Yuxuan Wang

Tao Sun, Mattia Segu, Janis Postels et al. · eth-zurich, stanford

Adapting to a continuously evolving environment is a safety-critical challenge inevitably faced by all autonomous driving systems. Existing image and video driving datasets, however, fall short of capturing the mutable nature of the real world. In this paper, we introduce the largest multi-task synthetic dataset for autonomous driving, SHIFT. It presents discrete and continuous shifts in cloudiness, rain and fog intensity, time of day, and vehicle and pedestrian density. Featuring a comprehensive sensor suite and annotations for several mainstream perception tasks, SHIFT allows investigating the degradation of a perception system performance at increasing levels of domain shift, fostering the development of continuous adaptation strategies to mitigate this problem and assess model robustness and generality. Our dataset and benchmark toolkit are publicly available at www.vis.xyz/shift.

Stan Weixian Lei, Difei Gao, Jay Zhangjie Wu et al.

VQA is an ambitious task aiming to answer any image-related question. However, in reality, it is hard to build such a system once for all since the needs of users are continuously updated, and the system has to implement new functions. Thus, Continual Learning (CL) ability is a must in developing advanced VQA systems. Recently, a pioneer work split a VQA dataset into disjoint answer sets to study this topic. However, CL on VQA involves not only the expansion of label sets (new Answer sets). It is crucial to study how to answer questions when deploying VQA systems to new environments (new Visual scenes) and how to answer questions requiring new functions (new Question types). Thus, we propose CLOVE, a benchmark for Continual Learning On Visual quEstion answering, which contains scene- and function-incremental settings for the two aforementioned CL scenarios. In terms of methodology, the main difference between CL on VQA and classification is that the former additionally involves expanding and preventing forgetting of reasoning mechanisms, while the latter focusing on class representation. Thus, we propose a real-data-free replay-based method tailored for CL on VQA, named Scene Graph as Prompt for Symbolic Replay. Using a piece of scene graph as a prompt, it replays pseudo scene graphs to represent the past images, along with correlated QA pairs. A unified VQA model is also proposed to utilize the current and replayed data to enhance its QA ability. Finally, experimental results reveal challenges in CLOVE and demonstrate the effectiveness of our method. The dataset and code will be available at https://github.com/showlab/CLVQA.

Xueliang Zhao, Yuxuan Wang, Chongyang Tao et al. · pku

We study video-grounded dialogue generation, where a response is generated based on the dialogue context and the associated video. The primary challenges of this task lie in (1) the difficulty of integrating video data into pre-trained language models (PLMs) which presents obstacles to exploiting the power of large-scale pre-training; and (2) the necessity of taking into account the complementarity of various modalities throughout the reasoning process. Although having made remarkable progress in video-grounded dialogue generation, existing methods still fall short when it comes to integrating with PLMs in a way that allows information from different modalities to complement each other. To alleviate these issues, we first propose extracting pertinent information from videos and turning it into reasoning paths that are acceptable to PLMs. Additionally, we propose a multi-agent reinforcement learning method to collaboratively perform reasoning on different modalities (i.e., video and dialogue context). Empirical experiment results on two public datasets indicate that the proposed model can significantly outperform state-of-the-art models by large margins on both automatic and human evaluations.

Jianghui Wang, Yuxuan Wang, Dongyan Zhao et al. · pku

We introduce MoviePuzzle, a novel challenge that targets visual narrative reasoning and holistic movie understanding. Despite the notable progress that has been witnessed in the realm of video understanding, most prior works fail to present tasks and models to address holistic video understanding and the innate visual narrative structures existing in long-form videos. To tackle this quandary, we put forth MoviePuzzle task that amplifies the temporal feature learning and structure learning of video models by reshuffling the shot, frame, and clip layers of movie segments in the presence of video-dialogue information. We start by establishing a carefully refined dataset based on MovieNet by dissecting movies into hierarchical layers and randomly permuting the orders. Besides benchmarking the MoviePuzzle with prior arts on movie understanding, we devise a Hierarchical Contrastive Movie Clustering (HCMC) model that considers the underlying structure and visual semantic orders for movie reordering. Specifically, through a pairwise and contrastive learning approach, we train models to predict the correct order of each layer. This equips them with the knack for deciphering the visual narrative structure of movies and handling the disorder lurking in video data. Experiments show that our approach outperforms existing state-of-the-art methods on the \MoviePuzzle benchmark, underscoring its efficacy.

Xiaowen Sun, Jiazhan Feng, Yuxuan Wang et al. · pku

A picture is worth a thousand words, thus, it is crucial for conversational agents to understand, perceive, and effectively respond with pictures. However, we find that directly employing conventional image generation techniques is inadequate for conversational agents to produce image responses effectively. In this paper, we focus on the innovative dialog-to-image generation task, where the model synthesizes a high-resolution image aligned with the given dialog context as a response. To tackle this problem, we design a tailored fine-tuning approach on the top of state-of-the-art text-to-image generation models to fully exploit the structural and semantic features in dialog context during image generation. Concretely, we linearize the dialog context with specific indicators to maintain the dialog structure, and employ in-domain data to alleviate the style mismatch between dialog-to-image and conventional image generation tasks. Empirical results on PhotoChat and MMDialog Corpus show that our approach brings consistent and remarkable improvement with 3 state-of-the-art pre-trained text-to-image generation backbones.

Yuxuan Wang, Yijun Liu, Fei Yu et al.

Despite the rapid development of Chinese vision-language models (VLMs), most existing Chinese vision-language (VL) datasets are constructed on Western-centric images from existing English VL datasets. The cultural bias in the images makes these datasets unsuitable for evaluating VLMs in Chinese culture. To remedy this issue, we present a new Chinese Vision- Language Understanding Evaluation (CVLUE) benchmark dataset, where the selection of object categories and images is entirely driven by Chinese native speakers, ensuring that the source images are representative of Chinese culture. The benchmark contains four distinct VL tasks ranging from image-text retrieval to visual question answering, visual grounding and visual dialogue. We present a detailed statistical analysis of CVLUE and provide a baseline performance analysis with several open-source multilingual VLMs on CVLUE and its English counterparts to reveal their performance gap between English and Chinese. Our in-depth category-level analysis reveals a lack of Chinese cultural knowledge in existing VLMs. We also find that fine-tuning on Chinese culture-related VL datasets effectively enhances VLMs' understanding of Chinese culture.

Tong Wu, Junzhe Shen, Zixia Jia et al. · pku

Generating ultra-long sequences with large language models (LLMs) has become increasingly crucial but remains a highly time-intensive task, particularly for sequences up to 100K tokens. While traditional speculative decoding methods exist, simply extending their generation limits fails to accelerate the process and can be detrimental. Through an in-depth analysis, we identify three major challenges hindering efficient generation: frequent model reloading, dynamic key-value (KV) management and repetitive generation. To address these issues, we introduce TOKENSWIFT, a novel framework designed to substantially accelerate the generation process of ultra-long sequences while maintaining the target model's inherent quality. Experimental results demonstrate that TOKENSWIFT achieves over 3 times speedup across models of varying scales (1.5B, 7B, 8B, 14B) and architectures (MHA, GQA). This acceleration translates to hours of time savings for ultra-long sequence generation, establishing TOKENSWIFT as a scalable and effective solution at unprecedented lengths. Code can be found at https://github.com/bigai-nlco/TokenSwift.

Yueqian Wang, Yuxuan Wang, Kai Chen et al. · pku

Recently we have witnessed the rapid development of video question answering models. However, most models can only handle simple videos in terms of temporal reasoning, and their performance tends to drop when answering temporal-reasoning questions on long and informative videos. To tackle this problem we propose STAIR, a Spatial-Temporal Reasoning model with Auditable Intermediate Results for video question answering. STAIR is a neural module network, which contains a program generator to decompose a given question into a hierarchical combination of several sub-tasks, and a set of lightweight neural modules to complete each of these sub-tasks. Though neural module networks are already widely studied on image-text tasks, applying them to videos is a non-trivial task, as reasoning on videos requires different abilities. In this paper, we define a set of basic video-text sub-tasks for video question answering and design a set of lightweight modules to complete them. Different from most prior works, modules of STAIR return intermediate outputs specific to their intentions instead of always returning attention maps, which makes it easier to interpret and collaborate with pre-trained models. We also introduce intermediate supervision to make these intermediate outputs more accurate. We conduct extensive experiments on several video question answering datasets under various settings to show STAIR's performance, explainability, compatibility with pre-trained models, and applicability when program annotations are not available. Code: https://github.com/yellow-binary-tree/STAIR

Yuxuan Wang, Feng Dong, Jinchao Zhu

Salient object detection (SOD) focuses on distinguishing the most conspicuous objects in the scene. However, most related works are based on RGB images, which lose massive useful information. Accordingly, with the maturity of thermal technology, RGB-T (RGB-Thermal) multi-modality tasks attain more and more attention. Thermal infrared images carry important information which can be used to improve the accuracy of SOD prediction. To accomplish it, the methods to integrate multi-modal information and suppress noises are critical. In this paper, we propose a novel network called Interactive Context-Aware Network (ICANet). It contains three modules that can effectively perform the cross-modal and cross-scale fusions. We design a Hybrid Feature Fusion (HFF) module to integrate the features of two modalities, which utilizes two types of feature extraction. The Multi-Scale Attention Reinforcement (MSAR) and Upper Fusion (UF) blocks are responsible for the cross-scale fusion that converges different levels of features and generate the prediction maps. We also raise a novel Context-Aware Multi-Supervised Network (CAMSNet) to calculate the content loss between the prediction and the ground truth (GT). Experiments prove that our network performs favorably against the state-of-the-art RGB-T SOD methods.

Stan Weixian Lei, Difei Gao, Yuxuan Wang et al.

It is still a pipe dream that personal AI assistants on the phone and AR glasses can assist our daily life in addressing our questions like ``how to adjust the date for this watch?'' and ``how to set its heating duration? (while pointing at an oven)''. The queries used in conventional tasks (i.e. Video Question Answering, Video Retrieval, Moment Localization) are often factoid and based on pure text. In contrast, we present a new task called Task-oriented Question-driven Video Segment Retrieval (TQVSR). Each of our questions is an image-box-text query that focuses on affordance of items in our daily life and expects relevant answer segments to be retrieved from a corpus of instructional video-transcript segments. To support the study of this TQVSR task, we construct a new dataset called AssistSR. We design novel guidelines to create high-quality samples. This dataset contains 3.2k multimodal questions on 1.6k video segments from instructional videos on diverse daily-used items. To address TQVSR, we develop a simple yet effective model called Dual Multimodal Encoders (DME) that significantly outperforms several baseline methods while still having large room for improvement in the future. Moreover, we present detailed ablation analyses. Code and data are available at \url{https://github.com/StanLei52/TQVSR}.

Jiaxiang Dong, Haixu Wu, Yuxuan Wang et al.

Time series pre-training has recently garnered wide attention for its potential to reduce labeling expenses and benefit various downstream tasks. Prior methods are mainly based on pre-training techniques well-acknowledged in vision or language, such as masked modeling and contrastive learning. However, randomly masking time series or calculating series-wise similarity will distort or neglect inherent temporal correlations crucial in time series data. To emphasize temporal correlation modeling, this paper proposes TimeSiam as a simple but effective self-supervised pre-training framework for Time series based on Siamese networks. Concretely, TimeSiam pre-trains Siamese encoders to capture intrinsic temporal correlations between randomly sampled past and current subseries. With a simple data augmentation method (e.g.~masking), TimeSiam can benefit from diverse augmented subseries and learn internal time-dependent representations through a past-to-current reconstruction. Moreover, learnable lineage embeddings are also introduced to distinguish temporal distance between sampled series and further foster the learning of diverse temporal correlations. TimeSiam consistently outperforms extensive advanced pre-training baselines, demonstrating superior forecasting and classification capabilities across 13 standard benchmarks in both intra- and cross-domain scenarios.

Zhitao Zeng, Zhu Zhuo, Xiaojun Jia et al. · pku

Foundation models have achieved transformative success across biomedical domains by enabling holistic understanding of multimodal data. However, their application in surgery remains underexplored. Surgical intelligence presents unique challenges - requiring surgical visual perception, temporal analysis, and reasoning. Existing general-purpose vision-language models fail to address these needs due to insufficient domain-specific supervision and the lack of a large-scale high-quality surgical database. To bridge this gap, we propose SurgVLM, one of the first large vision-language foundation models for surgical intelligence, where this single universal model can tackle versatile surgical tasks. To enable this, we construct a large-scale multimodal surgical database, SurgVLM-DB, comprising over 1.81 million frames with 7.79 million conversations, spanning more than 16 surgical types and 18 anatomical structures. We unify and reorganize 23 public datasets across 10 surgical tasks, followed by standardizing labels and doing hierarchical vision-language alignment to facilitate comprehensive coverage of gradually finer-grained surgical tasks, from visual perception, temporal analysis, to high-level reasoning. Building upon this comprehensive dataset, we propose SurgVLM, which is built upon Qwen2.5-VL, and undergoes instruction tuning to 10+ surgical tasks. We further construct a surgical multimodal benchmark, SurgVLM-Bench, for method evaluation. SurgVLM-Bench consists of 6 popular and widely-used datasets in surgical domain, covering several crucial downstream tasks. Based on SurgVLM-Bench, we evaluate the performance of our SurgVLM (3 SurgVLM variants: SurgVLM-7B, SurgVLM-32B, and SurgVLM-72B), and conduct comprehensive comparisons with 14 mainstream commercial VLMs (e.g., GPT-4o, Gemini 2.0 Flash, Qwen2.5-Max).

Zhitao Zeng, Guojian Yuan, Junyuan Mao et al.

Accurate temporal prediction is the bridge between comprehensive scene understanding and embodied artificial intelligence. However, predicting multiple fine-grained states of a scene at multiple temporal scales is difficult for vision-language models. We formalize the Multi-Scale Temporal Prediction (MSTP) task in general and surgical scenes by decomposing multi-scale into two orthogonal dimensions: the temporal scale, forecasting states of humans and surgery at varying look-ahead intervals, and the state scale, modeling a hierarchy of states in general and surgical scenes. For example, in general scenes, states of contact relationships are finer-grained than states of spatial relationships. In surgical scenes, medium-level steps are finer-grained than high-level phases yet remain constrained by their encompassing phase. To support this unified task, we introduce the first MSTP Benchmark, featuring synchronized annotations across multiple state scales and temporal scales. We further propose a method, Incremental Generation and Multi-agent Collaboration (IG-MC), which integrates two key innovations. First, we present a plug-and-play incremental generation module that continuously synthesizes up-to-date visual previews at expanding temporal scales to inform multiple decision-making agents, keeping decisions and generated visuals synchronized and preventing performance degradation as look-ahead intervals lengthen. Second, we present a decision-driven multi-agent collaboration framework for multi-state prediction, comprising generation, initiation, and multi-state assessment agents that dynamically trigger and evaluate prediction cycles to balance global coherence and local fidelity.

Tingle Li, Baihe Huang, Xiaobin Zhuang et al.

Generating accurate sounds for complex audio-visual scenes is challenging, especially in the presence of multiple objects and sound sources. In this paper, we propose an {\em interactive object-aware audio generation} model that grounds sound generation in user-selected visual objects within images. Our method integrates object-centric learning into a conditional latent diffusion model, which learns to associate image regions with their corresponding sounds through multi-modal attention. At test time, our model employs image segmentation to allow users to interactively generate sounds at the {\em object} level. We theoretically validate that our attention mechanism functionally approximates test-time segmentation masks, ensuring the generated audio aligns with selected objects. Quantitative and qualitative evaluations show that our model outperforms baselines, achieving better alignment between objects and their associated sounds. Project page: https://tinglok.netlify.app/files/avobject/

Yuxuan Wang, Zilong Zheng, Xueliang Zhao et al.

Video-grounded dialogue understanding is a challenging problem that requires machine to perceive, parse and reason over situated semantics extracted from weakly aligned video and dialogues. Most existing benchmarks treat both modalities the same as a frame-independent visual understanding task, while neglecting the intrinsic attributes in multimodal dialogues, such as scene and topic transitions. In this paper, we present Video-grounded Scene&Topic AwaRe dialogue (VSTAR) dataset, a large scale video-grounded dialogue understanding dataset based on 395 TV series. Based on VSTAR, we propose two benchmarks for video-grounded dialogue understanding: scene segmentation and topic segmentation, and one benchmark for video-grounded dialogue generation. Comprehensive experiments are performed on these benchmarks to demonstrate the importance of multimodal information and segments in video-grounded dialogue understanding and generation.

Max W. Y. Lam, Qiao Tian, Tang Li et al.

Recent progress in music generation has been remarkably advanced by the state-of-the-art MusicLM, which comprises a hierarchy of three LMs, respectively, for semantic, coarse acoustic, and fine acoustic modelings. Yet, sampling with the MusicLM requires processing through these LMs one by one to obtain the fine-grained acoustic tokens, making it computationally expensive and prohibitive for a real-time generation. Efficient music generation with a quality on par with MusicLM remains a significant challenge. In this paper, we present MeLoDy (M for music; L for LM; D for diffusion), an LM-guided diffusion model that generates music audios of state-of-the-art quality meanwhile reducing 95.7% or 99.6% forward passes in MusicLM, respectively, for sampling 10s or 30s music. MeLoDy inherits the highest-level LM from MusicLM for semantic modeling, and applies a novel dual-path diffusion (DPD) model and an audio VAE-GAN to efficiently decode the conditioning semantic tokens into waveform. DPD is proposed to simultaneously model the coarse and fine acoustics by incorporating the semantic information into segments of latents effectively via cross-attention at each denoising step. Our experimental results suggest the superiority of MeLoDy, not only in its practical advantages on sampling speed and infinitely continuable generation, but also in its state-of-the-art musicality, audio quality, and text correlation. Our samples are available at https://Efficient-MeLoDy.github.io/.

Guangyi Yang, Yang Zhan., Yuxuan Wang

The goal in a blind image quality assessment (BIQA) model is to simulate the process of evaluating images by human eyes and accurately assess the quality of the image. Although many approaches effectively identify degradation, they do not fully consider the semantic content in images resulting in distortion. In order to fill this gap, we propose a deep adaptive superpixel-based network, namely DSN-IQA, to assess the quality of image based on multi-scale and superpixel segmentation. The DSN-IQA can adaptively accept arbitrary scale images as input images, making the assessment process similar to human perception. The network uses two models to extract multi-scale semantic features and generate a superpixel adjacency map. These two elements are united together via feature fusion to accurately predict image quality. Experimental results on different benchmark databases demonstrate that our algorithm is highly competitive with other approaches when assessing challenging authentic image databases. Also, due to adaptive deep superpixel-based network, our model accurately assesses images with complicated distortion, much like the human eye.

Bochen Li, Yuxuan Wang, Zhiyao Duan

Separating a song into vocal and accompaniment components is an active research topic, and recent years witnessed an increased performance from supervised training using deep learning techniques. We propose to apply the visual information corresponding to the singers' vocal activities to further improve the quality of the separated vocal signals. The video frontend model takes the input of mouth movement and fuses it into the feature embeddings of an audio-based separation framework. To facilitate the network to learn audiovisual correlation of singing activities, we add extra vocal signals irrelevant to the mouth movement to the audio mixture during training. We create two audiovisual singing performance datasets for training and evaluation, respectively, one curated from audition recordings on the Internet, and the other recorded in house. The proposed method outperforms audio-based methods in terms of separation quality on most test recordings. This advantage is especially pronounced when there are backing vocals in the accompaniment, which poses a great challenge for audio-only methods.

Yu Gu, Xiang Yin, Yonghui Rao et al.

This paper presents ByteSing, a Chinese singing voice synthesis (SVS) system based on duration allocated Tacotron-like acoustic models and WaveRNN neural vocoders. Different from the conventional SVS models, the proposed ByteSing employs Tacotron-like encoder-decoder structures as the acoustic models, in which the CBHG models and recurrent neural networks (RNNs) are explored as encoders and decoders respectively. Meanwhile an auxiliary phoneme duration prediction model is utilized to expand the input sequence, which can enhance the model controllable capacity, model stability and tempo prediction accuracy. WaveRNN neural vocoders are also adopted as neural vocoders to further improve the voice quality of synthesized songs. Both objective and subjective experimental results prove that the SVS method proposed in this paper can produce quite natural, expressive and high-fidelity songs by improving the pitch and spectrogram prediction accuracy and the models using attention mechanism can achieve best performance.