Jingjing Liu

Xinlong Wang, Xiaosong Zhang, Zhengxiong Luo et al. · tsinghua

While next-token prediction is considered a promising path towards artificial general intelligence, it has struggled to excel in multimodal tasks, which are still dominated by diffusion models (e.g., Stable Diffusion) and compositional approaches (e.g., CLIP combined with LLMs). In this paper, we introduce Emu3, a new suite of state-of-the-art multimodal models trained solely with next-token prediction. By tokenizing images, text, and videos into a discrete space, we train a single transformer from scratch on a mixture of multimodal sequences. Emu3 outperforms several well-established task-specific models in both generation and perception tasks, surpassing flagship models such as SDXL and LLaVA-1.6, while eliminating the need for diffusion or compositional architectures. Emu3 is also capable of generating high-fidelity video via predicting the next token in a video sequence. We simplify complex multimodal model designs by converging on a singular focus: tokens, unlocking great potential for scaling both during training and inference. Our results demonstrate that next-token prediction is a promising path towards building general multimodal intelligence beyond language. We open-source key techniques and models to support further research in this direction.

Bu Jin, Xinyu Liu, Yupeng Zheng et al.

End-to-end autonomous driving has great potential in the transportation industry. However, the lack of transparency and interpretability of the automatic decision-making process hinders its industrial adoption in practice. There have been some early attempts to use attention maps or cost volume for better model explainability which is difficult for ordinary passengers to understand. To bridge the gap, we propose an end-to-end transformer-based architecture, ADAPT (Action-aware Driving cAPtion Transformer), which provides user-friendly natural language narrations and reasoning for each decision making step of autonomous vehicular control and action. ADAPT jointly trains both the driving caption task and the vehicular control prediction task, through a shared video representation. Experiments on BDD-X (Berkeley DeepDrive eXplanation) dataset demonstrate state-of-the-art performance of the ADAPT framework on both automatic metrics and human evaluation. To illustrate the feasibility of the proposed framework in real-world applications, we build a novel deployable system that takes raw car videos as input and outputs the action narrations and reasoning in real time. The code, models and data are available at https://github.com/jxbbb/ADAPT.

Quan Sun, Qiying Yu, Yufeng Cui et al. · tsinghua

We present Emu, a Transformer-based multimodal foundation model, which can seamlessly generate images and texts in multimodal context. This omnivore model can take in any single-modality or multimodal data input indiscriminately (e.g., interleaved image, text and video) through a one-model-for-all autoregressive training process. First, visual signals are encoded into embeddings, and together with text tokens form an interleaved input sequence. Emu is then end-to-end trained with a unified objective of classifying the next text token or regressing the next visual embedding in the multimodal sequence. This versatile multimodality empowers the exploration of diverse pretraining data sources at scale, such as videos with interleaved frames and text, webpages with interleaved images and text, as well as web-scale image-text pairs and video-text pairs. Emu can serve as a generalist multimodal interface for both image-to-text and text-to-image tasks, and supports in-context image and text generation. Across a broad range of zero-shot/few-shot tasks including image captioning, visual question answering, video question answering and text-to-image generation, Emu demonstrates superb performance compared to state-of-the-art large multimodal models. Extended capabilities such as multimodal assistants via instruction tuning are also demonstrated with impressive performance.

Jianxiong Li, Xianyuan Zhan, Haoran Xu et al. · tsinghua

In offline reinforcement learning (RL), one detrimental issue to policy learning is the error accumulation of deep Q function in out-of-distribution (OOD) areas. Unfortunately, existing offline RL methods are often over-conservative, inevitably hurting generalization performance outside data distribution. In our study, one interesting observation is that deep Q functions approximate well inside the convex hull of training data. Inspired by this, we propose a new method, DOGE (Distance-sensitive Offline RL with better GEneralization). DOGE marries dataset geometry with deep function approximators in offline RL, and enables exploitation in generalizable OOD areas rather than strictly constraining policy within data distribution. Specifically, DOGE trains a state-conditioned distance function that can be readily plugged into standard actor-critic methods as a policy constraint. Simple yet elegant, our algorithm enjoys better generalization compared to state-of-the-art methods on D4RL benchmarks. Theoretical analysis demonstrates the superiority of our approach to existing methods that are solely based on data distribution or support constraints.

Hideaki Takahashi, Jingjing Liu, Yang Liu

Federated Learning with Model Distillation (FedMD) is a nascent collaborative learning paradigm, where only output logits of public datasets are transmitted as distilled knowledge, instead of passing on private model parameters that are susceptible to gradient inversion attacks, a known privacy risk in federated learning. In this paper, we found that even though sharing output logits of public datasets is safer than directly sharing gradients, there still exists a substantial risk of data exposure caused by carefully designed malicious attacks. Our study shows that a malicious server can inject a PLI (Paired-Logits Inversion) attack against FedMD and its variants by training an inversion neural network that exploits the confidence gap between the server and client models. Experiments on multiple facial recognition datasets validate that under FedMD-like schemes, by using paired server-client logits of public datasets only, the malicious server is able to reconstruct private images on all tested benchmarks with a high success rate.

Qiying Yu, Quan Sun, Xiaosong Zhang et al. · tsinghua

Large multimodal models demonstrate remarkable generalist ability to perform diverse multimodal tasks in a zero-shot manner. Large-scale web-based image-text pairs contribute fundamentally to this success, but suffer from excessive noise. Recent studies use alternative captions synthesized by captioning models and have achieved notable benchmark performance. However, our experiments reveal significant Scalability Deficiency and World Knowledge Loss issues in models trained with synthetic captions, which have been largely obscured by their initial benchmark success. Upon closer examination, we identify the root cause as the overly-simplified language structure and lack of knowledge details in existing synthetic captions. To provide higher-quality and more scalable multimodal pretraining data, we propose CapsFusion, an advanced framework that leverages large language models to consolidate and refine information from both web-based image-text pairs and synthetic captions. Extensive experiments show that CapsFusion captions exhibit remarkable all-round superiority over existing captions in terms of model performance (e.g., 18.8 and 18.3 improvements in CIDEr score on COCO and NoCaps), sample efficiency (requiring 11-16 times less computation than baselines), world knowledge depth, and scalability. These effectiveness, efficiency and scalability advantages position CapsFusion as a promising candidate for future scaling of LMM training.

Shuyu Miao, Lin Zheng, Jingjing Liu et al.

The label-free model evaluation aims to predict the model performance on various test sets without relying on ground truths. The main challenge of this task is the absence of labels in the test data, unlike in classical supervised model evaluation. This paper presents our solutions for the 1st DataCV Challenge of the Visual Dataset Understanding workshop at CVPR 2023. Firstly, we propose a novel method called K-means Clustering Based Feature Consistency Alignment (KCFCA), which is tailored to handle the distribution shifts of various datasets. KCFCA utilizes the K-means algorithm to cluster labeled training sets and unlabeled test sets, and then aligns the cluster centers with feature consistency. Secondly, we develop a dynamic regression model to capture the relationship between the shifts in distribution and model accuracy. Thirdly, we design an algorithm to discover the outlier model factors, eliminate the outlier models, and combine the strengths of multiple autoeval models. On the DataCV Challenge leaderboard, our approach secured 2nd place with an RMSE of 6.8526. Our method significantly improved over the best baseline method by 36\% (6.8526 vs. 10.7378). Furthermore, our method achieves a relatively more robust and optimal single model performance on the validation dataset.

Haozhen Gong, Xiaozhong Ji, Yuansen Liu et al.

MLLMs MLLMs are beginning to appear in clinical workflows, but their ability to perform complex medical reasoning remains unclear. We present Med-CMR, a fine-grained Medical Complex Multimodal Reasoning benchmark. Med-CMR distinguishes from existing counterparts by three core features: 1) Systematic capability decomposition, splitting medical multimodal reasoning into fine-grained visual understanding and multi-step reasoning to enable targeted evaluation; 2) Challenging task design, with visual understanding across three key dimensions (small-object detection, fine-detail discrimination, spatial understanding) and reasoning covering four clinically relevant scenarios (temporal prediction, causal reasoning, long-tail generalization, multi-source integration); 3) Broad, high-quality data coverage, comprising 20,653 Visual Question Answering (VQA) pairs spanning 11 organ systems and 12 imaging modalities, validated via a rigorous two-stage (human expert + model-assisted) review to ensure clinical authenticity. We evaluate 18 state-of-the-art MLLMs with Med-CMR, revealing GPT-5 as the top-performing commercial model: 57.81 accuracy on multiple-choice questions (MCQs) and a 48.70 open-ended score, outperforming Gemini 2.5 Pro (49.87 MCQ accuracy, 45.98 open-ended score) and leading open-source model Qwen3-VL-235B-A22B (49.34 MCQ accuracy, 42.62 open-ended score). However, specialized medical MLLMs do not reliably outperform strong general models, and long-tail generalization emerges as the dominant failure mode. Med-CMR thus provides a stress test for visual-reasoning integration and rare-case robustness in medical MLLMs, and a rigorous yardstick for future clinical systems.

Tongda Xu, Qian Zhang, Yanghao Li et al.

We propose conditional perceptual quality, an extension of the perceptual quality defined in \citet{blau2018perception}, by conditioning it on user defined information. Specifically, we extend the original perceptual quality $d(p_{X},p_{\hat{X}})$ to the conditional perceptual quality $d(p_{X|Y},p_{\hat{X}|Y})$, where $X$ is the original image, $\hat{X}$ is the reconstructed, $Y$ is side information defined by user and $d(.,.)$ is divergence. We show that conditional perceptual quality has similar theoretical properties as rate-distortion-perception trade-off \citep{blau2019rethinking}. Based on these theoretical results, we propose an optimal framework for conditional perceptual quality preserving compression. Experimental results show that our codec successfully maintains high perceptual quality and semantic quality at all bitrate. Besides, by providing a lowerbound of common randomness required, we settle the previous arguments on whether randomness should be incorporated into generator for (conditional) perceptual quality compression. The source code is provided in supplementary material.

Zhe Wang, Siqi Fan, Xiaoliang Huo et al.

In autonomous driving, Vehicle-Infrastructure Cooperative 3D Object Detection (VIC3D) makes use of multi-view cameras from both vehicles and traffic infrastructure, providing a global vantage point with rich semantic context of road conditions beyond a single vehicle viewpoint. Two major challenges prevail in VIC3D: 1) inherent calibration noise when fusing multi-view images, caused by time asynchrony across cameras; 2) information loss when projecting 2D features into 3D space. To address these issues, We propose a novel 3D object detection framework, Vehicles-Infrastructure Multi-view Intermediate fusion (VIMI). First, to fully exploit the holistic perspectives from both vehicles and infrastructure, we propose a Multi-scale Cross Attention (MCA) module that fuses infrastructure and vehicle features on selective multi-scales to correct the calibration noise introduced by camera asynchrony. Then, we design a Camera-aware Channel Masking (CCM) module that uses camera parameters as priors to augment the fused features. We further introduce a Feature Compression (FC) module with channel and spatial compression blocks to reduce the size of transmitted features for enhanced efficiency. Experiments show that VIMI achieves 15.61% overall AP_3D and 21.44% AP_BEV on the new VIC3D dataset, DAIR-V2X-C, significantly outperforming state-of-the-art early fusion and late fusion methods with comparable transmission cost.

Siqi Fan, Zhe Wang, Xiaoliang Huo et al.

Effective BEV object detection on infrastructure can greatly improve traffic scenes understanding and vehicle-toinfrastructure (V2I) cooperative perception. However, cameras installed on infrastructure have various postures, and previous BEV detection methods rely on accurate calibration, which is difficult for practical applications due to inevitable natural factors (e.g., wind and snow). In this paper, we propose a Calibration-free BEV Representation (CBR) network, which achieves 3D detection based on BEV representation without calibration parameters and additional depth supervision. Specifically, we utilize two multi-layer perceptrons for decoupling the features from perspective view to front view and birdeye view under boxes-induced foreground supervision. Then, a cross-view feature fusion module matches features from orthogonal views according to similarity and conducts BEV feature enhancement with front view features. Experimental results on DAIR-V2X demonstrate that CBR achieves acceptable performance without any camera parameters and is naturally not affected by calibration noises. We hope CBR can serve as a baseline for future research addressing practical challenges of infrastructure perception.

Siqi Fan, Zhe Wang, Yan Wang et al.

For semantic segmentation in urban scene understanding, RGB cameras alone often fail to capture a clear holistic topology in challenging lighting conditions. Thermal signal is an informative additional channel that can bring to light the contour and fine-grained texture of blurred regions in low-quality RGB image. Aiming at practical RGB-T (thermal) segmentation, we systematically propose a Spatial-aware Demand-guided Recursive Meshing (SpiderMesh) framework that: 1) proactively compensates inadequate contextual semantics in optically-impaired regions via a demand-guided target masking algorithm; 2) refines multimodal semantic features with recursive meshing to improve pixel-level semantic analysis performance. We further introduce an asymmetric data augmentation technique M-CutOut, and enable semi-supervised learning to fully utilize RGB-T labels only sparsely available in practical use. Extensive experiments on MFNet and PST900 datasets demonstrate that SpiderMesh achieves state-of-the-art performance on standard RGB-T segmentation benchmarks.

Tongda Xu, Ziran Zhu, Dailan He et al.

Idempotence is the stability of image codec to re-compression. At the first glance, it is unrelated to perceptual image compression. However, we find that theoretically: 1) Conditional generative model-based perceptual codec satisfies idempotence; 2) Unconditional generative model with idempotence constraint is equivalent to conditional generative codec. Based on this newfound equivalence, we propose a new paradigm of perceptual image codec by inverting unconditional generative model with idempotence constraints. Our codec is theoretically equivalent to conditional generative codec, and it does not require training new models. Instead, it only requires a pre-trained mean-square-error codec and unconditional generative model. Empirically, we show that our proposed approach outperforms state-of-the-art methods such as HiFiC and ILLM, in terms of Fréchet Inception Distance (FID). The source code is provided in https://github.com/tongdaxu/Idempotence-and-Perceptual-Image-Compression.

Tianyuan Zou, Yang Liu, Peng Li et al. · tsinghua

Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis %that avoid fine-tuning large pre-trained generative models often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contrAstive private data Synthesis via Weighted multiple Pre-trained language models (PLM) framework, named as WASP. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-Q voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models.Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://anonymous.4open.science/r/WASP.

Quan Sun, Yufeng Cui, Xiaosong Zhang et al. · tsinghua

The human ability to easily solve multimodal tasks in context (i.e., with only a few demonstrations or simple instructions), is what current multimodal systems have largely struggled to imitate. In this work, we demonstrate that the task-agnostic in-context learning capabilities of large multimodal models can be significantly enhanced by effective scaling-up. We introduce Emu2, a generative multimodal model with 37 billion parameters, trained on large-scale multimodal sequences with a unified autoregressive objective. Emu2 exhibits strong multimodal in-context learning abilities, even emerging to solve tasks that require on-the-fly reasoning, such as visual prompting and object-grounded generation. The model sets a new record on multiple multimodal understanding tasks in few-shot settings. When instruction-tuned to follow specific instructions, Emu2 further achieves new state-of-the-art on challenging tasks such as question answering benchmarks for large multimodal models and open-ended subject-driven generation. These achievements demonstrate that Emu2 can serve as a base model and general-purpose interface for a wide range of multimodal tasks. Code and models are publicly available to facilitate future research.

Xin Ma, Yang Liu, Jingjing Liu et al.

Large language models (LLMs), although having revolutionized many fields, still suffer from the challenging extrapolation problem, where the inference ability of LLMs sharply declines beyond their max training lengths. In this work, we conduct a theoretical analysis to better understand why No Position Encoding (NoPE) fails outside its effective range, as well as examining the power of Position Encoding (PE) in this context. Our findings reveal that with meticulous weave position, PE can indeed be extended beyond effective range. Our theorems establish that LLMs equipped with weave PE can achieve improved extrapolation performance without additional cost. Furthermore, we introduce a novel weave PE method, Mesa-Extrapolation, which utilizes a chunk-based triangular attention matrix and applies Stair PE to manage the final chunk. This method not only retains competitive performance but also offers substantial benefits such as significantly reduced memory demand and faster inference speed. Extensive experiments validate the effectiveness of Mesa-Extrapolation, demonstrating its potential as a scalable solution to enhancing LLMs applicative reach. Our code is available at \url{https://github.com/soacker/Mesa-Extrapolation}.

He Jiang, Mujtaba Asad, Jingjing Liu et al.

Traditional image detail enhancement is local filter-based or global filter-based. In both approaches, the original image is first divided into the base layer and the detail layer, and then the enhanced image is obtained by amplifying the detail layer. Our method is different, and its innovation lies in the special way to get the image detail layer. The detail layer in our method is obtained by updating the residual features, and the updating mechanism is usually based on searching and matching similar patches. However, due to the diversity of image texture features, perfect matching is often not possible. In this paper, the process of searching and matching is treated as a thermodynamic process, where the Metropolis theorem can minimize the internal energy and get the global optimal solution of this task, that is, to find a more suitable feature for a better detail enhancement performance. Extensive experiments have proven that our algorithm can achieve better results in quantitative metrics testing and visual effects evaluation. The source code can be obtained from the link.

Xiaohan Chen, Yu Cheng, Shuohang Wang et al.

Lottery Ticket Hypothesis (LTH) raises keen attention to identifying sparse trainable subnetworks, or winning tickets, which can be trained in isolation to achieve similar or even better performance compared to the full models. Despite many efforts being made, the most effective method to identify such winning tickets is still Iterative Magnitude-based Pruning (IMP), which is computationally expensive and has to be run thoroughly for every different network. A natural question that comes in is: can we "transform" the winning ticket found in one network to another with a different architecture, yielding a winning ticket for the latter at the beginning, without re-doing the expensive IMP? Answering this question is not only practically relevant for efficient "once-for-all" winning ticket finding, but also theoretically appealing for uncovering inherently scalable sparse patterns in networks. We conduct extensive experiments on CIFAR-10 and ImageNet, and propose a variety of strategies to tweak the winning tickets found from different networks of the same model family (e.g., ResNets). Based on these results, we articulate the Elastic Lottery Ticket Hypothesis (E-LTH): by mindfully replicating (or dropping) and re-ordering layers for one network, its corresponding winning ticket could be stretched (or squeezed) into a subnetwork for another deeper (or shallower) network from the same family, whose performance is nearly the same competitive as the latter's winning ticket directly found by IMP. We have also extensively compared E-LTH with pruning-at-initialization and dynamic sparse training methods, as well as discussed the generalizability of E-LTH to different model families, layer types, and across datasets. Code is available at https://github.com/VITA-Group/ElasticLTH.

Tianlong Chen, Yu Cheng, Zhe Gan et al.

Recent advances in computer vision take advantage of adversarial data augmentation to ameliorate the generalization ability of classification models. Here, we present an effective and efficient alternative that advocates adversarial augmentation on intermediate feature embeddings, instead of relying on computationally-expensive pixel-level perturbations. We propose Adversarial Feature Augmentation and Normalization (A-FAN), which (i) first augments visual recognition models with adversarial features that integrate flexible scales of perturbation strengths, (ii) then extracts adversarial feature statistics from batch normalization, and re-injects them into clean features through feature normalization. We validate the proposed approach across diverse visual recognition tasks with representative backbone networks, including ResNets and EfficientNets for classification, Faster-RCNN for detection, and Deeplab V3+ for segmentation. Extensive experiments show that A-FAN yields consistent generalization improvement over strong baselines across various datasets for classification, detection and segmentation tasks, such as CIFAR-10, CIFAR-100, ImageNet, Pascal VOC2007, Pascal VOC2012, COCO2017, and Cityspaces. Comprehensive ablation studies and detailed analyses also demonstrate that adding perturbations to specific modules and layers of classification/detection/segmentation backbones yields optimal performance. Codes and pre-trained models will be made available at: https://github.com/VITA-Group/CV_A-FAN.

Siqi Sun, Yen-Chun Chen, Linjie Li et al.

Multimodal pre-training has propelled great advancement in vision-and-language research. These large-scale pre-trained models, although successful, fatefully suffer from slow inference speed due to enormous computation cost mainly from cross-modal attention in Transformer architecture. When applied to real-life applications, such latency and computation demand severely deter the practical use of pre-trained models. In this paper, we study Image-text retrieval (ITR), the most mature scenario of V+L application, which has been widely studied even prior to the emergence of recent pre-trained models. We propose a simple yet highly effective approach, LightningDOT that accelerates the inference time of ITR by thousands of times, without sacrificing accuracy. LightningDOT removes the time-consuming cross-modal attention by pre-training on three novel learning objectives, extracting feature indexes offline, and employing instant dot-product matching with further re-ranking, which significantly speeds up retrieval process. In fact, LightningDOT achieves new state of the art across multiple ITR benchmarks such as Flickr30k, COCO and Multi30K, outperforming existing pre-trained models that consume 1000x magnitude of computational hours. Code and pre-training checkpoints are available at https://github.com/intersun/LightningDOT.

Tianlong Chen, Yu Cheng, Zhe Gan et al.

Training generative adversarial networks (GANs) with limited real image data generally results in deteriorated performance and collapsed models. To conquer this challenge, we are inspired by the latest observation, that one can discover independently trainable and highly sparse subnetworks (a.k.a., lottery tickets) from GANs. Treating this as an inductive prior, we suggest a brand-new angle towards data-efficient GAN training: by first identifying the lottery ticket from the original GAN using the small training set of real images; and then focusing on training that sparse subnetwork by re-using the same set. We find our coordinated framework to offer orthogonal gains to existing real image data augmentation methods, and we additionally present a new feature-level augmentation that can be applied together with them. Comprehensive experiments endorse the effectiveness of our proposed framework, across various GAN architectures (SNGAN, BigGAN, and StyleGAN-V2) and diverse datasets (CIFAR-10, CIFAR-100, Tiny-ImageNet, ImageNet, and multiple few-shot generation datasets). Codes are available at: https://github.com/VITA-Group/Ultra-Data-Efficient-GAN-Training.

Jie Lei, Linjie Li, Luowei Zhou et al.

The canonical approach to video-and-language learning (e.g., video question answering) dictates a neural model to learn from offline-extracted dense video features from vision models and text features from language models. These feature extractors are trained independently and usually on tasks different from the target domains, rendering these fixed features sub-optimal for downstream tasks. Moreover, due to the high computational overload of dense video features, it is often difficult (or infeasible) to plug feature extractors directly into existing approaches for easy finetuning. To provide a remedy to this dilemma, we propose a generic framework ClipBERT that enables affordable end-to-end learning for video-and-language tasks, by employing sparse sampling, where only a single or a few sparsely sampled short clips from a video are used at each training step. Experiments on text-to-video retrieval and video question answering on six datasets demonstrate that ClipBERT outperforms (or is on par with) existing methods that exploit full-length videos, suggesting that end-to-end learning with just a few sparsely sampled clips is often more accurate than using densely extracted offline features from full-length videos, proving the proverbial less-is-more principle. Videos in the datasets are from considerably different domains and lengths, ranging from 3-second generic domain GIF videos to 180-second YouTube human activity videos, showing the generalization ability of our approach. Comprehensive ablation studies and thorough analyses are provided to dissect what factors lead to this success. Our code is publicly available at https://github.com/jayleicn/ClipBERT

Wenhu Chen, Zhe Gan, Linjie Li et al.

Neural Module Network (NMN) exhibits strong interpretability and compositionality thanks to its handcrafted neural modules with explicit multi-hop reasoning capability. However, most NMNs suffer from two critical drawbacks: 1) scalability: customized module for specific function renders it impractical when scaling up to a larger set of functions in complex tasks; 2) generalizability: rigid pre-defined module inventory makes it difficult to generalize to unseen functions in new tasks/domains. To design a more powerful NMN architecture for practical use, we propose Meta Module Network (MMN) centered on a novel meta module, which can take in function recipes and morph into diverse instance modules dynamically. The instance modules are then woven into an execution graph for complex visual reasoning, inheriting the strong explainability and compositionality of NMN. With such a flexible instantiation mechanism, the parameters of instance modules are inherited from the central meta module, retaining the same model complexity as the function set grows, which promises better scalability. Meanwhile, as functions are encoded into the embedding space, unseen functions can be readily represented based on its structural similarity with previously observed ones, which ensures better generalizability. Experiments on GQA and CLEVR datasets validate the superiority of MMN over state-of-the-art NMN designs. Synthetic experiments on held-out unseen functions from GQA dataset also demonstrate the strong generalizability of MMN. Our code and model are released in Github https://github.com/wenhuchen/Meta-Module-Network.

Zhe Wang, Siqi Fan, Xiaoliang Huo et al.

In autonomous driving, cooperative perception makes use of multi-view cameras from both vehicles and infrastructure, providing a global vantage point with rich semantic context of road conditions beyond a single vehicle viewpoint. Currently, two major challenges persist in vehicle-infrastructure cooperative 3D (VIC3D) object detection: $1)$ inherent pose errors when fusing multi-view images, caused by time asynchrony across cameras; $2)$ information loss in transmission process resulted from limited communication bandwidth. To address these issues, we propose a novel camera-based 3D detection framework for VIC3D task, Enhanced Multi-scale Image Feature Fusion (EMIFF). To fully exploit holistic perspectives from both vehicles and infrastructure, we propose Multi-scale Cross Attention (MCA) and Camera-aware Channel Masking (CCM) modules to enhance infrastructure and vehicle features at scale, spatial, and channel levels to correct the pose error introduced by camera asynchrony. We also introduce a Feature Compression (FC) module with channel and spatial compression blocks for transmission efficiency. Experiments show that EMIFF achieves SOTA on DAIR-V2X-C datasets, significantly outperforming previous early-fusion and late-fusion methods with comparable transmission costs.

Zhe Wang, Xiaoliang Huo, Siqi Fan et al.

In autonomous driving, The perception capabilities of the ego-vehicle can be improved with roadside sensors, which can provide a holistic view of the environment. However, existing monocular detection methods designed for vehicle cameras are not suitable for roadside cameras due to viewpoint domain gaps. To bridge this gap and Improve ROAdside Monocular 3D object detection, we propose IROAM, a semantic-geometry decoupled contrastive learning framework, which takes vehicle-side and roadside data as input simultaneously. IROAM has two significant modules. In-Domain Query Interaction module utilizes a transformer to learn content and depth information for each domain and outputs object queries. Cross-Domain Query Enhancement To learn better feature representations from two domains, Cross-Domain Query Enhancement decouples queries into semantic and geometry parts and only the former is used for contrastive learning. Experiments demonstrate the effectiveness of IROAM in improving roadside detector's performance. The results validate that IROAM has the capabilities to learn cross-domain information.

Jinyu Yang, Jingjing Liu, Ning Xu et al.

Unsupervised domain adaptation (UDA) aims to transfer the knowledge learnt from a labeled source domain to an unlabeled target domain. Previous work is mainly built upon convolutional neural networks (CNNs) to learn domain-invariant representations. With the recent exponential increase in applying Vision Transformer (ViT) to vision tasks, the capability of ViT in adapting cross-domain knowledge, however, remains unexplored in the literature. To fill this gap, this paper first comprehensively investigates the transferability of ViT on a variety of domain adaptation tasks. Surprisingly, ViT demonstrates superior transferability over its CNNs-based counterparts with a large margin, while the performance can be further improved by incorporating adversarial adaptation. Notwithstanding, directly using CNNs-based adaptation strategies fails to take the advantage of ViT's intrinsic merits (e.g., attention mechanism and sequential image representation) which play an important role in knowledge transfer. To remedy this, we propose an unified framework, namely Transferable Vision Transformer (TVT), to fully exploit the transferability of ViT for domain adaptation. Specifically, we delicately devise a novel and effective unit, which we term Transferability Adaption Module (TAM). By injecting learned transferabilities into attention blocks, TAM compels ViT focus on both transferable and discriminative features. Besides, we leverage discriminative clustering to enhance feature diversity and separation which are undermined during adversarial domain alignment. To verify its versatility, we perform extensive studies of TVT on four benchmarks and the experimental results demonstrate that TVT attains significant improvements compared to existing state-of-the-art UDA methods.

Linjie Li, Jie Lei, Zhe Gan et al.

Most existing video-and-language (VidL) research focuses on a single dataset, or multiple datasets of a single task. In reality, a truly useful VidL system is expected to be easily generalizable to diverse tasks, domains, and datasets. To facilitate the evaluation of such systems, we introduce Video-And-Language Understanding Evaluation (VALUE) benchmark, an assemblage of 11 VidL datasets over 3 popular tasks: (i) text-to-video retrieval; (ii) video question answering; and (iii) video captioning. VALUE benchmark aims to cover a broad range of video genres, video lengths, data volumes, and task difficulty levels. Rather than focusing on single-channel videos with visual information only, VALUE promotes models that leverage information from both video frames and their associated subtitles, as well as models that share knowledge across multiple tasks. We evaluate various baseline methods with and without large-scale VidL pre-training, and systematically investigate the impact of video input channels, fusion methods, and different video representations. We also study the transferability between tasks, and conduct multi-task learning under different settings. The significant gap between our best model and human performance calls for future study for advanced VidL models. VALUE is available at https://value-benchmark.github.io/.

Linjie Li, Jie Lei, Zhe Gan et al.

Benefiting from large-scale pre-training, we have witnessed significant performance boost on the popular Visual Question Answering (VQA) task. Despite rapid progress, it remains unclear whether these state-of-the-art (SOTA) models are robust when encountering examples in the wild. To study this, we introduce Adversarial VQA, a new large-scale VQA benchmark, collected iteratively via an adversarial human-and-model-in-the-loop procedure. Through this new benchmark, we discover several interesting findings. (i) Surprisingly, we find that during dataset collection, non-expert annotators can easily attack SOTA VQA models successfully. (ii) Both large-scale pre-trained models and adversarial training methods achieve far worse performance on the new benchmark than over standard VQA v2 dataset, revealing the fragility of these models while demonstrating the effectiveness of our adversarial dataset. (iii) When used for data augmentation, our dataset can effectively boost model performance on other robust VQA benchmarks. We hope our Adversarial VQA dataset can shed new light on robustness study in the community and serve as a valuable benchmark for future work.

Zhe Gan, Yen-Chun Chen, Linjie Li et al.

Large-scale pre-training has recently revolutionized vision-and-language (VL) research. Models such as LXMERT and UNITER have significantly lifted the state of the art over a wide range of VL tasks. However, the large number of parameters in such models hinders their application in practice. In parallel, work on the lottery ticket hypothesis (LTH) has shown that deep neural networks contain small matching subnetworks that can achieve on par or even better performance than the dense networks when trained in isolation. In this work, we perform the first empirical study to assess whether such trainable subnetworks also exist in pre-trained VL models. We use UNITER as the main testbed (also test on LXMERT and ViLT), and consolidate 7 representative VL tasks for experiments, including visual question answering, visual commonsense reasoning, visual entailment, referring expression comprehension, image-text retrieval, GQA, and NLVR$^2$. Through comprehensive analysis, we summarize our main findings as follows. ($i$) It is difficult to find subnetworks that strictly match the performance of the full model. However, we can find "relaxed" winning tickets at 50%-70% sparsity that maintain 99% of the full accuracy. ($ii$) Subnetworks found by task-specific pruning transfer reasonably well to the other tasks, while those found on the pre-training tasks at 60%/70% sparsity transfer universally, matching 98%/96% of the full accuracy on average over all the tasks. ($iii$) Besides UNITER, other models such as LXMERT and ViLT can also play lottery tickets. However, the highest sparsity we can achieve for ViLT is far lower than LXMERT and UNITER (30% vs. 70%). ($iv$) LTH also remains relevant when using other training methods (e.g., adversarial training).

Mingyang Zhou, Luowei Zhou, Shuohang Wang et al.

Vision-and-language pre-training has achieved impressive success in learning multimodal representations between vision and language. To generalize this success to non-English languages, we introduce UC2, the first machine translation-augmented framework for cross-lingual cross-modal representation learning. To tackle the scarcity problem of multilingual captions for image datasets, we first augment existing English-only datasets with other languages via machine translation (MT). Then we extend the standard Masked Language Modeling and Image-Text Matching training objectives to multilingual setting, where alignment between different languages is captured through shared visual context (i.e, using image as pivot). To facilitate the learning of a joint embedding space of images and all languages of interest, we further propose two novel pre-training tasks, namely Masked Region-to-Token Modeling (MRTM) and Visual Translation Language Modeling (VTLM), leveraging MT-enhanced translated data. Evaluation on multilingual image-text retrieval and multilingual visual question answering benchmarks demonstrates that our proposed framework achieves new state-of-the-art on diverse non-English benchmarks while maintaining comparable performance to monolingual pre-trained models on English tasks.

Luowei Zhou, Jingjing Liu, Yu Cheng et al.

This work concerns video-language pre-training and representation learning. In this now ubiquitous training scheme, a model first performs pre-training on paired videos and text (e.g., video clips and accompanied subtitles) from a large uncurated source corpus, before transferring to specific downstream tasks. This two-stage training process inevitably raises questions about the generalization ability of the pre-trained model, which is particularly pronounced when a salient domain gap exists between source and target data (e.g., instructional cooking videos vs. movies). In this paper, we first bring to light the sensitivity of pre-training objectives (contrastive vs. reconstructive) to domain discrepancy. Then, we propose a simple yet effective framework, CUPID, to bridge this domain gap by filtering and adapting source data to the target data, followed by domain-focused pre-training. Comprehensive experiments demonstrate that pre-training on a considerably small subset of domain-focused data can effectively close the source-target domain gap and achieve significant performance gain, compared to random sampling or even exploiting the full pre-training dataset. CUPID yields new state-of-the-art performance across multiple video-language and video tasks, including text-to-video retrieval [72, 37], video question answering [36], and video captioning [72], with consistent performance lift over different pre-training methods.

Liqun Chen, Dong Wang, Zhe Gan et al.

The primary goal of knowledge distillation (KD) is to encapsulate the information of a model learned from a teacher network into a student network, with the latter being more compact than the former. Existing work, e.g., using Kullback-Leibler divergence for distillation, may fail to capture important structural knowledge in the teacher network and often lacks the ability for feature generalization, particularly in situations when teacher and student are built to address different classification tasks. We propose Wasserstein Contrastive Representation Distillation (WCoRD), which leverages both primal and dual forms of Wasserstein distance for KD. The dual form is used for global knowledge transfer, yielding a contrastive learning objective that maximizes the lower bound of mutual information between the teacher and the student networks. The primal form is used for local contrastive knowledge transfer within a mini-batch, effectively matching the distributions of features between the teacher and the student networks. Experiments demonstrate that the proposed WCoRD method outperforms state-of-the-art approaches on privileged information distillation, model compression and cross-modal transfer.

Linjie Li, Zhe Gan, Jingjing Liu

Large-scale pre-trained multimodal transformers, such as ViLBERT and UNITER, have propelled the state of the art in vision-and-language (V+L) research to a new level. Although achieving impressive performance on standard tasks, to date, it still remains unclear how robust these pre-trained models are. To investigate, we conduct a host of thorough evaluations on existing pre-trained models over 4 different types of V+L specific model robustness: (i) Linguistic Variation; (ii) Logical Reasoning; (iii) Visual Content Manipulation; and (iv) Answer Distribution Shift. Interestingly, by standard model finetuning, pre-trained V+L models already exhibit better robustness than many task-specific state-of-the-art methods. To further enhance model robustness, we propose Mango, a generic and efficient approach that learns a Multimodal Adversarial Noise GeneratOr in the embedding space to fool pre-trained V+L models. Differing from previous studies focused on one specific type of robustness, Mango is task-agnostic, and enables universal performance lift for pre-trained models over diverse tasks designed to evaluate broad aspects of robustness. Comprehensive experiments demonstrate that Mango achieves new state of the art on 7 out of 9 robustness benchmarks, surpassing existing methods by a significant margin. As the first comprehensive study on V+L robustness, this work puts robustness of pre-trained models into sharper focus, pointing new directions for future study.

Jinghui Chen, Yu Cheng, Zhe Gan et al.

Adversarial training is so far the most effective strategy in defending against adversarial examples. However, it suffers from high computational costs due to the iterative adversarial attacks in each training step. Recent studies show that it is possible to achieve fast Adversarial Training by performing a single-step attack with random initialization. However, such an approach still lags behind state-of-the-art adversarial training algorithms on both stability and model robustness. In this work, we develop a new understanding towards Fast Adversarial Training, by viewing random initialization as performing randomized smoothing for better optimization of the inner maximization problem. Following this new perspective, we also propose a new initialization strategy, backward smoothing, to further improve the stability and model robustness over single-step robust training methods. Experiments on multiple benchmarks demonstrate that our method achieves similar model robustness as the original TRADES method while using much less training time ($\sim$3x improvement with the same training schedule).

Siqi Sun, Zhe Gan, Yu Cheng et al.

Existing language model compression methods mostly use a simple L2 loss to distill knowledge in the intermediate representations of a large BERT model to a smaller one. Although widely used, this objective by design assumes that all the dimensions of hidden representations are independent, failing to capture important structural knowledge in the intermediate layers of the teacher network. To achieve better distillation efficacy, we propose Contrastive Distillation on Intermediate Representations (CoDIR), a principled knowledge distillation framework where the student is trained to distill knowledge through intermediate layers of the teacher via a contrastive objective. By learning to distinguish positive sample from a large set of negative samples, CoDIR facilitates the student's exploitation of rich information in teacher's hidden layers. CoDIR can be readily applied to compress large-scale language models in both pre-training and finetuning stages, and achieves superb performance on the GLUE benchmark, outperforming state-of-the-art compression methods.

Shuohang Wang, Luowei Zhou, Zhe Gan et al.

Transformer has become ubiquitous in the deep learning field. One of the key ingredients that destined its success is the self-attention mechanism, which allows fully-connected contextual encoding over input tokens. However, despite its effectiveness in modeling short sequences, self-attention suffers when handling inputs with extreme long-range dependencies, as its complexity grows quadratically with respect to the sequence length. Therefore, long sequences are often encoded by Transformer in chunks using a sliding window. In this paper, we propose Cluster-Former, a novel clustering-based sparse Transformer to perform attention across chunked sequences. The proposed framework is pivoted on two unique types of Transformer layer: Sliding-Window Layer and Cluster-Former Layer, which encode local sequence information and global context jointly and iteratively. This new design allows information integration beyond local windows, which is especially beneficial for question answering (QA) tasks that rely on long-range dependencies. Experiments show that Cluster-Former achieves state-of-the-art performance on several major QA benchmarks.

Yuwei Fang, Shuohang Wang, Zhe Gan et al.

Although deep neural networks have achieved tremendous success for question answering (QA), they are still suffering from heavy computational and energy cost for real product deployment. Further, existing QA systems are bottlenecked by the encoding time of real-time questions with neural networks, thus suffering from detectable latency in deployment for large-volume traffic. To reduce the computational cost and accelerate real-time question answering (RTQA) for practical usage, we propose to remove all the neural networks from online QA systems, and present Ocean-Q (an Ocean of Questions), which introduces a new question generation (QG) model to generate a large pool of QA pairs offline, then in real time matches an input question with the candidate QA pool to predict the answer without question encoding. Ocean-Q can be readily deployed in existing distributed database systems or search engine for large-scale query usage, and much greener with no additional cost for maintaining large neural networks. Experiments on SQuAD(-open) and HotpotQA benchmarks demonstrate that Ocean-Q is able to accelerate the fastest state-of-the-art RTQA system by 4X times, with only a 3+% accuracy drop.

Liqun Chen, Zhe Gan, Yu Cheng et al.

Cross-domain alignment between two sets of entities (e.g., objects in an image, words in a sentence) is fundamental to both computer vision and natural language processing. Existing methods mainly focus on designing advanced attention mechanisms to simulate soft alignment, with no training signals to explicitly encourage alignment. The learned attention matrices are also dense and lacks interpretability. We propose Graph Optimal Transport (GOT), a principled framework that germinates from recent advances in Optimal Transport (OT). In GOT, cross-domain alignment is formulated as a graph matching problem, by representing entities into a dynamically-constructed graph. Two types of OT distances are considered: (i) Wasserstein distance (WD) for node (entity) matching; and (ii) Gromov-Wasserstein distance (GWD) for edge (structure) matching. Both WD and GWD can be incorporated into existing neural network models, effectively acting as a drop-in regularizer. The inferred transport plan also yields sparse and self-normalized alignment, enhancing the interpretability of the learned model. Experiments show consistent outperformance of GOT over baselines across a wide range of tasks, including image-text retrieval, visual question answering, image captioning, machine translation, and text summarization.

Zhe Gan, Yen-Chun Chen, Linjie Li et al.

We present VILLA, the first known effort on large-scale adversarial training for vision-and-language (V+L) representation learning. VILLA consists of two training stages: (i) task-agnostic adversarial pre-training; followed by (ii) task-specific adversarial finetuning. Instead of adding adversarial perturbations on image pixels and textual tokens, we propose to perform adversarial training in the embedding space of each modality. To enable large-scale training, we adopt the "free" adversarial training strategy, and combine it with KL-divergence-based regularization to promote higher invariance in the embedding space. We apply VILLA to current best-performing V+L models, and achieve new state of the art on a wide range of tasks, including Visual Question Answering, Visual Commonsense Reasoning, Image-Text Retrieval, Referring Expression Comprehension, Visual Entailment, and NLVR2.

Jize Cao, Zhe Gan, Yu Cheng et al.

Recent Transformer-based large-scale pre-trained models have revolutionized vision-and-language (V+L) research. Models such as ViLBERT, LXMERT and UNITER have significantly lifted state of the art across a wide range of V+L benchmarks with joint image-text pre-training. However, little is known about the inner mechanisms that destine their impressive success. To reveal the secrets behind the scene of these powerful models, we present VALUE (Vision-And-Language Understanding Evaluation), a set of meticulously designed probing tasks (e.g., Visual Coreference Resolution, Visual Relation Detection, Linguistic Probing Tasks) generalizable to standard pre-trained V+L models, aiming to decipher the inner workings of multimodal pre-training (e.g., the implicit knowledge garnered in individual attention heads, the inherent cross-modal alignment learned through contextualized multimodal embeddings). Through extensive analysis of each archetypal model architecture via these probing tasks, our key observations are: (i) Pre-trained models exhibit a propensity for attending over text rather than images during inference. (ii) There exists a subset of attention heads that are tailored for capturing cross-modal interactions. (iii) Learned attention matrix in pre-trained models demonstrates patterns coherent with the latent alignment between image regions and textual words. (iv) Plotted attention patterns reveal visually-interpretable relations among image regions. (v) Pure linguistic knowledge is also effectively encoded in the attention heads. These are valuable insights serving to guide future work towards designing better model architecture and objectives for multimodal pre-training.

Linjie Li, Yen-Chun Chen, Yu Cheng et al.

We present HERO, a novel framework for large-scale video+language omni-representation learning. HERO encodes multimodal inputs in a hierarchical structure, where local context of a video frame is captured by a Cross-modal Transformer via multimodal fusion, and global video context is captured by a Temporal Transformer. In addition to standard Masked Language Modeling (MLM) and Masked Frame Modeling (MFM) objectives, we design two new pre-training tasks: (i) Video-Subtitle Matching (VSM), where the model predicts both global and local temporal alignment; and (ii) Frame Order Modeling (FOM), where the model predicts the right order of shuffled video frames. HERO is jointly trained on HowTo100M and large-scale TV datasets to gain deep understanding of complex social dynamics with multi-character interactions. Comprehensive experiments demonstrate that HERO achieves new state of the art on multiple benchmarks over Text-based Video/Video-moment Retrieval, Video Question Answering (QA), Video-and-language Inference and Video Captioning tasks across different domains. We also introduce two new challenging benchmarks How2QA and How2R for Video QA and Retrieval, collected from diverse video content over multimodalities.

Yandong Li, Yu Cheng, Zhe Gan et al.

We propose a new task towards more practical application for image generation - high-quality image synthesis from salient object layout. This new setting allows users to provide the layout of salient objects only (i.e., foreground bounding boxes and categories), and lets the model complete the drawing with an invented background and a matching foreground. Two main challenges spring from this new task: (i) how to generate fine-grained details and realistic textures without segmentation map input; and (ii) how to create a background and weave it seamlessly into standalone objects. To tackle this, we propose Background Hallucination Generative Adversarial Network (BachGAN), which first selects a set of segmentation maps from a large candidate pool via a background retrieval module, then encodes these candidate layouts via a background fusion module to hallucinate a suitable background for the given objects. By generating the hallucinated background representation dynamically, our model can synthesize high-resolution images with both photo-realistic foreground and integral background. Experiments on Cityscapes and ADE20K datasets demonstrate the advantage of BachGAN over existing methods, measured on both visual fidelity of generated images and visual alignment between output images and input layouts.

Jingzhou Liu, Wenhu Chen, Yu Cheng et al.

We introduce a new task, Video-and-Language Inference, for joint multimodal understanding of video and text. Given a video clip with aligned subtitles as premise, paired with a natural language hypothesis based on the video content, a model needs to infer whether the hypothesis is entailed or contradicted by the given video clip. A new large-scale dataset, named Violin (VIdeO-and-Language INference), is introduced for this task, which consists of 95,322 video-hypothesis pairs from 15,887 video clips, spanning over 582 hours of video. These video clips contain rich content with diverse temporal dynamics, event shifts, and people interactions, collected from two sources: (i) popular TV shows, and (ii) movie clips from YouTube channels. In order to address our new multimodal inference task, a model is required to possess sophisticated reasoning skills, from surface-level grounding (e.g., identifying objects and characters in the video) to in-depth commonsense reasoning (e.g., inferring causal relations of events in the video). We present a detailed analysis of the dataset and an extensive evaluation over many strong baselines, providing valuable insights on the challenges of this new task.

Zongxian Li, Qixiang Ye, Chong Zhang et al.

Unsupervised domain adaptation (UDA) has achieved unprecedented success in improving the cross-domain robustness of object detection models. However, existing UDA methods largely ignore the instantaneous data distribution during model learning, which could deteriorate the feature representation given large domain shift. In this work, we propose a Self-Guided Adaptation (SGA) model, target at aligning feature representation and transferring object detection models across domains while considering the instantaneous alignment difficulty. The core of SGA is to calculate "hardness" factors for sample pairs indicating domain distance in a kernel space. With the hardness factor, the proposed SGA adaptively indicates the importance of samples and assigns them different constrains. Indicated by hardness factors, Self-Guided Progressive Sampling (SPS) is implemented in an "easy-to-hard" way during model adaptation. Using multi-stage convolutional features, SGA is further aggregated to fully align hierarchical representations of detection models. Extensive experiments on commonly used benchmarks show that SGA improves the state-of-the-art methods with significant margins, while demonstrating the effectiveness on large domain shift.

Shuohang Wang, Yunshi Lan, Yi Tay et al.

Transformer has been successfully applied to many natural language processing tasks. However, for textual sequence matching, simple matching between the representation of a pair of sequences might bring in unnecessary noise. In this paper, we propose a new approach to sequence pair matching with Transformer, by learning head-wise matching representations on multiple levels. Experiments show that our proposed approach can achieve new state-of-the-art performance on multiple tasks that rely only on pre-computed sequence-vector-representation, such as SNLI, MNLI-match, MNLI-mismatch, QQP, and SQuAD-binary.

Junjie Hu, Yu Cheng, Zhe Gan et al.

Previous storytelling approaches mostly focused on optimizing traditional metrics such as BLEU, ROUGE and CIDEr. In this paper, we re-examine this problem from a different angle, by looking deep into what defines a realistically-natural and topically-coherent story. To this end, we propose three assessment criteria: relevance, coherence and expressiveness, which we observe through empirical analysis could constitute a "high-quality" story to the human eye. Following this quality guideline, we propose a reinforcement learning framework, ReCo-RL, with reward functions designed to capture the essence of these quality criteria. Experiments on the Visual Storytelling Dataset (VIST) with both automatic and human evaluations demonstrate that our ReCo-RL model achieves better performance than state-of-the-art baselines on both traditional metrics and the proposed new criteria.

Shuyang Dai, Yu Cheng, Yizhe Zhang et al.

Recent unsupervised approaches to domain adaptation primarily focus on minimizing the gap between the source and the target domains through refining the feature generator, in order to learn a better alignment between the two domains. This minimization can be achieved via a domain classifier to detect target-domain features that are divergent from source-domain features. However, by optimizing via such domain classification discrepancy, ambiguous target samples that are not smoothly distributed on the low-dimensional data manifold are often missed. To solve this issue, we propose a novel Contrastively Smoothed Class Alignment (CoSCA) model, that explicitly incorporates both intra- and inter-class domain discrepancy to better align ambiguous target samples with the source domain. CoSCA estimates the underlying label hypothesis of target samples, and simultaneously adapts their feature representations by optimizing a proposed contrastive loss. In addition, Maximum Mean Discrepancy (MMD) is utilized to directly match features between source and target samples for better global alignment. Experiments on several benchmark datasets demonstrate that CoSCA can outperform state-of-the-art approaches for unsupervised domain adaptation by producing more discriminative features.

Linjie Li, Zhe Gan, Yu Cheng et al.

In order to answer semantically-complicated questions about an image, a Visual Question Answering (VQA) model needs to fully understand the visual scene in the image, especially the interactive dynamics between different objects. We propose a Relation-aware Graph Attention Network (ReGAT), which encodes each image into a graph and models multi-type inter-object relations via a graph attention mechanism, to learn question-adaptive relation representations. Two types of visual object relations are explored: (i) Explicit Relations that represent geometric positions and semantic interactions between objects; and (ii) Implicit Relations that capture the hidden dynamics between image regions. Experiments demonstrate that ReGAT outperforms prior state-of-the-art approaches on both VQA 2.0 and VQA-CP v2 datasets. We further show that ReGAT is compatible to existing VQA architectures, and can be used as a generic relation encoder to boost the model performance for VQA.

Zhe Gan, Yu Cheng, Ahmed El Kholy et al.

This paper presents a new model for visual dialog, Recurrent Dual Attention Network (ReDAN), using multi-step reasoning to answer a series of questions about an image. In each question-answering turn of a dialog, ReDAN infers the answer progressively through multiple reasoning steps. In each step of the reasoning process, the semantic representation of the question is updated based on the image and the previous dialog history, and the recurrently-refined representation is used for further reasoning in the subsequent step. On the VisDial v1.0 dataset, the proposed ReDAN model achieves a new state-of-the-art of 64.47% NDCG score. Visualization on the reasoning process further demonstrates that ReDAN can locate context-relevant visual and textual clues via iterative refinement, which can lead to the correct answer step-by-step.

Yu Cheng, Zhe Gan, Yitong Li et al.

Most existing text-to-image synthesis tasks are static single-turn generation, based on pre-defined textual descriptions of images. To explore more practical and interactive real-life applications, we introduce a new task - Interactive Image Editing, where users can guide an agent to edit images via multi-turn textual commands on-the-fly. In each session, the agent takes a natural language description from the user as the input and modifies the image generated in the previous turn to a new design, following the user description. The main challenges in this sequential and interactive image generation task are two-fold: 1) contextual consistency between a generated image and the provided textual description; 2) step-by-step region-level modification to maintain visual consistency across the generated image sequence in each session. To address these challenges, we propose a novel Sequential Attention Generative Adversarial Net-work (SeqAttnGAN), which applies a neural state tracker to encode the previous image and the textual description in each turn of the sequence, and uses a GAN framework to generate a modified version of the image that is consistent with the preceding images and coherent with the description. To achieve better region-specific refinement, we also introduce a sequential attention mechanism into the model. To benchmark on the new task, we introduce two new datasets, Zap-Seq and DeepFashion-Seq, which contain multi-turn sessions with image-description sequences in the fashion domain. Experiments on both datasets show that the proposed SeqAttnGANmodel outperforms state-of-the-art approaches on the interactive image editing task across all evaluation metrics including visual quality, image sequence coherence, and text-image consistency.