Wanggui He

Peng Zhang, Guanghao Zhang, Wanggui He et al.

Recent video multimodal large language models (MLLMs) increasingly couple step-by-step reasoning with on-demand visual evidence retrieval, allowing models to revisit relevant video segments during inference. However, two structural gaps remain in existing thinking-with-video systems. (i) Sampling density is not a learnable decision: existing methods may let the model decide where to look, but the per-window frame rate is largely fixed. As a result, fine-grained evidence is often recovered through repeated retrieval calls, which increases inference context length and training difficulty. (ii) Retrieval and answer generation are usually optimized with a single trajectory-level advantage, so the "where to look" tokens and the "how to answer" tokens receive the same credit even when one is correct and the other is not. To address these gaps, we present DynFrame, a framework that emits the temporal window and the sampling density as native tokens within a single autoregressive pass. This learnable span-density retrieval enables acquiring multi-granularity evidence with a single retrieval step. Based on the above tokenized retrieval interface, we further introduce Segment-Decoupled GRPO (SD-GRPO), which splits each rollout at the retrieval boundary and assigns role-specific token-level advantages, separately crediting the sampling decision and the answer. Trained on the curated DM-CoT-74k and DM-RL-45k, DynFrame-4B is competitive with strong 7B-8B baselines across six benchmarks (NExT-GQA, Charades-STA, ActivityNet-MR, Video-MME, MLVU, LVBench), and DynFrame-8B sets new state-of-the-art on most metrics. Code is available at https://github.com/zhangguanghao523/DynFrame.

Fangxun Shu, Yue Liao, Le Zhuo et al.

We introduce LLaVA-MoD, a novel framework designed to enable the efficient training of small-scale Multimodal Language Models (s-MLLM) by distilling knowledge from large-scale MLLM (l-MLLM). Our approach tackles two fundamental challenges in MLLM distillation. First, we optimize the network structure of s-MLLM by integrating a sparse Mixture of Experts (MoE) architecture into the language model, striking a balance between computational efficiency and model expressiveness. Second, we propose a progressive knowledge transfer strategy to ensure comprehensive knowledge migration. This strategy begins with mimic distillation, where we minimize the Kullback-Leibler (KL) divergence between output distributions to enable the student model to emulate the teacher network's understanding. Following this, we introduce preference distillation via Direct Preference Optimization (DPO), where the key lies in treating l-MLLM as the reference model. During this phase, the s-MLLM's ability to discriminate between superior and inferior examples is significantly enhanced beyond l-MLLM, leading to a better student that surpasses its teacher, particularly in hallucination benchmarks. Extensive experiments demonstrate that LLaVA-MoD outperforms existing models across various multimodal benchmarks while maintaining a minimal number of activated parameters and low computational costs. Remarkably, LLaVA-MoD, with only 2B activated parameters, surpasses Qwen-VL-Chat-7B by an average of 8.8% across benchmarks, using merely 0.3% of the training data and 23% trainable parameters. These results underscore LLaVA-MoD's ability to effectively distill comprehensive knowledge from its teacher model, paving the way for the development of more efficient MLLMs. The code will be available on: https://github.com/shufangxun/LLaVA-MoD.

Tianwei Lin, Jiang Liu, Wenqiao Zhang et al.

While Parameter-Efficient Fine-Tuning (PEFT) methods like LoRA have effectively addressed GPU memory constraints during fine-tuning, their performance often falls short, especially in multidimensional task scenarios. To address this issue, one straightforward solution is to introduce task-specific LoRA modules as domain experts, leveraging the modeling of multiple experts' capabilities and thus enhancing the general capability of multi-task learning. Despite promising, these additional components often add complexity to the training and inference process, contravening the efficient characterization of PEFT designed for. Considering this, we introduce an innovative PEFT method, TeamLoRA, consisting of a collaboration and competition module for experts, and thus achieving the right balance of effectiveness and efficiency: (i) For collaboration, a novel knowledge-sharing and -organizing mechanism is devised to appropriately reduce the scale of matrix operations, thereby boosting the training and inference speed. (ii) For competition, we propose leveraging a game-theoretic interaction mechanism for experts, encouraging experts to transfer their domain-specific knowledge while facing diverse downstream tasks, and thus enhancing the performance. By doing so, TeamLoRA elegantly connects the experts as a "Team" with internal collaboration and competition, enabling a faster and more accurate PEFT paradigm for multi-task learning. To validate the superiority of TeamLoRA, we curate a comprehensive multi-task evaluation(CME) benchmark to thoroughly assess the capability of multi-task learning. Experiments conducted on our CME and other benchmarks indicate the effectiveness and efficiency of TeamLoRA. Our project is available at https://github.com/Lin-Tianwei/TeamLoRA.

Jinlong Liu, Wanggui He, Peng Zhang et al.

Reinforcement learning (RL) can improve the prompt following capability of text-to-image (T2I) models, yet obtaining high-quality reward signals remains challenging: CLIP Score is too coarse-grained, while VLM-based reward models (e.g., RewardDance) require costly human-annotated preference data and additional fine-tuning. We propose PromptEcho, a reward construction method that requires \emph{no} annotation and \emph{no} reward model training. Given a generated image and a guiding query, PromptEcho computes the token-level cross-entropy loss of a frozen VLM with the original prompt as the label, directly extracting the image-text alignment knowledge encoded during VLM pretraining. The reward is deterministic, computationally efficient, and improves automatically as stronger open-source VLMs become available. For evaluation, we develop DenseAlignBench, a benchmark of concept-rich dense captions for rigorously testing prompt following capability. Experimental results on two state-of-the-art T2I models (Z-Image and QwenImage-2512) demonstrate that PromptEcho achieves substantial improvements on DenseAlignBench (+26.8pp / +16.2pp net win rate), along with consistent gains on GenEval, DPG-Bench, and TIIFBench without any task-specific training. Ablation studies confirm that PromptEcho comprehensively outperforms inference-based scoring with the same VLM, and that reward quality scales with VLM size. We will open-source the trained models and the DenseAlignBench.

Yunze Tong, Mushui Liu, Canyu Zhao et al.

Deploying GRPO on Flow Matching models has proven effective for text-to-image generation. However, existing paradigms typically propagate an outcome-based reward to all preceding denoising steps without distinguishing the local effect of each step. Moreover, current group-wise ranking mainly compares trajectories at matched timesteps and ignores within-trajectory dependencies, where certain early denoising actions can affect later states via delayed, implicit interactions. We propose TurningPoint-GRPO (TP-GRPO), a GRPO framework that alleviates step-wise reward sparsity and explicitly models long-term effects within the denoising trajectory. TP-GRPO makes two key innovations: (i) it replaces outcome-based rewards with step-level incremental rewards, providing a dense, step-aware learning signal that better isolates each denoising action's "pure" effect, and (ii) it identifies turning points-steps that flip the local reward trend and make subsequent reward evolution consistent with the overall trajectory trend-and assigns these actions an aggregated long-term reward to capture their delayed impact. Turning points are detected solely via sign changes in incremental rewards, making TP-GRPO efficient and hyperparameter-free. Extensive experiments also demonstrate that TP-GRPO exploits reward signals more effectively and consistently improves generation. Demo code is available at https://github.com/YunzeTong/TurningPoint-GRPO.

Wanggui He, Siming Fu, Mushui Liu et al.

Auto-regressive models have made significant progress in the realm of language generation, yet they do not perform on par with diffusion models in the domain of image synthesis. In this work, we introduce MARS, a novel framework for T2I generation that incorporates a specially designed Semantic Vision-Language Integration Expert (SemVIE). This innovative component integrates pre-trained LLMs by independently processing linguistic and visual information, freezing the textual component while fine-tuning the visual component. This methodology preserves the NLP capabilities of LLMs while imbuing them with exceptional visual understanding. Building upon the powerful base of the pre-trained Qwen-7B, MARS stands out with its bilingual generative capabilities corresponding to both English and Chinese language prompts and the capacity for joint image and text generation. The flexibility of this framework lends itself to migration towards any-to-any task adaptability. Furthermore, MARS employs a multi-stage training strategy that first establishes robust image-text alignment through complementary bidirectional tasks and subsequently concentrates on refining the T2I generation process, significantly augmenting text-image synchrony and the granularity of image details. Notably, MARS requires only 9% of the GPU days needed by SD1.5, yet it achieves remarkable results across a variety of benchmarks, illustrating the training efficiency and the potential for swift deployment in various applications.

Haoyuan Li, Hao Jiang, Tianke Zhang et al.

Training AI models has always been challenging, especially when there is a need for custom models to provide personalized services. Algorithm engineers often face a lengthy process to iteratively develop models tailored to specific business requirements, making it even more difficult for non-experts. The quest for high-quality and efficient model development, along with the emergence of Large Language Model (LLM) Agents, has become a key focus in the industry. Leveraging the powerful analytical, planning, and decision-making capabilities of LLM, we propose a TrainerAgent system comprising a multi-agent framework including Task, Data, Model and Server agents. These agents analyze user-defined tasks, input data, and requirements (e.g., accuracy, speed), optimizing them comprehensively from both data and model perspectives to obtain satisfactory models, and finally deploy these models as online service. Experimental evaluations on classical discriminative and generative tasks in computer vision and natural language processing domains demonstrate that our system consistently produces models that meet the desired criteria. Furthermore, the system exhibits the ability to critically identify and reject unattainable tasks, such as fantastical scenarios or unethical requests, ensuring robustness and safety. This research presents a significant advancement in achieving desired models with increased efficiency and quality as compared to traditional model development, facilitated by the integration of LLM-powered analysis, decision-making, and execution capabilities, as well as the collaboration among four agents. We anticipate that our work will contribute to the advancement of research on TrainerAgent in both academic and industry communities, potentially establishing it as a new paradigm for model development in the field of AI.

Tianwei Lin, Wenqiao Zhang, Sijing Li et al.

We present HealthGPT, a powerful Medical Large Vision-Language Model (Med-LVLM) that integrates medical visual comprehension and generation capabilities within a unified autoregressive paradigm. Our bootstrapping philosophy is to progressively adapt heterogeneous comprehension and generation knowledge to pre-trained large language models (LLMs). This is achieved through a novel heterogeneous low-rank adaptation (H-LoRA) technique, which is complemented by a tailored hierarchical visual perception approach and a three-stage learning strategy. To effectively learn the HealthGPT, we devise a comprehensive medical domain-specific comprehension and generation dataset called VL-Health. Experimental results demonstrate exceptional performance and scalability of HealthGPT in medical visual unified tasks. Our project can be accessed at https://github.com/DCDmllm/HealthGPT.

Wenyi Xiao, Zechuan Wang, Leilei Gan et al. · mit

With the rapid advancement of large language models (LLMs), aligning policy models with human preferences has become increasingly critical. Direct Preference Optimization (DPO) has emerged as a promising approach for alignment, acting as an RL-free alternative to Reinforcement Learning from Human Feedback (RLHF). Despite DPO's various advancements and inherent limitations, an in-depth review of these aspects is currently lacking in the literature. In this work, we present a comprehensive review of the challenges and opportunities in DPO, covering theoretical analyses, variants, relevant preference datasets, and applications. Specifically, we categorize recent studies on DPO based on key research questions to provide a thorough understanding of DPO's current landscape. Additionally, we propose several future research directions to offer insights on model alignment for the research community. An updated collection of relevant papers can be found on https://github.com/Mr-Loevan/DPO-Survey.

Qihan Huang, Weilong Dai, Jinlong Liu et al.

MLLM reasoning has drawn widespread research for its excellent problem-solving capability. Current reasoning methods fall into two types: PRM, which supervises the intermediate reasoning steps, and ORM, which supervises the final results. Recently, DeepSeek-R1 has challenged the traditional view that PRM outperforms ORM, which demonstrates strong generalization performance using an ORM method (i.e., GRPO). However, current MLLM's GRPO algorithms still struggle to handle challenging and complex multimodal reasoning tasks (e.g., mathematical reasoning). In this work, we reveal two problems that impede the performance of GRPO on the MLLM: Low data utilization and Text-bias. Low data utilization refers to that GRPO cannot acquire positive rewards to update the MLLM on difficult samples, and text-bias is a phenomenon that the MLLM bypasses image condition and solely relies on text condition for generation after GRPO training. To tackle these problems, this work proposes Hint-GRPO that improves data utilization by adaptively providing hints for samples of varying difficulty, and text-bias calibration that mitigates text-bias by calibrating the token prediction logits with image condition in test-time. Experiment results on three base MLLMs across eleven datasets demonstrate that our proposed methods advance the reasoning capability of original MLLM by a large margin, exhibiting superior performance to existing MLLM reasoning methods. Our code is available at https://github.com/hqhQAQ/Hint-GRPO.

Qihan Huang, Weilong Dai, Jinlong Liu et al.

Finetuning-free personalized image generation can synthesize customized images without test-time finetuning, attracting wide research interest owing to its high efficiency. Current finetuning-free methods simply adopt a single training stage with a simple image reconstruction task, and they typically generate low-quality images inconsistent with the reference images during test-time. To mitigate this problem, inspired by the recent DPO (i.e., direct preference optimization) technique, this work proposes an additional training stage to improve the pre-trained personalized generation models. However, traditional DPO only determines the overall superiority or inferiority of two samples, which is not suitable for personalized image generation because the generated images are commonly inconsistent with the reference images only in some local image patches. To tackle this problem, this work proposes PatchDPO that estimates the quality of image patches within each generated image and accordingly trains the model. To this end, PatchDPO first leverages the pre-trained vision model with a proposed self-supervised training method to estimate the patch quality. Next, PatchDPO adopts a weighted training approach to train the model with the estimated patch quality, which rewards the image patches with high quality while penalizing the image patches with low quality. Experiment results demonstrate that PatchDPO significantly improves the performance of multiple pre-trained personalized generation models, and achieves state-of-the-art performance on both single-object and multi-object personalized image generation. Our code is available at https://github.com/hqhQAQ/PatchDPO.

Jiang Liu, Bolin Li, Haoyuan Li et al.

Efficient multimodal large language models (EMLLMs), in contrast to multimodal large language models (MLLMs), reduce model size and computational costs and are often deployed on resource-constrained devices. However, due to data privacy concerns, existing open-source EMLLMs rarely have access to private domain-specific data during the pre-training process, making them difficult to directly apply in device-specific domains, such as certain business scenarios. To address this weakness, this paper focuses on the efficient adaptation of EMLLMs to private domains, specifically in two areas: 1) how to reduce data requirements, and 2) how to avoid parameter fine-tuning. Specifically, we propose a tun\textbf{\underline{I}}ng-free, a\textbf{\underline{D}}aptiv\textbf{\underline{E}}, univers\textbf{\underline{AL}} \textbf{\underline{Prompt}} Optimization Framework, abbreviated as \textit{\textbf{\ourmethod{}}} which consists of two stages: 1) Predefined Prompt, based on the reinforcement searching strategy, generate a prompt optimization strategy tree to acquire optimization priors; 2) Prompt Reflection initializes the prompt based on optimization priors, followed by self-reflection to further search and refine the prompt. By doing so, \ourmethod{} elegantly generates the ``ideal prompts'' for processing private domain-specific data. Note that our method requires no parameter fine-tuning and only a small amount of data to quickly adapt to the data distribution of private data. Extensive experiments across multiple tasks demonstrate that our proposed \ourmethod{} significantly improves both efficiency and performance compared to baselines.

Fangtai Wu, Mushui Liu, Weijie He et al.

The unified autoregressive (AR) model excels at multimodal understanding and generation, but its potential for customized image generation remains underexplored. Existing customized generation methods rely on full fine-tuning or adapters, making them costly and prone to overfitting or catastrophic forgetting. In this paper, we propose \textbf{CoAR}, a novel framework for injecting subject concepts into the unified AR models while keeping all pre-trained parameters completely frozen. CoAR learns effective, specific subject representations with only a minimal number of parameters using a Layerwise Multimodal Context Learning strategy. To address overfitting and language drift, we further introduce regularization that preserves the pre-trained distribution and anchors context tokens to improve subject fidelity and re-contextualization. Additionally, CoAR supports training-free subject customization in a user-provided style. Experiments demonstrate that CoAR achieves superior performance on both subject-driven personalization and style personalization, while delivering significant gains in computational and memory efficiency. Notably, CoAR tunes less than \textbf{0.05\%} of the parameters while achieving competitive performance compared to recent Proxy-Tuning. Code: https://github.com/KZF-kzf/CoAR

Wenyi Xiao, Ziwei Huang, Leilei Gan et al.

The rapidly developing Large Vision Language Models (LVLMs) have shown notable capabilities on a range of multi-modal tasks, but still face the hallucination phenomena where the generated texts do not align with the given contexts, significantly restricting the usages of LVLMs. Most previous work detects and mitigates hallucination at the coarse-grained level or requires expensive annotation (e.g., labeling by proprietary models or human experts). To address these issues, we propose detecting and mitigating hallucinations in LVLMs via fine-grained AI feedback. The basic idea is that we generate a small-size sentence-level hallucination annotation dataset by proprietary models, whereby we train a hallucination detection model which can perform sentence-level hallucination detection, covering primary hallucination types (i.e., object, attribute, and relationship). Then, we propose a detect-then-rewrite pipeline to automatically construct preference dataset for training hallucination mitigating model. Furthermore, we propose differentiating the severity of hallucinations, and introducing a Hallucination Severity-Aware Direct Preference Optimization (HSA-DPO) for mitigating hallucination in LVLMs by incorporating the severity of hallucinations into preference learning. Extensive experiments demonstrate the effectiveness of our method.

Shangzhe Di, Zhelun Yu, Guanghao Zhang et al.

We propose ReKV, a novel training-free approach that enables efficient streaming video question-answering (StreamingVQA), by seamlessly integrating with existing Video Large Language Models (Video-LLMs). Traditional VideoQA systems struggle with long videos, as they must process entire videos before responding to queries, and repeat this process for each new question. In contrast, our approach analyzes long videos in a streaming manner, allowing for prompt responses as soon as user queries are received. Building on a common Video-LLM, we first incorporate a sliding-window attention mechanism, ensuring that input frames attend to a limited number of preceding frames, thereby reducing computational overhead. To prevent information loss, we store processed video key-value caches (KV-Caches) in RAM and disk, reloading them into GPU memory as needed. Additionally, we introduce a retrieval method that leverages an external retriever or the parameters within Video-LLMs to retrieve only query-relevant KV-Caches, ensuring both efficiency and accuracy in question answering. ReKV enables the separation of video encoding and question-answering across different processes and GPUs, significantly enhancing the efficiency of StreamingVQA. Through comprehensive experimentation, we validate the efficacy and practicality of our approach, which significantly boosts efficiency and enhances applicability over existing VideoQA models.

Yi Wang, Mushui Liu, Wanggui He et al.

Unified generative models have shown remarkable performance in text and image generation. For image synthesis tasks, they adopt straightforward text-to-image (T2I) generation. However, direct T2I generation limits the models in handling complex compositional instructions, which frequently occur in real-world scenarios. Although this issue is vital, existing works mainly focus on improving the basic image generation capability of the models. While such improvements help to some extent, they still fail to adequately resolve the problem. Inspired by Chain of Thought (CoT) solving complex problems step by step, this work aims to introduce CoT into unified generative models to address the challenges of complex image generation that direct T2I generation cannot effectively solve, thereby endowing models with enhanced image generation ability. To achieve this, we first propose Functionality-oriented eXperts (FoXperts), an expert-parallel architecture in our model FoX, which assigns experts by function. FoXperts disentangles potential conflicts in mainstream modality-oriented designs and provides a solid foundation for CoT. When introducing CoT, the first question is how to design it for complex image generation. To this end, we emulate a human-like artistic workflow -- planning, acting, reflection, and correction -- and propose the Multimodal Chain of Thought (MCoT) approach, as the data involves both text and image. To address the subsequent challenge -- designing an effective MCoT training paradigm -- we develop a multi-task joint training scheme that equips the model with all capabilities required for each MCoT step in a disentangled manner. This paradigm avoids the difficulty of collecting consistent multi-step data tuples. Extensive experiments show that FoX consistently outperforms existing unified models on various T2I benchmarks, delivering notable improvements in complex image generation.

Guanghao Zhang, Tao Zhong, Yan Xia et al.

While previous multimodal slow-thinking methods have demonstrated remarkable success in single-image understanding scenarios, their effectiveness becomes fundamentally constrained when extended to more complex multi-image comprehension tasks. This limitation stems from their predominant reliance on text-based intermediate reasoning processes. While for human, when engaging in sophisticated multi-image analysis, they typically perform two complementary cognitive operations: (1) continuous cross-image visual comparison through region-of-interest matching, and (2) dynamic memorization of critical visual concepts throughout the reasoning chain. Motivated by these observations, we propose the Complex Multi-Modal Chain-of-Thought (CMMCoT) framework, a multi-step reasoning framework that mimics human-like "slow thinking" for multi-image understanding. Our approach incorporates two key innovations: 1. The construction of interleaved multimodal multi-step reasoning chains, which utilize critical visual region tokens, extracted from intermediate reasoning steps, as supervisory signals. This mechanism not only facilitates comprehensive cross-modal understanding but also enhances model interpretability. 2. The introduction of a test-time memory augmentation module that expands the model reasoning capacity during inference while preserving parameter efficiency. Furthermore, to facilitate research in this direction, we have curated a novel multi-image slow-thinking dataset. Extensive experiments demonstrate the effectiveness of our model.

D. She, Mushui Liu, Jingxuan Pang et al.

Customized generation has achieved significant progress in image synthesis, yet personalized video generation remains challenging due to temporal inconsistencies and quality degradation. In this paper, we introduce CustomVideoX, an innovative framework leveraging the video diffusion transformer for personalized video generation from a reference image. CustomVideoX capitalizes on pre-trained video networks by exclusively training the LoRA parameters to extract reference features, ensuring both efficiency and adaptability. To facilitate seamless interaction between the reference image and video content, we propose 3D Reference Attention, which enables direct and simultaneous engagement of reference image features with all video frames across spatial and temporal dimensions. To mitigate the excessive influence of reference image features and textual guidance on generated video content during inference, we implement the Time-Aware Reference Attention Bias (TAB) strategy, dynamically modulating reference bias over different time steps. Additionally, we introduce the Entity Region-Aware Enhancement (ERAE) module, aligning highly activated regions of key entity tokens with reference feature injection by adjusting attention bias. To thoroughly evaluate personalized video generation, we establish a new benchmark, VideoBench, comprising over 50 objects and 100 prompts for extensive assessment. Experimental results show that CustomVideoX significantly outperforms existing methods in terms of video consistency and quality.

Ziwei Huang, Wanggui He, Quanyu Long et al.

Evaluating the quality of synthesized images remains a significant challenge in the development of text-to-image (T2I) generation. Most existing studies in this area primarily focus on evaluating text-image alignment, image quality, and object composition capabilities, with comparatively fewer studies addressing the evaluation of the factuality of T2I models, particularly when the concepts involved are knowledge-intensive. To mitigate this gap, we present T2I-FactualBench in this work - the largest benchmark to date in terms of the number of concepts and prompts specifically designed to evaluate the factuality of knowledge-intensive concept generation. T2I-FactualBench consists of a three-tiered knowledge-intensive text-to-image generation framework, ranging from the basic memorization of individual knowledge concepts to the more complex composition of multiple knowledge concepts. We further introduce a multi-round visual question answering (VQA) based evaluation framework to assess the factuality of three-tiered knowledge-intensive text-to-image generation tasks. Experiments on T2I-FactualBench indicate that current state-of-the-art (SOTA) T2I models still leave significant room for improvement.

Xierui Wang, Siming Fu, Qihan Huang et al.

Recent advancements in text-to-image generation models have dramatically enhanced the generation of photorealistic images from textual prompts, leading to an increased interest in personalized text-to-image applications, particularly in multi-subject scenarios. However, these advances are hindered by two main challenges: firstly, the need to accurately maintain the details of each referenced subject in accordance with the textual descriptions; and secondly, the difficulty in achieving a cohesive representation of multiple subjects in a single image without introducing inconsistencies. To address these concerns, our research introduces the MS-Diffusion framework for layout-guided zero-shot image personalization with multi-subjects. This innovative approach integrates grounding tokens with the feature resampler to maintain detail fidelity among subjects. With the layout guidance, MS-Diffusion further improves the cross-attention to adapt to the multi-subject inputs, ensuring that each subject condition acts on specific areas. The proposed multi-subject cross-attention orchestrates harmonious inter-subject compositions while preserving the control of texts. Comprehensive quantitative and qualitative experiments affirm that this method surpasses existing models in both image and text fidelity, promoting the development of personalized text-to-image generation. The project page is https://MS-Diffusion.github.io.

Chenyi Lei, Shixian Luo, Yong Liu et al.

The pre-trained neural models have recently achieved impressive performances in understanding multimodal content. However, it is still very challenging to pre-train neural models for video and language understanding, especially for Chinese video-language data, due to the following reasons. Firstly, existing video-language pre-training algorithms mainly focus on the co-occurrence of words and video frames, but ignore other valuable semantic and structure information of video-language content, e.g., sequential order and spatiotemporal relationships. Secondly, there exist conflicts between video sentence alignment and other proxy tasks. Thirdly, there is a lack of large-scale and high-quality Chinese video-language datasets (e.g., including 10 million unique videos), which are the fundamental success conditions for pre-training techniques. In this work, we propose a novel video-language understanding framework named VICTOR, which stands for VIdeo-language understanding via Contrastive mulTimOdal pRe-training. Besides general proxy tasks such as masked language modeling, VICTOR constructs several novel proxy tasks under the contrastive learning paradigm, making the model be more robust and able to capture more complex multimodal semantic and structural relationships from different perspectives. VICTOR is trained on a large-scale Chinese video-language dataset, including over 10 million complete videos with corresponding high-quality textual descriptions. We apply the pre-trained VICTOR model to a series of downstream applications and demonstrate its superior performances, comparing against the state-of-the-art pre-training methods such as VideoBERT and UniVL. The codes and trained checkpoints will be publicly available to nourish further developments of the research community.