide empirical insights and highlight future CLIP-based TBPS research.

Zhenfeng Fan, Zhiheng Zhang, Shuang Yang et al.

While impressive progress has recently been made in image-oriented facial attribute translation, shape-oriented 3D facial attribute translation remains an unsolved issue. This is primarily limited by the lack of 3D generative models and ineffective usage of 3D facial data. We propose a learning framework for 3D facial attribute translation to relieve these limitations. Firstly, we customize a novel geometric map for 3D shape representation and embed it in an end-to-end generative adversarial network. The geometric map represents 3D shapes symmetrically on a square image grid, while preserving the neighboring relationship of 3D vertices in a local least-square sense. This enables effective learning for the latent representation of data with different attributes. Secondly, we employ a unified and unpaired learning framework for multi-domain attribute translation. It not only makes effective usage of data correlation from multiple domains, but also mitigates the constraint for hardly accessible paired data. Finally, we propose a hierarchical architecture for the discriminator to guarantee robust results against both global and local artifacts. We conduct extensive experiments to demonstrate the advantage of the proposed framework over the state-of-the-art in generating high-fidelity facial shapes. Given an input 3D facial shape, the proposed framework is able to synthesize novel shapes of different attributes, which covers some downstream applications, such as expression transfer, gender translation, and aging. Code at https://github.com/NaughtyZZ/3D_facial_shape_attribute_translation_ssgmap.

William Leach, Ru He, Sizhuo Ma et al.

Vision-language models (VLMs) are increasingly attractive for multimodal quality assessment, but their default reliance on autoregressive text generation and dynamic visual processing is poorly matched to scalar regression under strict latency budgets. We present a bounded-compute adaptation of SmolVLM2-256M-Video-Instruct for product-rating prediction in the LoViF 2026 Efficient VLM challenge. Motivated by recent multimodal engagement-prediction results showing that feature-based regression can outperform token-based score generation, we replace the language-modeling head with a lightweight two-layer MLP fed by pooled decoder states, and we enforce deterministic inputs through fixed 384x384 images and truncated metadata. Across controlled ablations, static global image processing slightly outperforms dynamic tiling, and scaling from 100K to 16M training examples substantially improves validation correlation. Under the official held-out evaluation, our 228M-parameter model achieves 0.39 PLCC and 0.40 CES, providing a strong and reproducible baseline for resource-constrained multimodal regression.

Ning Liao, Bowen Shi, Xiaopeng Zhang et al.

Prompt learning has achieved great success in efficiently exploiting large-scale pre-trained models in natural language processing (NLP). It reformulates the downstream tasks as the generative pre-training ones to achieve consistency, thus improving the performance stably. However, when transferring it to the vision area, current visual prompt learning methods are almost designed on discriminative pre-trained models, and there is also a lack of careful design to unify the forms of pre-training and downstream tasks. To explore prompt learning on the generative pre-trained visual model, as well as keeping the task consistency, we propose Visual Prompt learning as masked visual Token Modeling (VPTM) to transform the downstream visual classification into the pre-trained masked visual token prediction. In addition, we develop the prototypical verbalizer for mapping the predicted visual token with implicit semantics to explicit downstream labels. To our best knowledge, VPTM is the first visual prompt method on the generative pre-trained visual model, which achieves consistency between pre-training and downstream visual classification by task reformulation. Experiments show that VPTM outperforms other visual prompt methods and achieves excellent efficiency. Moreover, the task consistency of VPTM contributes to the robustness against prompt location, prompt length and prototype dimension, and could be deployed uniformly.

Min Cao, Yang Bai, Jingyao Wang et al.

Under the flourishing development in performance, current image-text retrieval methods suffer from $N$-related time complexity, which hinders their application in practice. Targeting at efficiency improvement, this paper presents a simple and effective keyword-guided pre-screening framework for the image-text retrieval. Specifically, we convert the image and text data into the keywords and perform the keyword matching across modalities to exclude a large number of irrelevant gallery samples prior to the retrieval network. For the keyword prediction, we transfer it into a multi-label classification problem and propose a multi-task learning scheme by appending the multi-label classifiers to the image-text retrieval network to achieve a lightweight and high-performance keyword prediction. For the keyword matching, we introduce the inverted index in the search engine and create a win-win situation on both time and space complexities for the pre-screening. Extensive experiments on two widely-used datasets, i.e., Flickr30K and MS-COCO, verify the effectiveness of the proposed framework. The proposed framework equipped with only two embedding layers achieves $O(1)$ querying time complexity, while improving the retrieval efficiency and keeping its performance, when applied prior to the common image-text retrieval methods. Our code will be released.

Qi Lv, Ziqiang Cao, Wenrui Xie et al. · tencent-ai

With the tremendously increasing number of videos, there is a great demand for techniques that help people quickly navigate to the video segments they are interested in. However, current works on video understanding mainly focus on video content summarization, while little effort has been made to explore the structure of a video. Inspired by textual outline generation, we introduce a novel video understanding task, namely video outline generation (VOG). This task is defined to contain two sub-tasks: (1) first segmenting the video according to the content structure and then (2) generating a heading for each segment. To learn and evaluate VOG, we annotate a 10k+ dataset, called DuVOG. Specifically, we use OCR tools to recognize subtitles of videos. Then annotators are asked to divide subtitles into chapters and title each chapter. In videos, highlighted text tends to be the headline since it is more likely to attract attention. Therefore we propose a Visual Subtitle feature Enhanced video outline generation model (VSENet) which takes as input the textual subtitles together with their visual font sizes and positions. We consider the VOG task as a sequence tagging problem that extracts spans where the headings are located and then rewrites them to form the final outlines. Furthermore, based on the similarity between video outlines and textual outlines, we use a large number of articles with chapter headings to pretrain our model. Experiments on DuVOG show that our model largely outperforms other baseline methods, achieving 77.1 of F1-score for the video segmentation level and 85.0 of ROUGE-L_F0.5 for the headline generation level.

Angen Ye, Boyuan Wang, Chaojun Ni et al.

World-Action Models (WAM) initialized from pre-trained video generation backbones have demonstrated remarkable potential for robot policy learning. However, existing approaches face two critical bottlenecks that hinder performance and deployment. First, jointly reasoning over future visual dynamics and corresponding actions incurs substantial inference overhead. Second, joint modeling often entangles visual and motion representations, making motion prediction accuracy heavily dependent on the quality of future video forecasts. To address these issues, we introduce GigaWorld-Policy, an action-centered WAM that learns 2D pixel-action dynamics while enabling efficient action decoding, with optional video generation. Specifically, we formulate policy training into two coupled components: the model predicts future action sequences conditioned on the current observation, and simultaneously generates future videos conditioned on the predicted actions and the same observation. The policy is supervised by both action prediction and video generation, providing richer learning signals and encouraging physically plausible actions through visual-dynamics constraints. With a causal design that prevents future-video tokens from influencing action tokens, explicit future-video generation is optional at inference time, allowing faster action prediction during deployment. To support this paradigm, we curate a diverse, large-scale robot dataset to pre-train an action-centered video generation model, which is then adapted as the backbone for robot policy learning. Experimental results on real-world robotic platforms show that GigaWorld-Policy runs 9x faster than the leading WAM baseline, Motus, while improving task success rates by 7%. Moreover, compared with pi-0.5, GigaWorld-Policy improves performance by 95% on RoboTwin 2.0.

Ning Liao, Xiaopeng Zhang, Min Cao et al.

In realistic open-set scenarios where labels of a part of testing data are totally unknown, when vision-language (VL) prompt learning methods encounter inputs related to unknown classes (i.e., not seen during training), they always predict them as one of the training classes. The exhibited label bias causes difficulty in open set recognition (OSR), in which an image should be correctly predicted as one of the known classes or the unknown one. To achieve this goal, we propose a vision-language prompt tuning method with mitigated label bias (M-Tuning). It introduces open words from the WordNet to extend the range of words forming the prompt texts from only closed-set label words to more, and thus prompts are tuned in a simulated open-set scenario. Besides, inspired by the observation that classifying directly on large datasets causes a much higher false positive rate than on small datasets, we propose a Combinatorial Tuning and Testing (CTT) strategy for improving performance. CTT decomposes M-Tuning on large datasets as multiple independent group-wise tuning on fewer classes, then makes accurate and comprehensive predictions by selecting the optimal sub-prompt. Finally, given the lack of VL-based OSR baselines in the literature, especially for prompt methods, we contribute new baselines for fair comparisons. Our method achieves the best performance on datasets with various scales, and extensive ablation studies also validate its effectiveness.

Min Cao, Cong Ding, Chen Chen et al.

Most existing person re-identification methods compute the matching relations between person images across camera views based on the ranking of the pairwise similarities. This matching strategy with the lack of the global viewpoint and the context's consideration inevitably leads to ambiguous matching results and sub-optimal performance. Based on a natural assumption that images belonging to the same person identity should not match with images belonging to multiple different person identities across views, called the unicity of person matching on the identity level, we propose an end-to-end person unicity matching architecture for learning and refining the person matching relations. First, we adopt the image samples' contextual information in feature space to generate the initial soft matching results by using graph neural networks. Secondly, we utilize the samples' global context relationship to refine the soft matching results and reach the matching unicity through bipartite graph matching. Given full consideration to real-world person re-identification applications, we achieve the unicity matching in both one-shot and multi-shot settings of person re-identification and further develop a fast version of the unicity matching without losing the performance. The proposed method is evaluated on five public benchmarks, including four multi-shot datasets MSMT17, DukeMTMC, Market1501, CUHK03, and a one-shot dataset VIPeR. Experimental results show the superiority of the proposed method on performance and efficiency.

Jiayang Sun, Zixin Guo, Min Cao et al.

Change captioning generates descriptions that explicitly describe the differences between two visually similar images. Existing methods operate on static image pairs, thus ignoring the rich temporal dynamics of the change procedure, which is the key to understand not only what has changed but also how it occurs. We introduce ProCap, a novel framework that reformulates change modeling from static image comparison to dynamic procedure modeling. ProCap features a two-stage design: The first stage trains a procedure encoder to learn the change procedure from a sparse set of keyframes. These keyframes are obtained by automatically generating intermediate frames to make the implicit procedural dynamics explicit and then sampling them to mitigate redundancy. Then the encoder learns to capture the latent dynamics of these keyframes via a caption-conditioned, masked reconstruction task. The second stage integrates this trained encoder within an encoder-decoder model for captioning. Instead of relying on explicit frames from the previous stage -- a process incurring computational overhead and sensitivity to visual noise -- we introduce learnable procedure queries to prompt the encoder for inferring the latent procedure representation, which the decoder then translates into text. The entire model is then trained end-to-end with a captioning loss, ensuring the encoder's output is both temporally coherent and captioning-aligned. Experiments on three datasets demonstrate the effectiveness of ProCap. Code and pre-trained models are available at https://github.com/BlueberryOreo/ProCap

Xu Yan, Chunhui Ai, Ziqiang Cao et al.

An outstanding image-text retrieval model depends on high-quality labeled data. While the builders of existing image-text retrieval datasets strive to ensure that the caption matches the linked image, they cannot prevent a caption from fitting other images. We observe that such a many-to-many matching phenomenon is quite common in the widely-used retrieval datasets, where one caption can describe up to 178 images. These large matching-lost data not only confuse the model in training but also weaken the evaluation accuracy. Inspired by visual and textual entailment tasks, we propose a multi-modal entailment classifier to determine whether a sentence is entailed by an image plus its linked captions. Subsequently, we revise the image-text retrieval datasets by adding these entailed captions as additional weak labels of an image and develop a universal variable learning rate strategy to teach a retrieval model to distinguish the entailed captions from other negative samples. In experiments, we manually annotate an entailment-corrected image-text retrieval dataset for evaluation. The results demonstrate that the proposed entailment classifier achieves about 78% accuracy and consistently improves the performance of image-text retrieval baselines.

Xinyu Zhou, Yu Wu, Jiayao Ma et al.

This work introduces Text-based Aerial-Ground Person Retrieval (TAG-PR), which aims to retrieve person images from heterogeneous aerial and ground views with textual descriptions. Unlike traditional Text-based Person Retrieval (T-PR), which focuses solely on ground-view images, TAG-PR introduces greater practical significance and presents unique challenges due to the large viewpoint discrepancy across images. To support this task, we contribute: (1) TAG-PEDES dataset, constructed from public benchmarks with automatically generated textual descriptions, enhanced by a diversified text generation paradigm to ensure robustness under view heterogeneity; and (2) TAG-CLIP, a novel retrieval framework that addresses view heterogeneity through a hierarchically-routed mixture of experts module to learn view-specific and view-agnostic features and a viewpoint decoupling strategy to decouple view-specific features for better cross-modal alignment. We evaluate the effectiveness of TAG-CLIP on both the proposed TAG-PEDES dataset and existing T-PR benchmarks. The dataset and code are available at https://github.com/Flame-Chasers/TAG-PR.

Mengxia Wu, Min Cao, Yang Bai et al.

Text-to-video retrieval (TVR) aims to find the most relevant video in a large video gallery given a query text. The intricate and abundant context of the video challenges the performance and efficiency of TVR. To handle the serialized video contexts, existing methods typically select a subset of frames within a video to represent the video content for TVR. How to select the most representative frames is a crucial issue, whereby the selected frames are required to not only retain the semantic information of the video but also promote retrieval efficiency by excluding temporally redundant frames. In this paper, we make the first empirical study of frame selection for TVR. We systemically classify existing frame selection methods into text-free and text-guided ones, under which we detailedly analyze six different frame selections in terms of effectiveness and efficiency. Among them, two frame selections are first developed in this paper. According to the comprehensive analysis on multiple TVR benchmarks, we empirically conclude that the TVR with proper frame selections can significantly improve the retrieval efficiency without sacrificing the retrieval performance.

Yingjia Xu, Jiulong Wu, Bowen Zhang et al.

Multimodal sarcasm detection requires reasoning over cross-modal incongruities between literal expression and intended meaning, yet the specific analytical perspectives needed vary across samples due to the diversity of sarcastic mechanisms. While recent methods make this analytical process explicit, they still rely on fixed, predefined perspectives that operate independently under hand-crafted routing rules. We argue that multimodal sarcasm detection instead calls for self-elicited multi-perspective reasoning, where a model autonomously generates the perspectives needed for each sample and progressively integrates them into a coherent analysis. To realize this goal, we propose ProCrit, a Proposal-Critic two-agent framework with a proposal agent for multi-perspective reasoning and a critic agent for external evaluation and targeted revision guidance. First, to overcome the lack of process-level supervision in existing sarcasm datasets, ProCrit synthesizes process-level reasoning annotations through a dynamic-role agentic rollout: a strong vision-language model sequentially spawns analytical roles within a shared context, and the resulting multi-role trajectories are flattened into sequences that preserve cross-perspective dependencies while enabling efficient autoregressive generation. Second, to improve reasoning reliability, ProCrit adopts a draft-critique-revise paradigm in which an independent critic identifies reasoning deficiencies and provides targeted natural-language feedback for directed revision. Finally, we develop a mutual-refinement training framework that jointly optimizes proposal drafting and feedback-guided revision via dual-stage reinforcement learning, while refining the critic agent according to the actual effectiveness of its feedback. Experiments on three widely used benchmarks demonstrate the effectiveness of ProCrit.

Jiulong Wu, Yucheng Shen, Lingyong Yan et al.

Facial Emotion Analysis (FEA) extends traditional facial emotion recognition by incorporating explainable, fine-grained reasoning. The task integrates three subtasks: emotion recognition, facial Action Unit (AU) recognition, and AU-based emotion reasoning to model affective states jointly. While recent approaches leverage Vision-Language Models (VLMs) and achieve promising results, they face two critical limitations: (1) hallucinated reasoning, where VLMs generate plausible but inaccurate explanations due to insufficient emotion-specific knowledge; and (2) misalignment between emotion reasoning and recognition, caused by fragmented connections between observed facial features and final labels. We propose Facial-R1, a three-stage alignment framework that effectively addresses both challenges with minimal supervision. First, we employ instruction fine-tuning to establish basic emotional reasoning capability. Second, we introduce reinforcement training guided by emotion and AU labels as reward signals, which explicitly aligns the generated reasoning process with the predicted emotion. Third, we design a data synthesis pipeline that iteratively leverages the prior stages to expand the training dataset, enabling scalable self-improvement of the model. Built upon this framework, we introduce FEA-20K, a benchmark dataset comprising 17,737 training and 1,688 test samples with fine-grained emotion analysis annotations. Extensive experiments across eight standard benchmarks demonstrate that Facial-R1 achieves state-of-the-art performance in FEA, with strong generalization and robust interpretability.

Ning Liao, Shaofeng Zhang, Renqiu Xia et al.

There is an emerging line of research on multimodal instruction tuning, and a line of benchmarks has been proposed for evaluating these models recently. Instead of evaluating the models directly, in this paper, we try to evaluate the Vision-Language Instruction-Tuning (VLIT) datasets. Also, we seek the way of building a dataset for developing an all-powerful VLIT model, which we believe could also be of utility for establishing a grounded protocol for benchmarking VLIT models. For effective evaluation of VLIT datasets that remains an open question, we propose a tune-cross-evaluation paradigm: tuning on one dataset and evaluating on the others in turn. For each single tune-evaluation experiment set, we define the Meta Quality (MQ) as the mean score obtained by a set of caption metrics including BLEU, METEOR, and ROUGE-L to quantify the quality of a certain dataset or a sample. On this basis, to evaluate the comprehensiveness of a dataset, we develop the Dataset Quality (DQ) covering all tune-evaluation sets. To lay the foundation for building a comprehensive dataset and developing an all-powerful model for practical applications, we define the Sample Quality (SQ) to quantify the all-sided quality of each sample. Extensive experiments validate the rationality of the proposed evaluation paradigm. Based on the holistic evaluation, we build a new dataset, REVO-LION (REfining VisiOn-Language InstructiOn tuNing), by collecting samples with higher SQ from each dataset. Remarkably, even with only half of the complete data, the model trained on REVO-LION can achieve the performance comparable to simply adding all VLIT datasets up. Furthermore, REVO-LION not only facilitates the development of a powerful model but also incorporates an evaluation set, which is designed to serve as a convenient benchmark for future research in the field.

Min Cao, Xinyu Zhou, Ding Jiang et al.

Text-to-image person retrieval (TIPR) aims to identify the target person using textual descriptions, facing challenge in modality heterogeneity. Prior works have attempted to address it by developing cross-modal global or local alignment strategies. However, global methods typically overlook fine-grained cross-modal differences, whereas local methods require prior information to explore explicit part alignments. Additionally, current methods are English-centric, restricting their application in multilingual contexts. To alleviate these issues, we pioneer a multilingual TIPR task by developing a multilingual TIPR benchmark, for which we leverage large language models for initial translations and refine them by integrating domain-specific knowledge. Correspondingly, we propose Bi-IRRA: a Bidirectional Implicit Relation Reasoning and Aligning framework to learn alignment across languages and modalities. Within Bi-IRRA, a bidirectional implicit relation reasoning module enables bidirectional prediction of masked image and text, implicitly enhancing the modeling of local relations across languages and modalities, a multi-dimensional global alignment module is integrated to bridge the modality heterogeneity. The proposed method achieves new state-of-the-art results on all multilingual TIPR datasets. Data and code are presented in https://github.com/Flame-Chasers/Bi-IRRA.

Jiulong Wu, Zhengliang Shi, Shuaiqiang Wang et al. · baidu

Large Visual Language Models (LVLMs) have demonstrated impressive capabilities across multiple tasks. However, their trustworthiness is often challenged by hallucinations, which can be attributed to the modality misalignment and the inherent hallucinations of their underlying Large Language Models (LLMs) backbone. Existing preference alignment methods focus on aligning model responses with human preferences while neglecting image-text modality alignment, resulting in over-reliance on LLMs and hallucinations. In this paper, we propose Entity-centric Multimodal Preference Optimization (EMPO), which achieves enhanced modality alignment compared to existing human preference alignment methods. Besides, to overcome the scarcity of high-quality multimodal preference data, we utilize open-source instruction datasets to automatically construct high-quality preference data across three aspects: image, instruction, and response. Experiments on two human preference datasets and five multimodal hallucination benchmarks demonstrate the effectiveness of EMPO, e.g., reducing hallucination rates by 85.9\% on Object-HalBench and 49.8\% on MM-HalBench.

Yang Bai, Min Cao, Daming Gao et al.

Text-based person search aims to retrieve the specified person images given a textual description. The key to tackling such a challenging task is to learn powerful multi-modal representations. Towards this, we propose a Relation and Sensitivity aware representation learning method (RaSa), including two novel tasks: Relation-Aware learning (RA) and Sensitivity-Aware learning (SA). For one thing, existing methods cluster representations of all positive pairs without distinction and overlook the noise problem caused by the weak positive pairs where the text and the paired image have noise correspondences, thus leading to overfitting learning. RA offsets the overfitting risk by introducing a novel positive relation detection task (i.e., learning to distinguish strong and weak positive pairs). For another thing, learning invariant representation under data augmentation (i.e., being insensitive to some transformations) is a general practice for improving representation's robustness in existing methods. Beyond that, we encourage the representation to perceive the sensitive transformation by SA (i.e., learning to detect the replaced words), thus promoting the representation's robustness. Experiments demonstrate that RaSa outperforms existing state-of-the-art methods by 6.94%, 4.45% and 15.35% in terms of Rank@1 on CUHK-PEDES, ICFG-PEDES and RSTPReid datasets, respectively. Code is available at: https://github.com/Flame-Chasers/RaSa.

Min Cao, Chen Chen, Hao Dou et al.

Most existing person re-identification methods compute pairwise similarity by extracting robust visual features and learning the discriminative metric. Owing to visual ambiguities, these content-based methods that determine the pairwise relationship only based on the similarity between them, inevitably produce a suboptimal ranking list. Instead, the pairwise similarity can be estimated more accurately along the geodesic path of the underlying data manifold by exploring the rich contextual information of the sample. In this paper, we propose a lightweight post-processing person re-identification method in which the pairwise measure is determined by the relationship between the sample and the counterpart's context in an unsupervised way. We translate the point-to-point comparison into the bilateral point-to-set comparison. The sample's context is composed of its neighbor samples with two different definition ways: the first order context and the second order context, which are used to compute the pairwise similarity in sequence, resulting in a progressive post-processing model. The experiments on four large-scale person re-identification benchmark datasets indicate that (1) the proposed method can consistently achieve higher accuracies by serving as a post-processing procedure after the content-based person re-identification methods, showing its state-of-the-art results, (2) the proposed lightweight method only needs about 6 milliseconds for optimizing the ranking results of one sample, showing its high-efficiency. Code is available at: https://github.com/123ci/PBCmodel.

Yuxin Lu, Qian Qiao, Jiayang Sun et al.

Recent advances in diffusion models have markedly enhanced the visual fidelity of audio-driven talking head generation. Nevertheless, existing methods are constrained by three critical limitations: causal inefficiency that impedes real-time inference, incompatibility with temporally coherent conditioning, and progressive drift over long-horizon generation, collectively hindering their deployment in real-time applications. To overcome these challenges, we introduce AsymK-Talker, a novel diffusion-distillation method designed for real-time and long-horizon talking head generation. AsymK-Talker comprises three key components: (1) Kernel-Conditioned Loop Generation (KCLG), a causal, chunk-wise generation paradigm that leverages motion kernels to enable temporally consistent propagation; (2) Temporal Reference Encoding (TRE), which converts a static identity reference into a time-aware latent representation to enhance audio-visual synchronization; and (3) Asymmetric Kernel Distillation (AKD), a teacher-student distillation framework wherein the teacher model conditions on ground-truth motion kernels for supervision, while the student learns to generate from generated kernels, thereby ensuring robustness during extended generation sequences. AsymK-Talker achieves promising results on both visual fidelity and lip synchronization metrics.

Qi Lv, Lei Geng, Ziqiang Cao et al.

Softmax with the cross entropy loss is the standard configuration for current neural classification models. The gold score for a target class is supposed to be 1, but it is never reachable under the softmax schema. Such a problem makes the training process continue forever and leads to overfitting. Moreover, the "target-approach-1" training goal forces the model to continuously learn all samples, leading to a waste of time in handling some samples which have already been classified correctly with high confidence, while the test goal simply requires the target class of each sample to hold the maximum score. To solve the above weaknesses, we propose the Adaptive Sparse softmax (AS-Softmax) which designs a reasonable and test-matching transformation on top of softmax. For more purposeful learning, we discard the classes with far smaller scores compared with the actual class during training. Then the model could focus on learning to distinguish the target class from its strong opponents, which is also the great challenge in test. In addition, since the training losses of easy samples will gradually drop to 0 in AS-Softmax, we develop an adaptive gradient accumulation strategy based on the masked sample ratio to speed up training. We verify the proposed AS-Softmax on a variety of text multi-class, text multi-label, text token classification, image classification and audio classification tasks with class sizes ranging from 5 to 5000+. The results show that AS-Softmax consistently outperforms softmax and its variants, and the loss of AS-Softmax is remarkably correlated with classification performance in validation. Furthermore, adaptive gradient accumulation strategy can bring about 1.2x training speedup comparing with the standard softmax while maintaining classification effectiveness.

Daming Gao, Yang Bai, Min Cao et al.

Text-based person search (TBPS) aims to retrieve images of a specific person from a large image gallery based on a natural language description. Existing methods rely on massive annotated image-text data to achieve satisfactory performance in fully-supervised learning. It poses a significant challenge in practice, as acquiring person images from surveillance videos is relatively easy, while obtaining annotated texts is challenging. The paper undertakes a pioneering initiative to explore TBPS under the semi-supervised setting, where only a limited number of person images are annotated with textual descriptions while the majority of images lack annotations. We present a two-stage basic solution based on generation-then-retrieval for semi-supervised TBPS. The generation stage enriches annotated data by applying an image captioning model to generate pseudo-texts for unannotated images. Later, the retrieval stage performs fully-supervised retrieval learning using the augmented data. Significantly, considering the noise interference of the pseudo-texts on retrieval learning, we propose a noise-robust retrieval framework that enhances the ability of the retrieval model to handle noisy data. The framework integrates two key strategies: Hybrid Patch-Channel Masking (PC-Mask) to refine the model architecture, and Noise-Guided Progressive Training (NP-Train) to enhance the training process. PC-Mask performs masking on the input data at both the patch-level and the channel-level to prevent overfitting noisy supervision. NP-Train introduces a progressive training schedule based on the noise level of pseudo-texts to facilitate noise-robust learning. Extensive experiments on multiple TBPS benchmarks show that the proposed framework achieves promising performance under the semi-supervised setting.

Yucheng Shen, Jiulong Wu, Jizhou Huang et al.

Visual Retrieval-Augmented Generation (VRAG) empowers Vision-Language Models to retrieve and reason over visually rich documents. To tackle complex queries requiring multi-step reasoning, agentic VRAG systems interleave reasoning with iterative retrieval.. However, existing agentic VRAG faces two critical bottlenecks. (1) Visual Evidence Sparsity: key evidence is scattered across pages yet processed in isolation, hindering cross-page reasoning; moreover, fine-grained intra-image evidence often requires precise visual actions, whose misuse degrades retrieval quality; (2) Search Drift in Long Horizons: the accumulation of visual tokens across retrieved pages dilutes context and causes cognitive overload, leading agents to deviate from their search objective. To address these challenges, we propose VISOR (Visual Retrieval-Augmented Generation via Iterative Search and Over-horizon Reasoning), a unified single-agent framework. VISOR features a structured Evidence Space for progressive cross-page reasoning, coupled with a Visual Action Evaluation and Correction mechanism to manage visual actions. Additionally, we introduce a Dynamic Trajectory with Sliding Window and Intent Injection to mitigate search drift. They anchor the evidence space while discarding earlier raw interactions, preventing context from being overwhelmed by visual tokens. We train VISOR using a Group Relative Policy Optimization-based Reinforcement Learning (GRPO-based RL) pipeline with state masking and credit assignment tailored for dynamic context reconstruction. Extensive experiments on ViDoSeek, SlideVQA, and MMLongBench demonstrate that VISOR achieves state-of-the-art performance with superior efficiency for long-horizon visual reasoning tasks.

Yingjia Xu, Jinlin Wu, Zhen Chen et al.

Text-based person retrieval aims to identify a target individual from a gallery of images based on a natural language description. It presents a significant challenge due to the complexity of real-world scenes and the ambiguity of appearance-related descriptions. Existing methods primarily emphasize appearance-based cross-modal retrieval, often neglecting the contextual information embedded within the scene, which can offer valuable complementary insights for retrieval. To address this, we introduce SCENEPERSON-13W, a large-scale dataset featuring over 100,000 scenes with rich annotations covering both pedestrian appearance and environmental cues. Based on this, we propose SA-Person, a two-stage retrieval framework. In the first stage, it performs discriminative appearance grounding by aligning textual cues with pedestrian-specific regions. In the second stage, it introduces SceneRanker, a training-free, scene-aware re-ranking method leveraging multimodal large language models to jointly reason over pedestrian appearance and the global scene context. Experiments on SCENEPERSON-13W validate the effectiveness of our framework in challenging scene-level retrieval scenarios. The code and dataset will be made publicly available.

Min Cao, ZiYin Zeng, YuXin Lu et al.

Data plays a pivotal role in Text-Based Person Retrieval (TBPR) research. Mainstream research paradigm necessitates real-world person images with manual textual annotations for training models, posing privacy-sensitive and labor-intensive issues. Several pioneering efforts explore synthetic data for TBPR but still rely on real data, keeping the aforementioned issues and also resulting in diversity-deficient issue in synthetic datasets, thus impacting TBPR performance. Moreover, these works tend to explore synthetic data for TBPR through limited perspectives, leading to exploration-restricted issue. In this paper, we conduct an empirical study to explore the potential of synthetic data for TBPR, highlighting three key aspects. (1) We propose an inter-class image generation pipeline, in which an automatic prompt construction strategy is introduced to guide generative Artificial Intelligence (AI) models in generating various inter-class images without reliance on original data. (2) We develop an intra-class image augmentation pipeline, in which the generative AI models are applied to further edit the images for obtaining various intra-class images. (3) Building upon the proposed pipelines and an automatic text generation pipeline, we explore the effectiveness of synthetic data in diverse scenarios through extensive experiments. Additionally, we experimentally investigate various noise-robust learning strategies to mitigate the inherent noise in synthetic data. We will release the code, along with the synthetic large-scale dataset generated by our pipelines, which are expected to advance practical TBPR research.

Yang Bai, Jingyao Wang, Min Cao et al.

Text-based person search (TBPS) aims to retrieve the images of the target person from a large image gallery based on a given natural language description. Existing methods are dominated by training models with parallel image-text pairs, which are very costly to collect. In this paper, we make the first attempt to explore TBPS without parallel image-text data ($μ$-TBPS), in which only non-parallel images and texts, or even image-only data, can be adopted. Towards this end, we propose a two-stage framework, generation-then-retrieval (GTR), to first generate the corresponding pseudo text for each image and then perform the retrieval in a supervised manner. In the generation stage, we propose a fine-grained image captioning strategy to obtain an enriched description of the person image, which firstly utilizes a set of instruction prompts to activate the off-the-shelf pretrained vision-language model to capture and generate fine-grained person attributes, and then converts the extracted attributes into a textual description via the finetuned large language model or the hand-crafted template. In the retrieval stage, considering the noise interference of the generated texts for training model, we develop a confidence score-based training scheme by enabling more reliable texts to contribute more during the training. Experimental results on multiple TBPS benchmarks (i.e., CUHK-PEDES, ICFG-PEDES and RSTPReid) show that the proposed GTR can achieve a promising performance without relying on parallel image-text data.

Shiping Li, Min Cao, Min Zhang

Text-based person search aims to retrieve images of a certain pedestrian by a textual description. The key challenge of this task is to eliminate the inter-modality gap and achieve the feature alignment across modalities. In this paper, we propose a semantic-aligned embedding method for text-based person search, in which the feature alignment across modalities is achieved by automatically learning the semantic-aligned visual features and textual features. First, we introduce two Transformer-based backbones to encode robust feature representations of the images and texts. Second, we design a semantic-aligned feature aggregation network to adaptively select and aggregate features with the same semantics into part-aware features, which is achieved by a multi-head attention module constrained by a cross-modality part alignment loss and a diversity loss. Experimental results on the CUHK-PEDES and Flickr30K datasets show that our method achieves state-of-the-art performances.

Chen Chen, Min Cao, Xiyuan Hu et al.

Ideally person re-identification seeks for perfect feature representation and metric model that re-identify all various pedestrians well in non-overlapping views at different locations with different camera configurations, which is very challenging. However, in most pedestrian sets, there always are some outstanding persons who are relatively easy to re-identify. Inspired by the existence of such data division, we propose a novel key person aided person re-identification framework based on the re-defined partially ordered pedestrian sets. The outstanding persons, namely "key persons", are selected by the K-nearest neighbor based saliency measurement. The partial order defined by pedestrian entering time in surveillance associates the key persons with the query person temporally and helps to locate the possible candidates. Experiments conducted on two video datasets show that the proposed key person aided framework outperforms the state-of-the-art methods and improves the matching accuracy greatly at all ranks.

Yi-bin Huang, Kang Li, Ge Wang et al.

For the problem whether Graphic Processing Unit(GPU),the stream processor with high performance of floating-point computing is applicable to neural networks, this paper proposes the parallel recognition algorithm of Convolutional Neural Networks(CNNs).It adopts Compute Unified Device Architecture(CUDA)technology, definite the parallel data structures, and describes the mapping mechanism for computing tasks on CUDA. It compares the parallel recognition algorithm achieved on GPU of GTX200 hardware architecture with the serial algorithm on CPU. It improves speed by nearly 60 times. Result shows that GPU based the stream processor architecture ate more applicable to some related applications about neural networks than CPU.