Qinghao Ye

Haiyang Xu, Qinghao Ye, Ming Yan et al.

Recent years have witnessed a big convergence of language, vision, and multi-modal pretraining. In this work, we present mPLUG-2, a new unified paradigm with modularized design for multi-modal pretraining, which can benefit from modality collaboration while addressing the problem of modality entanglement. In contrast to predominant paradigms of solely relying on sequence-to-sequence generation or encoder-based instance discrimination, mPLUG-2 introduces a multi-module composition network by sharing common universal modules for modality collaboration and disentangling different modality modules to deal with modality entanglement. It is flexible to select different modules for different understanding and generation tasks across all modalities including text, image, and video. Empirical study shows that mPLUG-2 achieves state-of-the-art or competitive results on a broad range of over 30 downstream tasks, spanning multi-modal tasks of image-text and video-text understanding and generation, and uni-modal tasks of text-only, image-only, and video-only understanding. Notably, mPLUG-2 shows new state-of-the-art results of 48.0 top-1 accuracy and 80.3 CIDEr on the challenging MSRVTT video QA and video caption tasks with a far smaller model size and data scale. It also demonstrates strong zero-shot transferability on vision-language and video-language tasks. Code and models will be released in https://github.com/alibaba/AliceMind.

Qinghao Ye, Haiyang Xu, Guohai Xu et al.

Large language models (LLMs) have demonstrated impressive zero-shot abilities on a variety of open-ended tasks, while recent research has also explored the use of LLMs for multi-modal generation. In this study, we introduce mPLUG-Owl, a novel training paradigm that equips LLMs with multi-modal abilities through modularized learning of foundation LLM, a visual knowledge module, and a visual abstractor module. This approach can support multiple modalities and facilitate diverse unimodal and multimodal abilities through modality collaboration. The training paradigm of mPLUG-Owl involves a two-stage method for aligning image and text, which learns visual knowledge with the assistance of LLM while maintaining and even improving the generation abilities of LLM. In the first stage, the visual knowledge module and abstractor module are trained with a frozen LLM module to align the image and text. In the second stage, language-only and multi-modal supervised datasets are used to jointly fine-tune a low-rank adaption (LoRA) module on LLM and the abstractor module by freezing the visual knowledge module. We carefully build a visually-related instruction evaluation set OwlEval. Experimental results show that our model outperforms existing multi-modal models, demonstrating mPLUG-Owl's impressive instruction and visual understanding ability, multi-turn conversation ability, and knowledge reasoning ability. Besides, we observe some unexpected and exciting abilities such as multi-image correlation and scene text understanding, which makes it possible to leverage it for harder real scenarios, such as vision-only document comprehension. Our code, pre-trained model, instruction-tuned models, and evaluation set are available at https://github.com/X-PLUG/mPLUG-Owl. The online demo is available at https://www.modelscope.cn/studios/damo/mPLUG-Owl.

Jiabo Ye, Anwen Hu, Haiyang Xu et al.

Document understanding refers to automatically extract, analyze and comprehend information from various types of digital documents, such as a web page. Existing Multi-model Large Language Models (MLLMs), including mPLUG-Owl, have demonstrated promising zero-shot capabilities in shallow OCR-free text recognition, indicating their potential for OCR-free document understanding. Nevertheless, without in-domain training, these models tend to ignore fine-grained OCR features, such as sophisticated tables or large blocks of text, which are essential for OCR-free document understanding. In this paper, we propose mPLUG-DocOwl based on mPLUG-Owl for OCR-free document understanding. Specifically, we first construct a instruction tuning dataset featuring a wide range of visual-text understanding tasks. Then, we strengthen the OCR-free document understanding ability by jointly train the model on language-only, general vision-and-language, and document instruction tuning dataset with our unified instruction tuning strategy. We also build an OCR-free document instruction understanding evaluation set LLMDoc to better compare models' capabilities on instruct compliance and document understanding. Experimental results show that our model outperforms existing multi-modal models, demonstrating its strong ability of document understanding. Besides, without specific fine-tuning, mPLUG-DocOwl generalizes well on various downstream tasks. Our code, models, training data and evaluation set are available at https://github.com/X-PLUG/mPLUG-DocOwl.

Junfeng Tian, Hehong Chen, Guohai Xu et al.

In this paper, we present ChatPLUG, a Chinese open-domain dialogue system for digital human applications that instruction finetunes on a wide range of dialogue tasks in a unified internet-augmented format. Different from other open-domain dialogue models that focus on large-scale pre-training and scaling up model size or dialogue corpus, we aim to build a powerful and practical dialogue system for digital human with diverse skills and good multi-task generalization by internet-augmented instruction tuning. To this end, we first conduct large-scale pre-training on both common document corpus and dialogue data with curriculum learning, so as to inject various world knowledge and dialogue abilities into ChatPLUG. Then, we collect a wide range of dialogue tasks spanning diverse features of knowledge, personality, multi-turn memory, and empathy, on which we further instruction tune \modelname via unified natural language instruction templates. External knowledge from an internet search is also used during instruction finetuning for alleviating the problem of knowledge hallucinations. We show that \modelname outperforms state-of-the-art Chinese dialogue systems on both automatic and human evaluation, and demonstrates strong multi-task generalization on a variety of text understanding and generation tasks. In addition, we deploy \modelname to real-world applications such as Smart Speaker and Instant Message applications with fast inference. Our models and code will be made publicly available on ModelScope: https://modelscope.cn/models/damo/ChatPLUG-3.7B and Github: https://github.com/X-PLUG/ChatPLUG .

Zhaorun Chen, Yichao Du, Zichen Wen et al.

While text-to-image models like DALLE-3 and Stable Diffusion are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with desired behaviors based on feedback from a multimodal judge. Despite their significance, current multimodal judges frequently undergo inadequate evaluation of their capabilities and limitations, potentially leading to misalignment and unsafe fine-tuning outcomes. To address this issue, we introduce MJ-Bench, a novel benchmark which incorporates a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image generation models across four key perspectives: alignment, safety, image quality, and bias. Specifically, we evaluate a large variety of multimodal judges including smaller-sized CLIP-based scoring models, open-source VLMs (e.g. LLaVA family), and close-source VLMs (e.g. GPT-4o, Claude 3) on each decomposed subcategory of our preference dataset. Experiments reveal that close-source VLMs generally provide better feedback, with GPT-4o outperforming other judges in average. Compared with open-source VLMs, smaller-sized scoring models can provide better feedback regarding text-image alignment and image quality, while VLMs provide more accurate feedback regarding safety and generation bias due to their stronger reasoning capabilities. Further studies in feedback scale reveal that VLM judges can generally provide more accurate and stable feedback in natural language (Likert-scale) than numerical scales. Notably, human evaluations on end-to-end fine-tuned models using separate feedback from these multimodal judges provide similar conclusions, further confirming the effectiveness of MJ-Bench. All data, code, models are available at https://huggingface.co/MJ-Bench.

Anwen Hu, Yaya Shi, Haiyang Xu et al.

Recently, the strong text creation ability of Large Language Models(LLMs) has given rise to many tools for assisting paper reading or even writing. However, the weak diagram analysis abilities of LLMs or Multimodal LLMs greatly limit their application scenarios, especially for scientific academic paper writing. In this work, towards a more versatile copilot for academic paper writing, we mainly focus on strengthening the multi-modal diagram analysis ability of Multimodal LLMs. By parsing Latex source files of high-quality papers, we carefully build a multi-modal diagram understanding dataset M-Paper. By aligning diagrams in the paper with related paragraphs, we construct professional diagram analysis samples for training and evaluation. M-Paper is the first dataset to support joint comprehension of multiple scientific diagrams, including figures and tables in the format of images or Latex codes. Besides, to better align the copilot with the user's intention, we introduce the `outline' as the control signal, which could be directly given by the user or revised based on auto-generated ones. Comprehensive experiments with a state-of-the-art Mumtimodal LLM demonstrate that training on our dataset shows stronger scientific diagram understanding performance, including diagram captioning, diagram analysis, and outline recommendation. The dataset, code, and model are available at https://github.com/X-PLUG/mPLUG-DocOwl/tree/main/PaperOwl.

Qinghao Ye, Haiyang Xu, Jiabo Ye et al.

Multi-modal Large Language Models (MLLMs) have demonstrated impressive instruction abilities across various open-ended tasks. However, previous methods primarily focus on enhancing multi-modal capabilities. In this work, we introduce a versatile multi-modal large language model, mPLUG-Owl2, which effectively leverages modality collaboration to improve performance in both text and multi-modal tasks. mPLUG-Owl2 utilizes a modularized network design, with the language decoder acting as a universal interface for managing different modalities. Specifically, mPLUG-Owl2 incorporates shared functional modules to facilitate modality collaboration and introduces a modality-adaptive module that preserves modality-specific features. Extensive experiments reveal that mPLUG-Owl2 is capable of generalizing both text tasks and multi-modal tasks and achieving state-of-the-art performances with a single generic model. Notably, mPLUG-Owl2 is the first MLLM model that demonstrates the modality collaboration phenomenon in both pure-text and multi-modal scenarios, setting a pioneering path in the development of future multi-modal foundation models.

Qinghao Ye, Guohai Xu, Ming Yan et al.

Video-language pre-training has advanced the performance of various downstream video-language tasks. However, most previous methods directly inherit or adapt typical image-language pre-training paradigms to video-language pre-training, thus not fully exploiting the unique characteristic of video, i.e., temporal. In this paper, we propose a Hierarchical Temporal-Aware video-language pre-training framework, HiTeA, with two novel pre-training tasks for modeling cross-modal alignment between moments and texts as well as the temporal relations of video-text pairs. Specifically, we propose a cross-modal moment exploration task to explore moments in videos, which results in detailed video moment representation. Besides, the inherent temporal relations are captured by aligning video-text pairs as a whole in different time resolutions with multi-modal temporal relation exploration task. Furthermore, we introduce the shuffling test to evaluate the temporal reliance of datasets and video-language pre-training models. We achieve state-of-the-art results on 15 well-established video-language understanding and generation tasks, especially on temporal-oriented datasets (e.g., SSv2-Template and SSv2-Label) with 8.6% and 11.1% improvement respectively. HiTeA also demonstrates strong generalization ability when directly transferred to downstream tasks in a zero-shot manner. Models and demo will be available on ModelScope.

Jiabo Ye, Anwen Hu, Haiyang Xu et al.

Text is ubiquitous in our visual world, conveying crucial information, such as in documents, websites, and everyday photographs. In this work, we propose UReader, a first exploration of universal OCR-free visually-situated language understanding based on the Multimodal Large Language Model (MLLM). By leveraging the shallow text recognition ability of the MLLM, we only finetuned 1.2% parameters and the training cost is much lower than previous work following domain-specific pretraining and finetuning paradigms. Concretely, UReader is jointly finetuned on a wide range of Visually-situated Language Understanding tasks via a unified instruction format. To enhance the visual text and semantic understanding, we further apply two auxiliary tasks with the same format, namely text reading and key points generation tasks. We design a shape-adaptive cropping module before the encoder-decoder architecture of MLLM to leverage the frozen low-resolution vision encoder for processing high-resolution images. Without downstream finetuning, our single model achieves state-of-the-art ocr-free performance in 8 out of 10 visually-situated language understanding tasks, across 5 domains: documents, tables, charts, natural images, and webpage screenshots. Codes and instruction-tuning datasets will be released.

Junyang Wang, Yiyang Zhou, Guohai Xu et al.

Large Vision-Language Models (LVLMs) have recently achieved remarkable success. However, LVLMs are still plagued by the hallucination problem, which limits the practicality in many scenarios. Hallucination refers to the information of LVLMs' responses that does not exist in the visual input, which poses potential risks of substantial consequences. There has been limited work studying hallucination evaluation in LVLMs. In this paper, we propose Hallucination Evaluation based on Large Language Models (HaELM), an LLM-based hallucination evaluation framework. HaELM achieves an approximate 95% performance comparable to ChatGPT and has additional advantages including low cost, reproducibility, privacy preservation and local deployment. Leveraging the HaELM, we evaluate the hallucination in current LVLMs. Furthermore, we analyze the factors contributing to hallucination in LVLMs and offer helpful suggestions to mitigate the hallucination problem. Our training data and human annotation hallucination data will be made public soon.

Qi Bi, Beichen Zhou, Kun Qin et al.

Aerial scene classification remains challenging as: 1) the size of key objects in determining the scene scheme varies greatly; 2) many objects irrelevant to the scene scheme are often flooded in the image. Hence, how to effectively perceive the region of interests (RoIs) from a variety of sizes and build more discriminative representation from such complicated object distribution is vital to understand an aerial scene. In this paper, we propose a novel all grains, one scheme (AGOS) framework to tackle these challenges. To the best of our knowledge, it is the first work to extend the classic multiple instance learning into multi-grain formulation. Specially, it consists of a multi-grain perception module (MGP), a multi-branch multi-instance representation module (MBMIR) and a self-aligned semantic fusion (SSF) module. Firstly, our MGP preserves the differential dilated convolutional features from the backbone, which magnifies the discriminative information from multi-grains. Then, our MBMIR highlights the key instances in the multi-grain representation under the MIL formulation. Finally, our SSF allows our framework to learn the same scene scheme from multi-grain instance representations and fuses them, so that the entire framework is optimized as a whole. Notably, our AGOS is flexible and can be easily adapted to existing CNNs in a plug-and-play manner. Extensive experiments on UCM, AID and NWPU benchmarks demonstrate that our AGOS achieves a comparable performance against the state-of-the-art methods.

Xu Yang, Zhangzikang Li, Haiyang Xu et al.

In a video, an object usually appears as the trajectory, i.e., it spans over a few spatial but longer temporal patches, that contains abundant spatiotemporal contexts. However, modern Video-Language BERTs (VDL-BERTs) neglect this trajectory characteristic that they usually follow image-language BERTs (IL-BERTs) to deploy the patch-to-word (P2W) attention that may over-exploit trivial spatial contexts and neglect significant temporal contexts. To amend this, we propose a novel TW-BERT to learn Trajectory-Word alignment by a newly designed trajectory-to-word (T2W) attention for solving video-language tasks. Moreover, previous VDL-BERTs usually uniformly sample a few frames into the model while different trajectories have diverse graininess, i.e., some trajectories span longer frames and some span shorter, and using a few frames will lose certain useful temporal contexts. However, simply sampling more frames will also make pre-training infeasible due to the largely increased training burdens. To alleviate the problem, during the fine-tuning stage, we insert a novel Hierarchical Frame-Selector (HFS) module into the video encoder. HFS gradually selects the suitable frames conditioned on the text context for the later cross-modal encoder to learn better trajectory-word alignments. By the proposed T2W attention and HFS, our TW-BERT achieves SOTA performances on text-to-video retrieval tasks, and comparable performances on video question-answering tasks with some VDL-BERTs trained on much more data. The code will be available in the supplementary material.

Chaoya Jiang, Haiyang Xu, Wei Ye et al.

Vision Transformer (ViT) based Vision-Language Pre-training (VLP) models have demonstrated impressive performance in various tasks. However, the lengthy visual token sequences fed into ViT can lead to training inefficiency and ineffectiveness. Existing efforts address the challenge by either bottom-level patch extraction in the ViT backbone or top-level patch abstraction outside, not balancing training efficiency and effectiveness well. Inspired by text summarization in natural language processing, we propose a Bottom-Up Patch Summarization approach named BUS, coordinating bottom-level extraction and top-level abstraction to learn a concise summary of lengthy visual token sequences efficiently. Specifically, We incorporate a Text-Semantics-Aware Patch Selector (TSPS) into the ViT backbone to perform a coarse-grained visual token extraction and then attach a flexible Transformer-based Patch Abstraction Decoder (PAD) upon the backbone for top-level visual abstraction. This bottom-up collaboration enables our BUS to yield high training efficiency while maintaining or even improving effectiveness. We evaluate our approach on various visual-language understanding and generation tasks and show competitive downstream task performance while boosting the training efficiency by 50\%. Additionally, our model achieves state-of-the-art performance on many downstream tasks by increasing input image resolution without increasing computational costs over baselines.

Haiyang Xu, Qinghao Ye, Xuan Wu et al.

To promote the development of Vision-Language Pre-training (VLP) and multimodal Large Language Model (LLM) in the Chinese community, we firstly release the largest public Chinese high-quality video-language dataset named Youku-mPLUG, which is collected from Youku, a well-known Chinese video-sharing website, with strict criteria of safety, diversity, and quality. Youku-mPLUG contains 10 million Chinese video-text pairs filtered from 400 million raw videos across a wide range of 45 diverse categories for large-scale pre-training. In addition, to facilitate a comprehensive evaluation of video-language models, we carefully build the largest human-annotated Chinese benchmarks covering three popular video-language tasks of cross-modal retrieval, video captioning, and video category classification. Youku-mPLUG can enable researchers to conduct more in-depth multimodal research and develop better applications in the future. Furthermore, we release popular video-language pre-training models, ALPRO and mPLUG-2, and our proposed modularized decoder-only model mPLUG-video pre-trained on Youku-mPLUG. Experiments show that models pre-trained on Youku-mPLUG gain up to 23.1% improvement in video category classification. Besides, mPLUG-video achieves a new state-of-the-art result on these benchmarks with 80.5% top-1 accuracy in video category classification and 68.9 CIDEr score in video captioning, respectively. Finally, we scale up mPLUG-video based on the frozen Bloomz with only 1.7% trainable parameters as Chinese multimodal LLM, and demonstrate impressive instruction and video understanding ability. The zero-shot instruction understanding experiment indicates that pretraining with Youku-mPLUG can enhance the ability to comprehend overall and detailed visual semantics, recognize scene text, and leverage open-domain knowledge.

Chaoya Jiang, Haiyang Xu, Mengfan Dong et al.

Multi-modal large language models (MLLMs) have been shown to efficiently integrate natural language with visual information to handle multi-modal tasks. However, MLLMs still face a fundamental limitation of hallucinations, where they tend to generate erroneous or fabricated information. In this paper, we address hallucinations in MLLMs from a novel perspective of representation learning. We first analyzed the representation distribution of textual and visual tokens in MLLM, revealing two important findings: 1) there is a significant gap between textual and visual representations, indicating unsatisfactory cross-modal representation alignment; 2) representations of texts that contain and do not contain hallucinations are entangled, making it challenging to distinguish them. These two observations inspire us with a simple yet effective method to mitigate hallucinations. Specifically, we introduce contrastive learning into MLLMs and use text with hallucination as hard negative examples, naturally bringing representations of non-hallucinative text and visual samples closer while pushing way representations of non-hallucinating and hallucinative text. We evaluate our method quantitatively and qualitatively, showing its effectiveness in reducing hallucination occurrences and improving performance across multiple benchmarks. On the MMhal-Bench benchmark, our method obtains a 34.66% /29.5% improvement over the baseline MiniGPT-4/LLaVA. Our code is available on https://github.com/X-PLUG/mPLUG-HalOwl/tree/main/hacl.

Yiyang Zhou, Haoqin Tu, Zijun Wang et al.

We propose MIRA, a new benchmark designed to evaluate models in scenarios where generating intermediate visual images is essential for successful reasoning. Unlike traditional CoT methods that rely solely on text, tasks in MIRA require models to generate and utilize intermediate images - such as sketches, structural diagrams, or path drawings - to guide their reasoning process. This setup closely mirrors how humans solve complex problems through "drawing to think". To solve this, MIRA focuses on tasks that are intrinsically challenging and involve complex structures, spatial relationships, or reasoning steps that are difficult to express through language alone. To ensure that our evaluation data is of high-quality, we include 546 multimodal problems, annotated with intermediate visual images and final answers. We also propose a unified evaluation protocol for MIRA that spans three levels of evaluation input: direct input with image and question only, text-only CoT input with image and thinking prompts, and Visual-CoT input with both annotated image clues and textual thinking prompts. To probe the upper bound of model capacity on our benchmark, we also report pass@k and majority voting accuracies under different k settings. Experimental results show that existing multimodal large language models, including strongest private models as well as strong open-weight models, perform poorly when relying solely on textual prompts. However, when intermediate visual cues are provided, model performance improves consistently, yielding an average relative gain of 33.7% across all models and tasks. We also probe the upper bound by expanding the search space and designing textual prompts aligned with Visual-CoT, but both yield only limited improvements compared to our Visual-CoT setting. These results underscore the critical role of imagined visual information in enabling successful reasoning on MIRA.

Zilong Huang, Qinghao Ye, Bingyi Kang et al.

We introduce SuperClass, a super simple classification method for vision-language pre-training on image-text data. Unlike its contrastive counterpart CLIP who contrast with a text encoder, SuperClass directly utilizes tokenized raw text as supervised classification labels, without the need for additional text filtering or selection. Due to the absence of the text encoding as contrastive target, SuperClass does not require a text encoder and does not need to maintain a large batch size as CLIP does. SuperClass demonstrated superior performance on various downstream tasks, including classic computer vision benchmarks and vision language downstream tasks. We further explored the scaling behavior of SuperClass on model size, training length, or data size, and reported encouraging results and comparisons to CLIP. https://github.com/x-cls/superclass

Dong Guo, Faming Wu, Feida Zhu et al. · pku

We present Seed1.5-VL, a vision-language foundation model designed to advance general-purpose multimodal understanding and reasoning. Seed1.5-VL is composed with a 532M-parameter vision encoder and a Mixture-of-Experts (MoE) LLM of 20B active parameters. Despite its relatively compact architecture, it delivers strong performance across a wide spectrum of public VLM benchmarks and internal evaluation suites, achieving the state-of-the-art performance on 38 out of 60 public benchmarks. Moreover, in agent-centric tasks such as GUI control and gameplay, Seed1.5-VL outperforms leading multimodal systems, including OpenAI CUA and Claude 3.7. Beyond visual and video understanding, it also demonstrates strong reasoning abilities, making it particularly effective for multimodal reasoning challenges such as visual puzzles. We believe these capabilities will empower broader applications across diverse tasks. In this report, we mainly provide a comprehensive review of our experiences in building Seed1.5-VL across model design, data construction, and training at various stages, hoping that this report can inspire further research. Seed1.5-VL is now accessible at https://www.volcengine.com/ (Volcano Engine Model ID: doubao-1-5-thinking-vision-pro-250428)

Yunhao Fang, Weihao Yu, Shu Zhong et al.

Long-sequence modeling faces a fundamental trade-off between the efficiency of compressive fixed-size memory in RNN-like models and the fidelity of lossless growing memory in attention-based Transformers. Inspired by the Multi-Store Model in cognitive science, we introduce a memory framework of artificial neural networks. Our method maintains a sliding window of the Transformer's KV cache as lossless short-term memory, while a learnable module termed Artificial Hippocampus Network (AHN) recurrently compresses out-of-window information into a fixed-size compact long-term memory. To validate this framework, we instantiate AHNs using modern RNN-like architectures, including Mamba2, DeltaNet, and Gated DeltaNet. Extensive experiments on long-context benchmarks LV-Eval and InfiniteBench demonstrate that AHN-augmented models consistently outperform sliding window baselines and achieve performance comparable or even superior to full-attention models, while substantially reducing computational and memory requirements. For instance, augmenting the Qwen2.5-3B-Instruct with AHNs reduces inference FLOPs by 40.5% and memory cache by 74.0%, while improving its average score on LV-Eval (128k sequence length) from 4.41 to 5.88. Code is available at: https://github.com/ByteDance-Seed/AHN.

Xu Yang, Jiawei Peng, Zihua Wang et al.

We propose to Transform Scene Graphs (TSG) into more descriptive captions. In TSG, we apply multi-head attention (MHA) to design the Graph Neural Network (GNN) for embedding scene graphs. After embedding, different graph embeddings contain diverse specific knowledge for generating the words with different part-of-speech, e.g., object/attribute embedding is good for generating nouns/adjectives. Motivated by this, we design a Mixture-of-Expert (MOE)-based decoder, where each expert is built on MHA, for discriminating the graph embeddings to generate different kinds of words. Since both the encoder and decoder are built based on the MHA, as a result, we construct a homogeneous encoder-decoder unlike the previous heterogeneous ones which usually apply Fully-Connected-based GNN and LSTM-based decoder. The homogeneous architecture enables us to unify the training configuration of the whole model instead of specifying different training strategies for diverse sub-networks as in the heterogeneous pipeline, which releases the training difficulty. Extensive experiments on the MS-COCO captioning benchmark validate the effectiveness of our TSG. The code is in: https://github.com/GaryJiajia/TSG.

Zhipeng Chen, Xiaobo Qin, Youbin Wu et al.

Reinforcement learning with verifiable rewards (RLVR), which typically adopts Pass@1 as the reward, has faced the issues in balancing exploration and exploitation, causing policies to prefer conservative actions, converging to a local optimum. Identifying an appropriate reward metric is therefore crucial. Regarding the prior work, although Pass@k has been used in evaluation, its connection to LLM exploration ability in RLVR remains largely overlooked. To investigate this, we first use Pass@k as the reward to train the policy model (i.e., $\textbf{Pass@k Training}$), and observe the improvement on its exploration ability. Next, we derive an analytical solution for the advantage of Pass@k Training, leading to an efficient and effective process. Building on this, our analysis reveals that exploration and exploitation are not inherently conflicting objectives, while they can mutually enhance each other. Moreover, Pass@k Training with analytical derivation essentially involves directly designing the advantage function. Inspired by this, we preliminarily explore the advantage design for RLVR, showing promising results and highlighting a potential future direction.

Haoming Wang, Haoyang Zou, Huatong Song et al. · pku

The development of autonomous agents for graphical user interfaces (GUIs) presents major challenges in artificial intelligence. While recent advances in native agent models have shown promise by unifying perception, reasoning, action, and memory through end-to-end learning, open problems remain in data scalability, multi-turn reinforcement learning (RL), the limitations of GUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a native GUI-centered agent model that addresses these challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts. Empirical evaluation demonstrates that UI-TARS-2 achieves significant improvements over its predecessor UI-TARS-1.5. On GUI benchmarks, it reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines such as Claude and OpenAI agents. In game environments, it attains a mean normalized score of 59.8 across a 15-game suite-roughly 60% of human-level performance-and remains competitive with frontier proprietary models (e.g., OpenAI o3) on LMGame-Bench. Additionally, the model can generalize to long-horizon information-seeking tasks and software engineering benchmarks, highlighting its robustness across diverse agent tasks. Detailed analyses of training dynamics further provide insights into achieving stability and efficiency in large-scale agent RL. These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.

Qinghao Ye, Xianhan Zeng, Fu Li et al.

Image captioning has long been a pivotal task in visual understanding, with recent advancements in vision-language models (VLMs) significantly enhancing the ability to generate detailed image captions. However, the evaluation of detailed image captioning remains underexplored due to outdated evaluation metrics and coarse annotations. In this paper, we introduce DeCapBench along with a novel metric, DCScore, specifically designed for detailed captioning tasks. DCScore evaluates hallucinations and fine-grained comprehensiveness by deconstructing responses into the smallest self-sufficient units, termed primitive information units, and assessing them individually. Our evaluation shows that DCScore aligns more closely with human judgment than other rule-based or model-based metrics. Concurrently, DeCapBench exhibits a high correlation with VLM arena results on descriptive tasks, surpassing existing benchmarks for vision-language models. Additionally, we present an automatic fine-grained feedback collection method, FeedQuill, for preference optimization based on our advanced metric, showing robust generalization capabilities across auto-generated preference data. Extensive experiments on multiple VLMs demonstrate that our method not only significantly reduces hallucinations but also enhances performance across various benchmarks, achieving superior detail captioning performance while surpassing GPT-4o.

Chaoya Jiang, Wei ye, Haiyang Xu et al.

Self-supervised Multi-modal Contrastive Learning (SMCL) remarkably advances modern Vision-Language Pre-training (VLP) models by aligning visual and linguistic modalities. Due to noises in web-harvested text-image pairs, however, scaling up training data volume in SMCL presents considerable obstacles in terms of computational cost and data inefficiency. To improve data efficiency in VLP, we propose Text-aware Image Mixing (TiMix), which integrates mix-based data augmentation techniques into SMCL, yielding significant performance improvements without significantly increasing computational overhead. We provide a theoretical analysis of TiMixfrom a mutual information (MI) perspective, showing that mixed data samples for cross-modal contrastive learning implicitly serve as a regularizer for the contrastive loss. The experimental results demonstrate that TiMix exhibits a comparable performance on downstream tasks, even with a reduced amount of training data and shorter training time, when benchmarked against existing methods. This work empirically and theoretically demonstrates the potential of data mixing for data-efficient and computationally viable VLP, benefiting broader VLP model adoption in practical scenarios.

Haowei Liu, Yaya Shi, Haiyang Xu et al.

In video-text retrieval, most existing methods adopt the dual-encoder architecture for fast retrieval, which employs two individual encoders to extract global latent representations for videos and texts. However, they face challenges in capturing fine-grained semantic concepts. In this work, we propose the UNIFY framework, which learns lexicon representations to capture fine-grained semantics and combines the strengths of latent and lexicon representations for video-text retrieval. Specifically, we map videos and texts into a pre-defined lexicon space, where each dimension corresponds to a semantic concept. A two-stage semantics grounding approach is proposed to activate semantically relevant dimensions and suppress irrelevant dimensions. The learned lexicon representations can thus reflect fine-grained semantics of videos and texts. Furthermore, to leverage the complementarity between latent and lexicon representations, we propose a unified learning scheme to facilitate mutual learning via structure sharing and self-distillation. Experimental results show our UNIFY framework largely outperforms previous video-text retrieval methods, with 4.8% and 8.2% Recall@1 improvement on MSR-VTT and DiDeMo respectively.

Haowei Liu, Yaya Shi, Haiyang Xu et al.

In vision-language pre-training (VLP), masked image modeling (MIM) has recently been introduced for fine-grained cross-modal alignment. However, in most existing methods, the reconstruction targets for MIM lack high-level semantics, and text is not sufficiently involved in masked modeling. These two drawbacks limit the effect of MIM in facilitating cross-modal semantic alignment. In this work, we propose a semantics-enhanced cross-modal MIM framework (SemMIM) for vision-language representation learning. Specifically, to provide more semantically meaningful supervision for MIM, we propose a local semantics enhancing approach, which harvest high-level semantics from global image features via self-supervised agreement learning and transfer them to local patch encodings by sharing the encoding space. Moreover, to achieve deep involvement of text during the entire MIM process, we propose a text-guided masking strategy and devise an efficient way of injecting textual information in both masked modeling and reconstruction target acquisition. Experimental results validate that our method improves the effectiveness of the MIM task in facilitating cross-modal semantic alignment. Compared to previous VLP models with similar model size and data scale, our SemMIM model achieves state-of-the-art or competitive performance on multiple downstream vision-language tasks.

Xi Zhou, Qinghao Ye, Xiaolin Yang et al.

Based on CT and MRI images acquired from normal pressure hydrocephalus (NPH) patients, using machine learning methods, we aim to establish a multi-modal and high-performance automatic ventricle segmentation method to achieve efficient and accurate automatic measurement of the ventricular volume. First, we extract the brain CT and MRI images of 143 definite NPH patients. Second, we manually label the ventricular volume (VV) and intracranial volume (ICV). Then, we use machine learning method to extract features and establish automatic ventricle segmentation model. Finally, we verify the reliability of the model and achieved automatic measurement of VV and ICV. In CT images, the Dice similarity coefficient (DSC), Intraclass Correlation Coefficient (ICC), Pearson correlation, and Bland-Altman analysis of the automatic and manual segmentation result of the VV were 0.95, 0.99, 0.99, and 4.2$\pm$2.6 respectively. The results of ICV were 0.96, 0.99, 0.99, and 6.0$\pm$3.8 respectively. The whole process takes 3.4$\pm$0.3 seconds. In MRI images, the DSC, ICC, Pearson correlation, and Bland-Altman analysis of the automatic and manual segmentation result of the VV were 0.94, 0.99, 0.99, and 2.0$\pm$0.6 respectively. The results of ICV were 0.93, 0.99, 0.99, and 7.9$\pm$3.8 respectively. The whole process took 1.9$\pm$0.1 seconds. We have established a multi-modal and high-performance automatic ventricle segmentation method to achieve efficient and accurate automatic measurement of the ventricular volume of NPH patients. This can help clinicians quickly and accurately understand the situation of NPH patient's ventricles.

Yingchao Pan, Ouhan Huang, Qinghao Ye et al.

Video summarization aims to automatically generate a diverse and concise summary which is useful in large-scale video processing. Most of the methods tend to adopt self-attention mechanism across video frames, which fails to model the diversity of video frames. To alleviate this problem, we revisit the pairwise similarity measurement in self-attention mechanism and find that the existing inner-product affinity leads to discriminative features rather than diversified features. In light of this phenomenon, we propose global diverse attention which uses the squared Euclidean distance instead to compute the affinities. Moreover, we model the local contextual information by novel local contextual attention to remove the redundancy in the video. By combining these two attention mechanisms, a video SUMmarization model with Diversified Contextual Attention scheme is developed, namely SUM-DCA. Extensive experiments are conducted on benchmark data sets to verify the effectiveness and the superiority of SUM-DCA in terms of F-score and rank-based evaluation without any bells and whistles.

Qinghao Ye, Yuan Gao, Weiping Ding et al.

The world is currently experiencing an ongoing pandemic of an infectious disease named coronavirus disease 2019 (i.e., COVID-19), which is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Computed Tomography (CT) plays an important role in assessing the severity of the infection and can also be used to identify those symptomatic and asymptomatic COVID-19 carriers. With a surge of the cumulative number of COVID-19 patients, radiologists are increasingly stressed to examine the CT scans manually. Therefore, an automated 3D CT scan recognition tool is highly in demand since the manual analysis is time-consuming for radiologists and their fatigue can cause possible misjudgment. However, due to various technical specifications of CT scanners located in different hospitals, the appearance of CT images can be significantly different leading to the failure of many automated image recognition approaches. The multi-domain shift problem for the multi-center and multi-scanner studies is therefore nontrivial that is also crucial for a dependable recognition and critical for reproducible and objective diagnosis and prognosis. In this paper, we proposed a COVID-19 CT scan recognition model namely coronavirus information fusion and diagnosis network (CIFD-Net) that can efficiently handle the multi-domain shift problem via a new robust weakly supervised learning paradigm. Our model can resolve the problem of different appearance in CT scan images reliably and efficiently while attaining higher accuracy compared to other state-of-the-art methods.

Qinghao Ye, Jun Xia, Guang Yang

Artificial Intelligence (AI) has made leapfrogs in development across all the industrial sectors especially when deep learning has been introduced. Deep learning helps to learn the behaviour of an entity through methods of recognising and interpreting patterns. Despite its limitless potential, the mystery is how deep learning algorithms make a decision in the first place. Explainable AI (XAI) is the key to unlocking AI and the black-box for deep learning. XAI is an AI model that is programmed to explain its goals, logic, and decision making so that the end users can understand. The end users can be domain experts, regulatory agencies, managers and executive board members, data scientists, users that use AI, with or without awareness, or someone who is affected by the decisions of an AI model. Chest CT has emerged as a valuable tool for the clinical diagnostic and treatment management of the lung diseases associated with COVID-19. AI can support rapid evaluation of CT scans to differentiate COVID-19 findings from other lung diseases. However, how these AI tools or deep learning algorithms reach such a decision and which are the most influential features derived from these neural networks with typically deep layers are not clear. The aim of this study is to propose and develop XAI strategies for COVID-19 classification models with an investigation of comparison. The results demonstrate promising quantification and qualitative visualisations that can further enhance the clinician's understanding and decision making with more granular information from the results given by the learned XAI models.

Guang Yang, Qinghao Ye, Jun Xia

Explainable Artificial Intelligence (XAI) is an emerging research topic of machine learning aimed at unboxing how AI systems' black-box choices are made. This research field inspects the measures and models involved in decision-making and seeks solutions to explain them explicitly. Many of the machine learning algorithms can not manifest how and why a decision has been cast. This is particularly true of the most popular deep neural network approaches currently in use. Consequently, our confidence in AI systems can be hindered by the lack of explainability in these black-box models. The XAI becomes more and more crucial for deep learning powered applications, especially for medical and healthcare studies, although in general these deep neural networks can return an arresting dividend in performance. The insufficient explainability and transparency in most existing AI systems can be one of the major reasons that successful implementation and integration of AI tools into routine clinical practice are uncommon. In this study, we first surveyed the current progress of XAI and in particular its advances in healthcare applications. We then introduced our solutions for XAI leveraging multi-modal and multi-centre data fusion, and subsequently validated in two showcases following real clinical scenarios. Comprehensive quantitative and qualitative analyses can prove the efficacy of our proposed XAI solutions, from which we can envisage successful applications in a broader range of clinical questions.

Ping Li, Qinghao Ye, Luming Zhang et al.

Video summarization is an effective way to facilitate video searching and browsing. Most of existing systems employ encoder-decoder based recurrent neural networks, which fail to explicitly diversify the system-generated summary frames while requiring intensive computations. In this paper, we propose an efficient convolutional neural network architecture for video SUMmarization via Global Diverse Attention called SUM-GDA, which adapts attention mechanism in a global perspective to consider pairwise temporal relations of video frames. Particularly, the GDA module has two advantages: 1) it models the relations within paired frames as well as the relations among all pairs, thus capturing the global attention across all frames of one video; 2) it reflects the importance of each frame to the whole video, leading to diverse attention on these frames. Thus, SUM-GDA is beneficial for generating diverse frames to form satisfactory video summary. Extensive experiments on three data sets, i.e., SumMe, TVSum, and VTW, have demonstrated that SUM-GDA and its extension outperform other competing state-of-the-art methods with remarkable improvements. In addition, the proposed models can be run in parallel with significantly less computational costs, which helps the deployment in highly demanding applications.

Qinghao Ye, Kaiyuan Hu, Yizhe Wang

With the improvements of Los Angeles in many aspects, people in mounting numbers tend to live or travel to the city. The primary objective of this paper is to apply a set of methods for the time series analysis of traffic accidents in Los Angeles in the past few years. The number of traffic accidents, collected from 2010 to 2019 monthly reveals that the traffic accident happens seasonally and increasing with fluctuation. This paper utilizes the ensemble methods to combine several different methods to model the data from various perspectives, which can lead to better forecasting accuracy. The IMA(1, 1), ETS(A, N, A), and two models with Fourier items are failed in independence assumption checking. However, the Online Gradient Descent (OGD) model generated by the ensemble method shows the perfect fit in the data modeling, which is the state-of-the-art model among our candidate models. Therefore, it can be easier to accurately forecast future traffic accidents based on previous data through our model, which can help designers to make better plans.