Jie Lei

Nicolas Carion, Laura Gustafson, Yuan-Ting Hu et al.

We present Segment Anything Model (SAM) 3, a unified model that detects, segments, and tracks objects in images and videos based on concept prompts, which we define as either short noun phrases (e.g., "yellow school bus"), image exemplars, or a combination of both. Promptable Concept Segmentation (PCS) takes such prompts and returns segmentation masks and unique identities for all matching object instances. To advance PCS, we build a scalable data engine that produces a high-quality dataset with 4M unique concept labels, including hard negatives, across images and videos. Our model consists of an image-level detector and a memory-based video tracker that share a single backbone. Recognition and localization are decoupled with a presence head, which boosts detection accuracy. SAM 3 doubles the accuracy of existing systems in both image and video PCS, and improves previous SAM capabilities on visual segmentation tasks. We open source SAM 3 along with our new Segment Anything with Concepts (SA-Co) benchmark for promptable concept segmentation.

Zineng Tang, Jaemin Cho, Jie Lei et al. · allen-ai

We present Perceiver-VL, a vision-and-language framework that efficiently handles high-dimensional multimodal inputs such as long videos and text. Powered by the iterative latent cross-attention of Perceiver, our framework scales with linear complexity, in contrast to the quadratic complexity of self-attention used in many state-of-the-art transformer-based models. To further improve the efficiency of our framework, we also study applying LayerDrop on cross-attention layers and introduce a mixed-stream architecture for cross-modal retrieval. We evaluate Perceiver-VL on diverse video-text and image-text benchmarks, where Perceiver-VL achieves the lowest GFLOPs and latency while maintaining competitive performance. In addition, we also provide comprehensive analyses of various aspects of our framework, including pretraining data, scalability of latent size and input size, dropping cross-attention layers at inference to reduce latency, modality aggregation strategy, positional encoding, and weight initialization strategy. Our code and checkpoints are available at: https://github.com/zinengtang/Perceiver_VL

Zhenhailong Wang, Manling Li, Ruochen Xu et al.

The goal of this work is to build flexible video-language models that can generalize to various video-to-text tasks from few examples, such as domain-specific captioning, question answering, and future event prediction. Existing few-shot video-language learners focus exclusively on the encoder, resulting in the absence of a video-to-text decoder to handle generative tasks. Video captioners have been pretrained on large-scale video-language datasets, but they rely heavily on finetuning and lack the ability to generate text for unseen tasks in a few-shot setting. We propose VidIL, a few-shot Video-language Learner via Image and Language models, which demonstrates strong performance on few-shot video-to-text tasks without the necessity of pretraining or finetuning on any video datasets. We use the image-language models to translate the video content into frame captions, object, attribute, and event phrases, and compose them into a temporal structure template. We then instruct a language model, with a prompt containing a few in-context examples, to generate a target output from the composed content. The flexibility of prompting allows the model to capture any form of text input, such as automatic speech recognition (ASR) transcripts. Our experiments demonstrate the power of language models in understanding videos on a wide variety of video-language tasks, including video captioning, video question answering, video caption retrieval, and video future event prediction. Especially, on video future event prediction, our few-shot model significantly outperforms state-of-the-art supervised models trained on large-scale video datasets. Code and resources are publicly available for research purposes at https://github.com/MikeWangWZHL/VidIL .

Jie Lei, Tamara L. Berg, Mohit Bansal

Training an effective video-and-language model intuitively requires multiple frames as model inputs. However, it is unclear whether using multiple frames is beneficial to downstream tasks, and if yes, whether the performance gain is worth the drastically-increased computation and memory costs resulting from using more frames. In this work, we explore single-frame models for video-and-language learning. On a diverse set of video-and-language tasks (including text-to-video retrieval and video question answering), we show the surprising result that, with large-scale pre-training and a proper frame ensemble strategy at inference time, a single-frame trained model that does not consider temporal information can achieve better performance than existing methods that use multiple frames for training. This result reveals the existence of a strong "static appearance bias" in popular video-and-language datasets. Therefore, to allow for a more comprehensive evaluation of video-and-language models, we propose two new retrieval tasks based on existing fine-grained action recognition datasets that encourage temporal modeling. Our code is available at https://github.com/jayleicn/singularity

Yan-Bo Lin, Yi-Lin Sung, Jie Lei et al.

Vision transformers (ViTs) have achieved impressive results on various computer vision tasks in the last several years. In this work, we study the capability of frozen ViTs, pretrained only on visual data, to generalize to audio-visual data without finetuning any of its original parameters. To do so, we propose a latent audio-visual hybrid (LAVISH) adapter that adapts pretrained ViTs to audio-visual tasks by injecting a small number of trainable parameters into every layer of a frozen ViT. To efficiently fuse visual and audio cues, our LAVISH adapter uses a small set of latent tokens, which form an attention bottleneck, thus, eliminating the quadratic cost of standard cross-attention. Compared to the existing modality-specific audio-visual methods, our approach achieves competitive or even better performance on various audio-visual tasks while using fewer tunable parameters and without relying on costly audio pretraining or external audio encoders. Our code is available at https://genjib.github.io/project_page/LAVISH/

Feng Cheng, Xizi Wang, Jie Lei et al.

The last several years have witnessed remarkable progress in video-and-language (VidL) understanding. However, most modern VidL approaches use complex and specialized model architectures and sophisticated pretraining protocols, making the reproducibility, analysis and comparisons of these frameworks difficult. Hence, instead of proposing yet another new VidL model, this paper conducts a thorough empirical study demystifying the most important factors in the VidL model design. Among the factors that we investigate are (i) the spatiotemporal architecture design, (ii) the multimodal fusion schemes, (iii) the pretraining objectives, (iv) the choice of pretraining data, (v) pretraining and finetuning protocols, and (vi) dataset and model scaling. Our empirical study reveals that the most important design factors include: temporal modeling, video-to-text multimodal fusion, masked modeling objectives, and joint training on images and videos. Using these empirical insights, we then develop a step-by-step recipe, dubbed VindLU, for effective VidL pretraining. Our final model trained using our recipe achieves comparable or better than state-of-the-art results on several VidL tasks without relying on external CLIP pretraining. In particular, on the text-to-video retrieval task, our approach obtains 61.2% on DiDeMo, and 55.0% on ActivityNet, outperforming current SOTA by 7.8% and 6.1% respectively. Furthermore, our model also obtains state-of-the-art video question-answering results on ActivityNet-QA, MSRVTT-QA, MSRVTT-MC and TVQA. Our code and pretrained models are publicly available at: https://github.com/klauscc/VindLU.

Jiaqing Zhang, Jie Lei, Weiying Xie et al.

Accurately and timely detecting multiscale small objects that contain tens of pixels from remote sensing images (RSI) remains challenging. Most of the existing solutions primarily design complex deep neural networks to learn strong feature representations for objects separated from the background, which often results in a heavy computation burden. In this article, we propose an accurate yet fast object detection method for RSI, named SuperYOLO, which fuses multimodal data and performs high-resolution (HR) object detection on multiscale objects by utilizing the assisted super resolution (SR) learning and considering both the detection accuracy and computation cost. First, we utilize a symmetric compact multimodal fusion (MF) to extract supplementary information from various data for improving small object detection in RSI. Furthermore, we design a simple and flexible SR branch to learn HR feature representations that can discriminate small objects from vast backgrounds with low-resolution (LR) input, thus further improving the detection accuracy. Moreover, to avoid introducing additional computation, the SR branch is discarded in the inference stage, and the computation of the network model is reduced due to the LR input. Experimental results show that, on the widely used VEDAI RS dataset, SuperYOLO achieves an accuracy of 75.09% (in terms of mAP50 ), which is more than 10% higher than the SOTA large models, such as YOLOv5l, YOLOv5x, and RS designed YOLOrs. Meanwhile, the parameter size and GFLOPs of SuperYOLO are about 18 times and 3.8 times less than YOLOv5x. Our proposed model shows a favorable accuracy and speed tradeoff compared to the state-of-the-art models. The code will be open-sourced at https://github.com/icey-zhang/SuperYOLO.

Jiaqing Zhang, Jie Lei, Weiying Xie et al.

Considering the computation complexity, we propose a Guided Hybrid Quantization with One-to-one Self-Teaching (GHOST}) framework. More concretely, we first design a structure called guided quantization self-distillation (GQSD), which is an innovative idea for realizing lightweight through the synergy of quantization and distillation. The training process of the quantization model is guided by its full-precision model, which is time-saving and cost-saving without preparing a huge pre-trained model in advance. Second, we put forward a hybrid quantization (HQ) module to obtain the optimal bit width automatically under a constrained condition where a threshold for distribution distance between the center and samples is applied in the weight value search space. Third, in order to improve information transformation, we propose a one-to-one self-teaching (OST) module to give the student network a ability of self-judgment. A switch control machine (SCM) builds a bridge between the student network and teacher network in the same location to help the teacher to reduce wrong guidance and impart vital knowledge to the student. This distillation method allows a model to learn from itself and gain substantial improvement without any additional supervision. Extensive experiments on a multimodal dataset (VEDAI) and single-modality datasets (DOTA, NWPU, and DIOR) show that object detection based on GHOST outperforms the existing detectors. The tiny parameters (<9.7 MB) and Bit-Operations (BOPs) (<2158 G) compared with any remote sensing-based, lightweight or distillation-based algorithms demonstrate the superiority in the lightweight design domain. Our code and model will be released at https://github.com/icey-zhang/GHOST.

Jie Lei, Xinlei Chen, Ning Zhang et al. · meta-ai

Dual encoders and cross encoders have been widely used for image-text retrieval. Between the two, the dual encoder encodes the image and text independently followed by a dot product, while the cross encoder jointly feeds image and text as the input and performs dense multi-modal fusion. These two architectures are typically modeled separately without interaction. In this work, we propose LoopITR, which combines them in the same network for joint learning. Specifically, we let the dual encoder provide hard negatives to the cross encoder, and use the more discriminative cross encoder to distill its predictions back to the dual encoder. Both steps are efficiently performed together in the same model. Our work centers on empirical analyses of this combined architecture, putting the main focus on the design of the distillation objective. Our experimental results highlight the benefits of training the two encoders in the same network, and demonstrate that distillation can be quite effective with just a few hard negative examples. Experiments on two standard datasets (Flickr30K and COCO) show our approach achieves state-of-the-art dual encoder performance when compared with approaches using a similar amount of data.

Guoyu Yang, Jie Lei, Zhikuan Zhu et al.

Multi-scale features are of great importance in encoding objects with scale variance in object detection tasks. A common strategy for multi-scale feature extraction is adopting the classic top-down and bottom-up feature pyramid networks. However, these approaches suffer from the loss or degradation of feature information, impairing the fusion effect of non-adjacent levels. This paper proposes an asymptotic feature pyramid network (AFPN) to support direct interaction at non-adjacent levels. AFPN is initiated by fusing two adjacent low-level features and asymptotically incorporates higher-level features into the fusion process. In this way, the larger semantic gap between non-adjacent levels can be avoided. Given the potential for multi-object information conflicts to arise during feature fusion at each spatial location, adaptive spatial fusion operation is further utilized to mitigate these inconsistencies. We incorporate the proposed AFPN into both two-stage and one-stage object detection frameworks and evaluate with the MS-COCO 2017 validation and test datasets. Experimental evaluation shows that our method achieves more competitive results than other state-of-the-art feature pyramid networks. The code is available at \href{https://github.com/gyyang23/AFPN}{https://github.com/gyyang23/AFPN}.

Yan-Bo Lin, Jie Lei, Mohit Bansal et al.

We introduce an audiovisual method for long-range text-to-video retrieval. Unlike previous approaches designed for short video retrieval (e.g., 5-15 seconds in duration), our approach aims to retrieve minute-long videos that capture complex human actions. One challenge of standard video-only approaches is the large computational cost associated with processing hundreds of densely extracted frames from such long videos. To address this issue, we propose to replace parts of the video with compact audio cues that succinctly summarize dynamic audio events and are cheap to process. Our method, named ECLIPSE (Efficient CLIP with Sound Encoding), adapts the popular CLIP model to an audiovisual video setting, by adding a unified audiovisual transformer block that captures complementary cues from the video and audio streams. In addition to being 2.92x faster and 2.34x memory-efficient than long-range video-only approaches, our method also achieves better text-to-video retrieval accuracy on several diverse long-range video datasets such as ActivityNet, QVHighlights, YouCook2, DiDeMo and Charades.

Weiying Xie, Yusi Zhang, Tianlin Hui et al.

Multimodal object detection offers a promising prospect to facilitate robust detection in various visual conditions. However, existing two-stream backbone networks are challenged by complex fusion and substantial parameter increments. This is primarily due to large data distribution biases of multimodal homogeneous information. In this paper, we propose a novel multimodal object detector, named Low-rank Modal Adaptors (LMA) with a shared backbone. The shared parameters enhance the consistency of homogeneous information, while lightweight modal adaptors focus on modality unique features. Furthermore, we design an adaptive rank allocation strategy to adapt to the varying heterogeneity at different feature levels. When applied to two multimodal object detection datasets, experiments validate the effectiveness of our method. Notably, on DroneVehicle, LMA attains a 10.4% accuracy improvement over the state-of-the-art method with a 149M-parameters reduction. The code is available at https://github.com/zyszxhy/FoRA. Our work was submitted to ACM MM in April 2024, but was rejected. We will continue to refine our work and paper writing next, mainly including proof of theory and multi-task applications of FoRA.

Jie Lei, José Flich, Enrique S. Quintana-Ortí

The remarkable positive impact of Deep Neural Networks on many Artificial Intelligence (AI) tasks has led to the development of various high performance algorithms as well as specialized processors and accelerators. In this paper we address this scenario by demonstrating that the principles underlying the modern realization of the general matrix multiplication (GEMM) in conventional processor architectures, are also valid to achieve high performance for the type of operations that arise in deep learning (DL) on an exotic accelerator such as the AI Engine (AIE) tile embedded in Xilinx Versal platforms. In particular, our experimental results with a prototype implementation of the GEMM kernel, on a Xilinx Versal VCK190, delivers performance close to 86.7% of the theoretical peak that can be expected on an AIE tile, for 16-bit integer operands.

Jie Lei, Zhao Liu, Zeyu Zou et al.

Facial expression is an essential factor in conveying human emotional states and intentions. Although remarkable advancement has been made in facial expression recognition (FER) task, challenges due to large variations of expression patterns and unavoidable data uncertainties still remain. In this paper, we propose mid-level representation enhancement (MRE) and graph embedded uncertainty suppressing (GUS) addressing these issues. On one hand, MRE is introduced to avoid expression representation learning being dominated by a limited number of highly discriminative patterns. On the other hand, GUS is introduced to suppress the feature ambiguity in the representation space. The proposed method not only has stronger generalization capability to handle different variations of expression patterns but also more robustness to capture expression representations. Experimental evaluation on Aff-Wild2 have verified the effectiveness of the proposed method.

Lin Chen, Zhijie Jia, Tian Qiu et al.

A surge of interest has emerged in utilizing Transformers in diverse vision tasks owing to its formidable performance. However, existing approaches primarily focus on optimizing internal model architecture designs that often entail significant trial and error with high burdens. In this work, we propose a new paradigm dubbed Decision Stream Calibration that boosts the performance of general Vision Transformers. To achieve this, we shed light on the information propagation mechanism in the learning procedure by exploring the correlation between different tokens and the relevance coefficient of multiple dimensions. Upon further analysis, it was discovered that 1) the final decision is associated with tokens of foreground targets, while token features of foreground target will be transmitted into the next layer as much as possible, and the useless token features of background area will be eliminated gradually in the forward propagation. 2) Each category is solely associated with specific sparse dimensions in the tokens. Based on the discoveries mentioned above, we designed a two-stage calibration scheme, namely ViT-Calibrator, including token propagation calibration stage and dimension propagation calibration stage. Extensive experiments on commonly used datasets show that the proposed approach can achieve promising results. The source codes are given in the supplements.

Jiaqing Zhang, Jie Lei, Weiying Xie et al.

In multimodal land cover classification (MLCC), a common challenge is the redundancy in data distribution, where irrelevant information from multiple modalities can hinder the effective integration of their unique features. To tackle this, we introduce the Multimodal Informative Vit (MIVit), a system with an innovative information aggregate-distributing mechanism. This approach redefines redundancy levels and integrates performance-aware elements into the fused representation, facilitating the learning of semantics in both forward and backward directions. MIVit stands out by significantly reducing redundancy in the empirical distribution of each modality's separate and fused features. It employs oriented attention fusion (OAF) for extracting shallow local features across modalities in horizontal and vertical dimensions, and a Transformer feature extractor for extracting deep global features through long-range attention. We also propose an information aggregation constraint (IAC) based on mutual information, designed to remove redundant information and preserve complementary information within embedded features. Additionally, the information distribution flow (IDF) in MIVit enhances performance-awareness by distributing global classification information across different modalities' feature maps. This architecture also addresses missing modality challenges with lightweight independent modality classifiers, reducing the computational load typically associated with Transformers. Our results show that MIVit's bidirectional aggregate-distributing mechanism between modalities is highly effective, achieving an average overall accuracy of 95.56% across three multimodal datasets. This performance surpasses current state-of-the-art methods in MLCC. The code for MIVit is accessible at https://github.com/icey-zhang/MIViT.

Jiazhen Yang, Junjun Zheng, Kejia Chen et al.

With the rapid development of generative models and multimodal content editing technologies, the key challenge faced by synthetic image detection (SID) lies in cross-distribution generalization to unknown generation sources. In recent years, visual foundation models (VFM), which acquire rich visual priors through large scale image-text alignment pretraining, have become a promising technical route for improving the generalization ability of SID. However, existing VFM-based methods remain relatively coarse-grained in their adaptation strategies. They typically either directly use the final layer representations of VFM or simply fuse multi layer features, lacking explicit modeling of the optimal representational hierarchy for transferable forgery cues. Meanwhile, although directly fine-tuning VFM can enhance task adaptation, it may also damage the cross-modal pretrained structure that supports open-set generalization. To address this task specific tension, we reformulate VFM adaptation for SID as a joint optimization problem: it is necessary both to identify the critical representational layer that is more suitable for carrying forgery discriminative information and to constrain the disturbance caused by task knowledge injection to the pretrained structure. Based on this, we propose I2P, an SID framework centered on intrinsic importance perception. I2P first adaptively identifies the critical layer representations that are most discriminative for SID, and then constrains task-driven parameter updates within a low sensitivity parameter subspace, thereby improving task specificity while preserving the transferable structure of pretrained representations as much as possible.

Jiaqing Zhang, Jie Lei, Weiying Xie et al.

Accurate cloud recognition and warning are crucial for various applications, including in-flight support, weather forecasting, and climate research. However, recent deep learning algorithms have predominantly focused on detecting cloud regions in satellite imagery, with insufficient attention to the specificity required for accurate cloud recognition. This limitation inspired us to develop the novel FY-4A-Himawari-8 (FYH) dataset, which includes nine distinct cloud categories and uses precise domain adaptation methods to align 70,419 image-label pairs in terms of projection, temporal resolution, and spatial resolution, thereby facilitating the training of supervised deep learning networks. Given the complexity and diversity of cloud formations, we have thoroughly analyzed the challenges inherent to cloud recognition tasks, examining the intricate characteristics and distribution of the data. To effectively address these challenges, we designed a Distribution-aware Interactive-Attention Network (DIAnet), which preserves pixel-level details through a high-resolution branch and a parallel multi-resolution cross-branch. We also integrated a distribution-aware loss (DAL) to mitigate the imbalance across cloud categories. An Interactive Attention Module (IAM) further enhances the robustness of feature extraction combined with spatial and channel information. Empirical evaluations on the FYH dataset demonstrate that our method outperforms other cloud recognition networks, achieving superior performance in terms of mean Intersection over Union (mIoU). The code for implementing DIAnet is available at https://github.com/icey-zhang/DIAnet.

Mingxiang Cao, Weiying Xie, Xin Zhang et al.

Multi-modal fusion holds great promise for integrating information from different modalities. However, due to a lack of consideration for modal consistency, existing multi-modal fusion methods in the field of remote sensing still face challenges of incomplete semantic information and low computational efficiency in their fusion designs. Inspired by the observation that the visual language pre-training model CLIP can effectively extract strong semantic information from visual features, we propose M$^3$amba, a novel end-to-end CLIP-driven Mamba model for multi-modal fusion to address these challenges. Specifically, we introduce CLIP-driven modality-specific adapters in the fusion architecture to avoid the bias of understanding specific domains caused by direct inference, making the original CLIP encoder modality-specific perception. This unified framework enables minimal training to achieve a comprehensive semantic understanding of different modalities, thereby guiding cross-modal feature fusion. To further enhance the consistent association between modality mappings, a multi-modal Mamba fusion architecture with linear complexity and a cross-attention module Cross-SS2D are designed, which fully considers effective and efficient information interaction to achieve complete fusion. Extensive experiments have shown that M$^3$amba has an average performance improvement of at least 5.98\% compared with the state-of-the-art methods in multi-modal hyperspectral image classification tasks in the remote sensing field, while also demonstrating excellent training efficiency, achieving a double improvement in accuracy and efficiency. The code is released at https://github.com/kaka-Cao/M3amba.

Zichun Yu, Fei Peng, Jie Lei et al.

In this paper, we introduce Group-MATES, an efficient group-level data selection approach to optimize the speed-quality frontier of language model pretraining. Specifically, Group-MATES parameterizes costly group-level selection with a relational data influence model. To train this model, we sample training trajectories of the language model and collect oracle data influences alongside. The relational data influence model approximates the oracle data influence by weighting individual influence with relationships among training data. To enable efficient selection with our relational data influence model, we partition the dataset into small clusters using relationship weights and select data within each cluster independently. Experiments on DCLM 400M-4x, 1B-1x, and 3B-1x show that Group-MATES achieves 3.5%-9.4% relative performance gains over random selection across 22 downstream tasks, nearly doubling the improvements achieved by state-of-the-art individual data selection baselines. Furthermore, Group-MATES reduces the number of tokens required to reach a certain downstream performance by up to 1.75x, substantially elevating the speed-quality frontier. Further analyses highlight the critical role of relationship weights in the relational data influence model and the effectiveness of our cluster-based inference. Our code is open-sourced at https://github.com/facebookresearch/Group-MATES.

Jiaqing Zhang, Mingxiang Cao, Weiying Xie et al.

Multimodal image fusion and object detection are crucial for autonomous driving. While current methods have advanced the fusion of texture details and semantic information, their complex training processes hinder broader applications. Addressing this challenge, we introduce E2E-MFD, a novel end-to-end algorithm for multimodal fusion detection. E2E-MFD streamlines the process, achieving high performance with a single training phase. It employs synchronous joint optimization across components to avoid suboptimal solutions tied to individual tasks. Furthermore, it implements a comprehensive optimization strategy in the gradient matrix for shared parameters, ensuring convergence to an optimal fusion detection configuration. Our extensive testing on multiple public datasets reveals E2E-MFD's superior capabilities, showcasing not only visually appealing image fusion but also impressive detection outcomes, such as a 3.9% and 2.0% mAP50 increase on horizontal object detection dataset M3FD and oriented object detection dataset DroneVehicle, respectively, compared to state-of-the-art approaches. The code is released at https://github.com/icey-zhang/E2E-MFD.

Li Xu, Gang He, Jinjia Zhou et al.

In most video platforms, such as Youtube, and TikTok, the played videos usually have undergone multiple video encodings such as hardware encoding by recording devices, software encoding by video editing apps, and single/multiple video transcoding by video application servers. Previous works in compressed video restoration typically assume the compression artifacts are caused by one-time encoding. Thus, the derived solution usually does not work very well in practice. In this paper, we propose a new method, temporal spatial auxiliary network (TSAN), for transcoded video restoration. Our method considers the unique traits between video encoding and transcoding, and we consider the initial shallow encoded videos as the intermediate labels to assist the network to conduct self-supervised attention training. In addition, we employ adjacent multi-frame information and propose the temporal deformable alignment and pyramidal spatial fusion for transcoded video restoration. The experimental results demonstrate that the performance of the proposed method is superior to that of the previous techniques. The code is available at https://github.com/icecherylXuli/TSAN.

Lechao Cheng, Zunlei Feng, Xinchao Wang et al.

Given a reference object of an unknown type in an image, human observers can effortlessly find the objects of the same category in another image and precisely tell their visual boundaries. Such visual cognition capability of humans seems absent from the current research spectrum of computer vision. Existing segmentation networks, for example, rely on a humongous amount of labeled data, which is laborious and costly to collect and annotate; besides, the performance of segmentation networks tend to downgrade as the number of the category increases. In this paper, we introduce a novel Reference semantic segmentation Network (Ref-Net) to conduct visual boundary knowledge translation. Ref-Net contains a Reference Segmentation Module (RSM) and a Boundary Knowledge Translation Module (BKTM). Inspired by the human recognition mechanism, RSM is devised only to segment the same category objects based on the features of the reference objects. BKTM, on the other hand, introduces two boundary discriminator branches to conduct inner and outer boundary segmentation of the target objectin an adversarial manner, and translate the annotated boundary knowledge of open-source datasets into the segmentation network. Exhaustive experiments demonstrate that, with tens of finely-grained annotated samples as guidance, Ref-Net achieves results on par with fully supervised methods on six datasets.

Zunlei Feng, Zhonghua Wang, Xinchao Wang et al.

The microvascular invasion (MVI) is a major prognostic factor in hepatocellular carcinoma, which is one of the malignant tumors with the highest mortality rate. The diagnosis of MVI needs discovering the vessels that contain hepatocellular carcinoma cells and counting their number in each vessel, which depends heavily on experiences of the doctor, is largely subjective and time-consuming. However, there is no algorithm as yet tailored for the MVI detection from pathological images. This paper collects the first pathological liver image dataset containing 522 whole slide images with labels of vessels, MVI, and hepatocellular carcinoma grades. The first and essential step for the automatic diagnosis of MVI is the accurate segmentation of vessels. The unique characteristics of pathological liver images, such as super-large size, multi-scale vessel, and blurred vessel edges, make the accurate vessel segmentation challenging. Based on the collected dataset, we propose an Edge-competing Vessel Segmentation Network (EVS-Net), which contains a segmentation network and two edge segmentation discriminators. The segmentation network, combined with an edge-aware self-supervision mechanism, is devised to conduct vessel segmentation with limited labeled patches. Meanwhile, two discriminators are introduced to distinguish whether the segmented vessel and background contain residual features in an adversarial manner. In the training stage, two discriminators are devised tocompete for the predicted position of edges. Exhaustive experiments demonstrate that, with only limited labeled patches, EVS-Net achieves a close performance of fully supervised methods, which provides a convenient tool for the pathological liver vessel segmentation. Code is publicly available at https://github.com/zju-vipa/EVS-Net.

Jie Lei, Tamara L. Berg, Mohit Bansal

We introduce mTVR, a large-scale multilingual video moment retrieval dataset, containing 218K English and Chinese queries from 21.8K TV show video clips. The dataset is collected by extending the popular TVR dataset (in English) with paired Chinese queries and subtitles. Compared to existing moment retrieval datasets, mTVR is multilingual, larger, and comes with diverse annotations. We further propose mXML, a multilingual moment retrieval model that learns and operates on data from both languages, via encoder parameter sharing and language neighborhood constraints. We demonstrate the effectiveness of mXML on the newly collected MTVR dataset, where mXML outperforms strong monolingual baselines while using fewer parameters. In addition, we also provide detailed dataset analyses and model ablations. Data and code are publicly available at https://github.com/jayleicn/mTVRetrieval

Jie Lei, Tamara L. Berg, Mohit Bansal

Detecting customized moments and highlights from videos given natural language (NL) user queries is an important but under-studied topic. One of the challenges in pursuing this direction is the lack of annotated data. To address this issue, we present the Query-based Video Highlights (QVHIGHLIGHTS) dataset. It consists of over 10,000 YouTube videos, covering a wide range of topics, from everyday activities and travel in lifestyle vlog videos to social and political activities in news videos. Each video in the dataset is annotated with: (1) a human-written free-form NL query, (2) relevant moments in the video w.r.t. the query, and (3) five-point scale saliency scores for all query-relevant clips. This comprehensive annotation enables us to develop and evaluate systems that detect relevant moments as well as salient highlights for diverse, flexible user queries. We also present a strong baseline for this task, Moment-DETR, a transformer encoder-decoder model that views moment retrieval as a direct set prediction problem, taking extracted video and query representations as inputs and predicting moment coordinates and saliency scores end-to-end. While our model does not utilize any human prior, we show that it performs competitively when compared to well-engineered architectures. With weakly supervised pretraining using ASR captions, MomentDETR substantially outperforms previous methods. Lastly, we present several ablations and visualizations of Moment-DETR. Data and code is publicly available at https://github.com/jayleicn/moment_detr

Hao Tan, Jie Lei, Thomas Wolf et al.

Video understanding relies on perceiving the global content and modeling its internal connections (e.g., causality, movement, and spatio-temporal correspondence). To learn these interactions, we apply a mask-then-predict pre-training task on discretized video tokens generated via VQ-VAE. Unlike language, where the text tokens are more independent, neighboring video tokens typically have strong correlations (e.g., consecutive video frames usually look very similar), and hence uniformly masking individual tokens will make the task too trivial to learn useful representations. To deal with this issue, we propose a block-wise masking strategy where we mask neighboring video tokens in both spatial and temporal domains. We also add an augmentation-free contrastive learning method to further capture the global content by predicting whether the video clips are sampled from the same video. We pre-train our model on uncurated videos and show that our pre-trained model can reach state-of-the-art results on several video understanding datasets (e.g., SSV2, Diving48). Lastly, we provide detailed analyses on model scalability and pre-training method design. Code is released at https://github.com/airsplay/vimpac.

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

Jaemin Cho, Jie Lei, Hao Tan et al.

Existing methods for vision-and-language learning typically require designing task-specific architectures and objectives for each task. For example, a multi-label answer classifier for visual question answering, a region scorer for referring expression comprehension, and a language decoder for image captioning, etc. To alleviate these hassles, in this work, we propose a unified framework that learns different tasks in a single architecture with the same language modeling objective, i.e., multimodal conditional text generation, where our models learn to generate labels in text based on the visual and textual inputs. On 7 popular vision-and-language benchmarks, including visual question answering, referring expression comprehension, visual commonsense reasoning, most of which have been previously modeled as discriminative tasks, our generative approach (with a single unified architecture) reaches comparable performance to recent task-specific state-of-the-art vision-and-language models. Moreover, our generative approach shows better generalization ability on questions that have rare answers. Also, we show that our framework allows multi-task learning in a single architecture with a single set of parameters, achieving similar performance to separately optimized single-task models. Our code is publicly available at: https://github.com/j-min/VL-T5

Jie Lei, Licheng Yu, Tamara L. Berg et al.

Given a video with aligned dialogue, people can often infer what is more likely to happen next. Making such predictions requires not only a deep understanding of the rich dynamics underlying the video and dialogue, but also a significant amount of commonsense knowledge. In this work, we explore whether AI models are able to learn to make such multimodal commonsense next-event predictions. To support research in this direction, we collect a new dataset, named Video-and-Language Event Prediction (VLEP), with 28,726 future event prediction examples (along with their rationales) from 10,234 diverse TV Show and YouTube Lifestyle Vlog video clips. In order to promote the collection of non-trivial challenging examples, we employ an adversarial human-and-model-in-the-loop data collection procedure. We also present a strong baseline incorporating information from video, dialogue, and commonsense knowledge. Experiments show that each type of information is useful for this challenging task, and that compared to the high human performance on VLEP, our model provides a good starting point but leaves large room for future work. Our dataset and code are available at: https://github.com/jayleicn/VideoLanguageFuturePred

Jie Lei, Liwei Wang, Yelong Shen et al.

Generating multi-sentence descriptions for videos is one of the most challenging captioning tasks due to its high requirements for not only visual relevance but also discourse-based coherence across the sentences in the paragraph. Towards this goal, we propose a new approach called Memory-Augmented Recurrent Transformer (MART), which uses a memory module to augment the transformer architecture. The memory module generates a highly summarized memory state from the video segments and the sentence history so as to help better prediction of the next sentence (w.r.t. coreference and repetition aspects), thus encouraging coherent paragraph generation. Extensive experiments, human evaluations, and qualitative analyses on two popular datasets ActivityNet Captions and YouCookII show that MART generates more coherent and less repetitive paragraph captions than baseline methods, while maintaining relevance to the input video events. All code is available open-source at: https://github.com/jayleicn/recurrent-transformer

Jie Lei, Licheng Yu, Tamara L. Berg et al.

We present the task of Spatio-Temporal Video Question Answering, which requires intelligent systems to simultaneously retrieve relevant moments and detect referenced visual concepts (people and objects) to answer natural language questions about videos. We first augment the TVQA dataset with 310.8K bounding boxes, linking depicted objects to visual concepts in questions and answers. We name this augmented version as TVQA+. We then propose Spatio-Temporal Answerer with Grounded Evidence (STAGE), a unified framework that grounds evidence in both spatial and temporal domains to answer questions about videos. Comprehensive experiments and analyses demonstrate the effectiveness of our framework and how the rich annotations in our TVQA+ dataset can contribute to the question answering task. Moreover, by performing this joint task, our model is able to produce insightful and interpretable spatio-temporal attention visualizations. Dataset and code are publicly available at: http: //tvqa.cs.unc.edu, https://github.com/jayleicn/TVQAplus

Jiayuan Tian, Jie Lei, Jiaqing Zhang et al.

With recent advancements in aerospace technology, the volume of unlabeled remote sensing image (RSI) data has increased dramatically. Effectively leveraging this data through self-supervised learning (SSL) is vital in the field of remote sensing. However, current methodologies, particularly contrastive learning (CL), a leading SSL method, encounter specific challenges in this domain. Firstly, CL often mistakenly identifies geographically adjacent samples with similar semantic content as negative pairs, leading to confusion during model training. Secondly, as an instance-level discriminative task, it tends to neglect the essential fine-grained features and complex details inherent in unstructured RSIs. To overcome these obstacles, we introduce SwiMDiff, a novel self-supervised pre-training framework designed for RSIs. SwiMDiff employs a scene-wide matching approach that effectively recalibrates labels to recognize data from the same scene as false negatives. This adjustment makes CL more applicable to the nuances of remote sensing. Additionally, SwiMDiff seamlessly integrates CL with a diffusion model. Through the implementation of pixel-level diffusion constraints, we enhance the encoder's ability to capture both the global semantic information and the fine-grained features of the images more comprehensively. Our proposed framework significantly enriches the information available for downstream tasks in remote sensing. Demonstrating exceptional performance in change detection and land-cover classification tasks, SwiMDiff proves its substantial utility and value in the field of remote sensing.

Carolina Zheng, Minhui Huang, Dmitrii Pedchenko et al.

The exponential growth of online content has posed significant challenges to ID-based models in industrial recommendation systems, ranging from extremely high cardinality and dynamically growing ID space, to highly skewed engagement distributions, to prediction instability as a result of natural id life cycles (e.g, the birth of new IDs and retirement of old IDs). To address these issues, many systems rely on random hashing to handle the id space and control the corresponding model parameters (i.e embedding table). However, this approach introduces data pollution from multiple ids sharing the same embedding, leading to degraded model performance and embedding representation instability. This paper examines these challenges and introduces Semantic ID prefix ngram, a novel token parameterization technique that significantly improves the performance of the original Semantic ID. Semantic ID prefix ngram creates semantically meaningful collisions by hierarchically clustering items based on their content embeddings, as opposed to random assignments. Through extensive experimentation, we demonstrate that Semantic ID prefix ngram not only addresses embedding instability but also significantly improves tail id modeling, reduces overfitting, and mitigates representation shifts. We further highlight the advantages of Semantic ID prefix ngram in attention-based models that contextualize user histories, showing substantial performance improvements. We also report our experience of integrating Semantic ID into Meta production Ads Ranking system, leading to notable performance gains and enhanced prediction stability in live deployments.

Chiyu Zhang, Yifei Sun, Minghao Wu et al.

Content-based recommendation systems play a crucial role in delivering personalized content to users in the digital world. In this work, we introduce EmbSum, a novel framework that enables offline pre-computations of users and candidate items while capturing the interactions within the user engagement history. By utilizing the pretrained encoder-decoder model and poly-attention layers, EmbSum derives User Poly-Embedding (UPE) and Content Poly-Embedding (CPE) to calculate relevance scores between users and candidate items. EmbSum actively learns the long user engagement histories by generating user-interest summary with supervision from large language model (LLM). The effectiveness of EmbSum is validated on two datasets from different domains, surpassing state-of-the-art (SoTA) methods with higher accuracy and fewer parameters. Additionally, the model's ability to generate summaries of user interests serves as a valuable by-product, enhancing its usefulness for personalized content recommendations.

Chiyu Zhang, Yifei Sun, Jun Chen et al.

Leveraging users' long engagement histories is essential for personalized content recommendations. The success of pretrained language models (PLMs) in NLP has led to their use in encoding user histories and candidate items, framing content recommendations as textual semantic matching tasks. However, existing works still struggle with processing very long user historical text and insufficient user-item interaction. In this paper, we introduce a content-based recommendation framework, SPAR, which effectively tackles the challenges of holistic user interest extraction from the long user engagement history. It achieves so by leveraging PLM, poly-attention layers and attention sparsity mechanisms to encode user's history in a session-based manner. The user and item side features are sufficiently fused for engagement prediction while maintaining standalone representations for both sides, which is efficient for practical model deployment. Moreover, we enhance user profiling by exploiting large language model (LLM) to extract global interests from user engagement history. Extensive experiments on two benchmark datasets demonstrate that our framework outperforms existing state-of-the-art (SoTA) methods.

Jie Lei, Héctor Martínez, Adrián Castelló

The growing adoption of RISC-V in high-performance and scientific computing has increased the need for performance-portable code targeting the RISC-V Vector (RVV) extension. However, current compiler infrastructures provide limited end-to-end support for generating optimized RVV code from high-level representations to low-level implementations. In particular, existing MLIR distributions lack practical lowering paths that map high-level abstractions to RVV intrinsics, limiting their applicability for production-ready RISC-V kernels. This paper presents a compilation approach that combines MLIR with xDSL to bridge the missing lowering stages required for RVV code generation. Using custom intermediate representations and transformation passes implemented in xDSL, we systematically translate high-level operations into specialized, hardware-aware C code invoking RVV intrinsics. The resulting kernels are emitted as portable C functions that can be directly integrated into existing applications, enabling incremental adoption without modifying surrounding software stacks. We demonstrate the approach on the General Matrix Multiplication (GEMM) kernel and evaluate the generated micro-kernels on two real RISC-V platforms, the K230 and the BananaPi F3, comparing against OpenBLAS for both square-matrix benchmarks and transformer-based workloads derived from the BERT-Large model. When integrated into a matrix multiplication kernel, the proposed approach consistently outperforms OpenBLAS, reaching up to 12.2 GFLOPS compared to the baseline's 5.1 GFLOPS and providing performance improvements between 10--35\% across the evaluated workloads. These results demonstrate that combining MLIR with xDSL provides a practical pathway to portable, optimized code generation for RISC-V platforms.

Shuhan Dong, Yunsong Li, Weiying Xie et al.

Multimodal object detection leverages diverse modal information to enhance the accuracy and robustness of detectors. By learning long-term dependencies, Transformer can effectively integrate multimodal features in the feature extraction stage, which greatly improves the performance of multimodal object detection. However, current methods merely stack Transformer-guided fusion techniques without exploring their capability to extract features at various depth layers of network, thus limiting the improvements in detection performance. In this paper, we introduce an accurate and efficient object detection method named SeaDATE. Initially, we propose a novel dual attention Feature Fusion (DTF) module that, under Transformer's guidance, integrates local and global information through a dual attention mechanism, strengthening the fusion of modal features from orthogonal perspectives using spatial and channel tokens. Meanwhile, our theoretical analysis and empirical validation demonstrate that the Transformer-guided fusion method, treating images as sequences of pixels for fusion, performs better on shallow features' detail information compared to deep semantic information. To address this, we designed a contrastive learning (CL) module aimed at learning features of multimodal samples, remedying the shortcomings of Transformer-guided fusion in extracting deep semantic features, and effectively utilizing cross-modal information. Extensive experiments and ablation studies on the FLIR, LLVIP, and M3FD datasets have proven our method to be effective, achieving state-of-the-art detection performance.

Kai Jiang, Jiaxing Huang, Weiying Xie et al.

Large-vocabulary object detectors (LVDs) aim to detect objects of many categories, which learn super objectness features and can locate objects accurately while applied to various downstream data. However, LVDs often struggle in recognizing the located objects due to domain discrepancy in data distribution and object vocabulary. At the other end, recent vision-language foundation models such as CLIP demonstrate superior open-vocabulary recognition capability. This paper presents KGD, a Knowledge Graph Distillation technique that exploits the implicit knowledge graphs (KG) in CLIP for effectively adapting LVDs to various downstream domains. KGD consists of two consecutive stages: 1) KG extraction that employs CLIP to encode downstream domain data as nodes and their feature distances as edges, constructing KG that inherits the rich semantic relations in CLIP explicitly; and 2) KG encapsulation that transfers the extracted KG into LVDs to enable accurate cross-domain object classification. In addition, KGD can extract both visual and textual KG independently, providing complementary vision and language knowledge for object localization and object classification in detection tasks over various downstream domains. Experiments over multiple widely adopted detection benchmarks show that KGD outperforms the state-of-the-art consistently by large margins.

Mingxiang Cao, Weiying Xie, Jie Lei et al.

Deep learning has driven significant progress in object detection using Synthetic Aperture Radar (SAR) imagery. Existing methods, while achieving promising results, often struggle to effectively integrate local and global information, particularly direction-aware features. This paper proposes SAR-Net, a novel framework specifically designed for global fusion of direction-aware information in SAR object detection. SAR-Net leverages two key innovations: the Unity Compensation Mechanism (UCM) and the Direction-aware Attention Module (DAM). UCM facilitates the establishment of complementary relationships among features across different scales, enabling efficient global information fusion and transmission. Additionally, DAM, through bidirectional attention polymerization, captures direction-aware information, effectively eliminating background interference. Extensive experiments demonstrate the effectiveness of SAR-Net, achieving state-of-the-art results on aircraft (SAR-AIRcraft-1.0) and ship datasets (SSDD, HRSID), confirming its generalization capability and robustness.

Yuxin Wang, Zunlei Feng, Haofei Zhang et al.

Due to the inability to receive signals from the Global Navigation Satellite System (GNSS) in extreme conditions, achieving accurate and robust navigation for Unmanned Aerial Vehicles (UAVs) is a challenging task. Recently emerged, vision-based navigation has been a promising and feasible alternative to GNSS-based navigation. However, existing vision-based techniques are inadequate in addressing flight deviation caused by environmental disturbances and inaccurate position predictions in practical settings. In this paper, we present a novel angle robustness navigation paradigm to deal with flight deviation in point-to-point navigation tasks. Additionally, we propose a model that includes the Adaptive Feature Enhance Module, Cross-knowledge Attention-guided Module and Robust Task-oriented Head Module to accurately predict direction angles for high-precision navigation. To evaluate the vision-based navigation methods, we collect a new dataset termed as UAV_AR368. Furthermore, we design the Simulation Flight Testing Instrument (SFTI) using Google Earth to simulate different flight environments, thereby reducing the expenses associated with real flight testing. Experiment results demonstrate that the proposed model outperforms the state-of-the-art by achieving improvements of 26.0% and 45.6% in the success rate of arrival under ideal and disturbed circumstances, respectively.

Wenjin Guo, Donglai Liu, Weiying Xie et al.

Neural network training is a memory- and compute-intensive task. Quantization, which enables low-bitwidth formats in training, can significantly mitigate the workload. To reduce quantization error, recent methods have developed new data formats and additional pre-processing operations on quantizers. However, it remains quite challenging to achieve high accuracy and efficiency simultaneously. In this paper, we explore sub-8-bit integer training from its essence of gradient descent optimization. Our integer training framework includes two components: ShiftQuant to realize accurate gradient estimation, and L1 normalization to smoothen the loss landscape. ShiftQuant attains performance that approaches the theoretical upper bound of group quantization. Furthermore, it liberates group quantization from inefficient memory rearrangement. The L1 normalization facilitates the implementation of fully quantized normalization layers with impressive convergence accuracy. Our method frees sub-8-bit integer training from pre-processing and supports general devices. This framework achieves negligible accuracy loss across various neural networks and tasks ($0.92\%$ on 4-bit ResNets, $0.61\%$ on 6-bit Transformers). The prototypical implementation of ShiftQuant achieves more than $1.85\times/15.3\%$ performance improvement on CPU/GPU compared to its FP16 counterparts, and $33.9\%$ resource consumption reduction on FPGA than the FP16 counterparts. The proposed fully-quantized L1 normalization layers achieve more than $35.54\%$ improvement in throughout on CPU compared to traditional L2 normalization layers. Moreover, theoretical analysis verifies the advancement of our method.

Jitong Liao, Yulu Gao, Shaofei Huang et al.

Cross-view object correspondence involves matching objects between egocentric (first-person) and exocentric (third-person) views. It is a critical yet challenging task for visual understanding. In this work, we propose the Dense Object Matching and Refinement (DOMR) framework to establish dense object correspondences across views. The framework centers around the Dense Object Matcher (DOM) module, which jointly models multiple objects. Unlike methods that directly match individual object masks to image features, DOM leverages both positional and semantic relationships among objects to find correspondences. DOM integrates a proposal generation module with a dense matching module that jointly encodes visual, spatial, and semantic cues, explicitly constructing inter-object relationships to achieve dense matching among objects. Furthermore, we combine DOM with a mask refinement head designed to improve the completeness and accuracy of the predicted masks, forming the complete DOMR framework. Extensive evaluations on the Ego-Exo4D benchmark demonstrate that our approach achieves state-of-the-art performance with a mean IoU of 49.7% on Ego$\to$Exo and 55.2% on Exo$\to$Ego. These results outperform those of previous methods by 5.8% and 4.3%, respectively, validating the effectiveness of our integrated approach for cross-view understanding.

Jie Lei, Ruofan Jia, J. Andrew Zhang et al.

In wireless communication systems, stringent requirements such as ultra-low latency and power consumption have significantly increased the demand for efficient algorithm-to-hardware deployment. However, a persistent and substantial gap remains between algorithm design and hardware implementation. Bridging this gap traditionally requires extensive domain expertise and time-consuming manual development, due to fundamental mismatches between high-level programming languages like MATLAB and hardware description languages (HDLs) such as Verilog-in terms of memory access patterns, data processing manners, and datatype representations. To address this challenge, we propose A2HCoder: a Hierarchical Algorithm-to-HDL Coding Agent, powered by large language models (LLMs), designed to enable agile and reliable algorithm-to-hardware translation. A2HCoder introduces a hierarchical framework that enhances both robustness and interpretability while suppressing common hallucination issues in LLM-generated code. In the horizontal dimension, A2HCoder decomposes complex algorithms into modular functional blocks, simplifying code generation and improving consistency. In the vertical dimension, instead of relying on end-to-end generation, A2HCoder performs step-by-step, fine-grained translation, leveraging external toolchains such as MATLAB and Vitis HLS for debugging and circuit-level synthesis. This structured process significantly mitigates hallucinations and ensures hardware-level correctness. We validate A2HCoder through a real-world deployment case in the 5G wireless communication domain, demonstrating its practicality, reliability, and deployment efficiency.

Binjia Zhou, Hengrui Lou, Lizhe Chen et al.

With the swift progression of image generation technology, the widespread emergence of facial deepfakes poses significant challenges to the field of security, thus amplifying the urgent need for effective deepfake detection.Existing techniques for face forgery detection can broadly be categorized into two primary groups: visual-based methods and multimodal approaches. The former often lacks clear explanations for forgery details, while the latter, which merges visual and linguistic modalities, is more prone to the issue of hallucinations.To address these shortcomings, we introduce a visual detail enhanced self-correction framework, designated CorrDetail, for interpretable face forgery detection. CorrDetail is meticulously designed to rectify authentic forgery details when provided with error-guided questioning, with the aim of fostering the ability to uncover forgery details rather than yielding hallucinated responses. Additionally, to bolster the reliability of its findings, a visual fine-grained detail enhancement module is incorporated, supplying CorrDetail with more precise visual forgery details. Ultimately, a fusion decision strategy is devised to further augment the model's discriminative capacity in handling extreme samples, through the integration of visual information compensation and model bias reduction.Experimental results demonstrate that CorrDetail not only achieves state-of-the-art performance compared to the latest methodologies but also excels in accurately identifying forged details, all while exhibiting robust generalization capabilities.

Ruofan Jia, Weiying Xie, Jie Lei et al.

While large pre-trained models have achieved impressive performance across AI tasks, their deployment in privacy-sensitive and distributed environments remains challenging. Federated learning (FL) offers a viable solution by enabling decentralized fine-tuning without data sharing, but real-world applications face significant obstacles due to heterogeneous client resources in compute and memory. To address this, we propose HeteroTune, a novel federated fine-tuning paradigm for large, heterogeneous models operating under limited communication and computation budgets. The core of our method lies in a novel architecture, DeMA (Dense Mixture of Adapters), which enables flexible and efficient aggregation of heterogeneous models by preserving their full representational capacity while facilitating seamless cross-model knowledge fusion. We further introduce CMGA (Cross-Model Gradient Alignment), a lightweight yet effective mechanism that enhances training stability by harmonizing gradient directions across heterogeneous client models during aggregation, mitigating update conflicts and promoting more consistent convergence in federated settings. We provide both theoretical analysis and empirical evidence showing that HeteroTune achieves state-of-the-art performance and efficiency across diverse tasks and model architectures. For example, on LLaMA models, it reduces communication overhead by 99.5%, cuts peak memory usage by ~50%, and improves performance by 4.61%.

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.

Jie Lei, Tousif Rahman, Rishad Shafik et al.

The emergence of Artificial Intelligence (AI) driven Keyword Spotting (KWS) technologies has revolutionized human to machine interaction. Yet, the challenge of end-to-end energy efficiency, memory footprint and system complexity of current Neural Network (NN) powered AI-KWS pipelines has remained ever present. This paper evaluates KWS utilizing a learning automata powered machine learning algorithm called the Tsetlin Machine (TM). Through significant reduction in parameter requirements and choosing logic over arithmetic based processing, the TM offers new opportunities for low-power KWS while maintaining high learning efficacy. In this paper we explore a TM based keyword spotting (KWS) pipeline to demonstrate low complexity with faster rate of convergence compared to NNs. Further, we investigate the scalability with increasing keywords and explore the potential for enabling low-power on-chip KWS.

K. Darshana Abeyrathna, Ole-Christoffer Granmo, Rishad Shafik et al.

Due to the high energy consumption and scalability challenges of deep learning, there is a critical need to shift research focus towards dealing with energy consumption constraints. Tsetlin Machines (TMs) are a recent approach to machine learning that has demonstrated significantly reduced energy usage compared to neural networks alike, while performing competitively accuracy-wise on several benchmarks. However, TMs rely heavily on energy-costly random number generation to stochastically guide a team of Tsetlin Automata to a Nash Equilibrium of the TM game. In this paper, we propose a novel finite-state learning automaton that can replace the Tsetlin Automata in TM learning, for increased determinism. The new automaton uses multi-step deterministic state jumps to reinforce sub-patterns. Simultaneously, flipping a coin to skip every $d$'th state update ensures diversification by randomization. The $d$-parameter thus allows the degree of randomization to be finely controlled. E.g., $d=1$ makes every update random and $d=\infty$ makes the automaton completely deterministic. Our empirical results show that, overall, only substantial degrees of determinism reduces accuracy. Energy-wise, random number generation constitutes switching energy consumption of the TM, saving up to 11 mW power for larger datasets with high $d$ values. We can thus use the new $d$-parameter to trade off accuracy against energy consumption, to facilitate low-energy machine learning.