Masked image modeling (MIM) performs strongly in pre-training large vision Transformers (ViTs). However, small models that are critical for real-world applications cannot or only marginally benefit from this pre-training approach. In this paper, we explore distillation techniques to transfer the success of large MIM-based pre-trained models to smaller ones. We systematically study different options in the distillation framework, including distilling targets, losses, input, network regularization, sequential distillation, etc, revealing that: 1) Distilling token relations is more effective than CLS token- and feature-based distillation; 2) An intermediate layer of the teacher network as target perform better than that using the last layer when the depth of the student mismatches that of the teacher; 3) Weak regularization is preferred; etc. With these findings, we achieve significant fine-tuning accuracy improvements over the scratch MIM pre-training on ImageNet-1K classification, using all the ViT-Tiny, ViT-Small, and ViT-base models, with +4.2%/+2.4%/+1.4% gains, respectively. Our TinyMIM model of base size achieves 52.2 mIoU in AE20K semantic segmentation, which is +4.1 higher than the MAE baseline. Our TinyMIM model of tiny size achieves 79.6% top-1 accuracy on ImageNet-1K image classification, which sets a new record for small vision models of the same size and computation budget. This strong performance suggests an alternative way for developing small vision Transformer models, that is, by exploring better training methods rather than introducing inductive biases into architectures as in most previous works. Code is available at https://github.com/OliverRensu/TinyMIM.

Reading comprehension of legal text can be a particularly challenging task due to the length and complexity of legal clauses and a shortage of expert-annotated datasets. To address this challenge, we introduce the Merger Agreement Understanding Dataset (MAUD), an expert-annotated reading comprehension dataset based on the American Bar Association's 2021 Public Target Deal Points Study, with over 39,000 examples and over 47,000 total annotations. Our fine-tuned Transformer baselines show promising results, with models performing well above random on most questions. However, on a large subset of questions, there is still room for significant improvement. As the only expert-annotated merger agreement dataset, MAUD is valuable as a benchmark for both the legal profession and the NLP community.

Blockchain, also coined as decentralized AI, has the potential to empower AI to be more trustworthy by creating a decentralized trust of privacy, security, and audibility. However, systematic studies on the design principle of Blockchain as a trust engine for an integrated society of Cyber-Physical-Socia-System (CPSS) are still absent. In this article, we provide an initiative for seeking the design principle of Blockchain for a better digital world. Using a hybrid method of qualitative and quantitative studies, we examine the past origin, the current development, and the future directions of Blockchain design principles. We have three findings. First, the answers to whether Blockchain lives up to its original design principle as a distributed database are controversial. Second, the current development of Blockchain community reveals a taxonomy of 7 categories, including privacy and security, scalability, decentralization, applicability, governance and regulation, system design, and cross-chain interoperability. Both research and practice are more centered around the first category of privacy and security and the fourth category of applicability. Future scholars, practitioners, and policy-makers have vast opportunities in other, much less exploited facets and the synthesis at the interface of multiple aspects. Finally, in counter-examples, we conclude that a synthetic solution that crosses discipline boundaries is necessary to close the gaps between the current design of Blockchain and the design principle of a trust engine for a truly intelligent world.

In this paper, we develop an efficient multi-scale network to predict action classes in partial videos in an end-to-end manner. Unlike most existing methods with offline feature generation, our method directly takes frames as input and further models motion evolution on two different temporal scales.Therefore, we solve the complexity problems of the two stages of modeling and the problem of insufficient temporal and spatial information of a single scale. Our proposed End-to-End MultiScale Network (E2EMSNet) is composed of two scales which are named segment scale and observed global scale. The segment scale leverages temporal difference over consecutive frames for finer motion patterns by supplying 2D convolutions. For observed global scale, a Long Short-Term Memory (LSTM) is incorporated to capture motion features of observed frames. Our model provides a simple and efficient modeling framework with a small computational cost. Our E2EMSNet is evaluated on three challenging datasets: BIT, HMDB51, and UCF101. The extensive experiments demonstrate the effectiveness of our method for action prediction in videos.

With the attention mechanism, transformers achieve significant empirical successes. Despite the intuitive understanding that transformers perform relational inference over long sequences to produce desirable representations, we lack a rigorous theory on how the attention mechanism achieves it. In particular, several intriguing questions remain open: (a) What makes a desirable representation? (b) How does the attention mechanism infer the desirable representation within the forward pass? (c) How does a pretraining procedure learn to infer the desirable representation through the backward pass? We observe that, as is the case in BERT and ViT, input tokens are often exchangeable since they already include positional encodings. The notion of exchangeability induces a latent variable model that is invariant to input sizes, which enables our theoretical analysis. - To answer (a) on representation, we establish the existence of a sufficient and minimal representation of input tokens. In particular, such a representation instantiates the posterior distribution of the latent variable given input tokens, which plays a central role in predicting output labels and solving downstream tasks. - To answer (b) on inference, we prove that attention with the desired parameter infers the latent posterior up to an approximation error, which is decreasing in input sizes. In detail, we quantify how attention approximates the conditional mean of the value given the key, which characterizes how it performs relational inference over long sequences. - To answer (c) on learning, we prove that both supervised and self-supervised objectives allow empirical risk minimization to learn the desired parameter up to a generalization error, which is independent of input sizes. Particularly, in the self-supervised setting, we identify a condition number that is pivotal to solving downstream tasks.

Pioneers of autonomous vehicles (AVs) promised to revolutionize the driving experience and driving safety. However, milestones in AVs have materialized slower than forecast. Two culprits are (1) the lack of verifiability of proposed state-of-the-art AV components, and (2) stagnation of pursuing next-level evaluations, e.g., vehicle-to-infrastructure (V2I) and multi-agent collaboration. In part, progress has been hampered by: the large volume of software in AVs, the multiple disparate conventions, the difficulty of testing across datasets and simulators, and the inflexibility of state-of-the-art AV components. To address these challenges, we present AVstack, an open-source, reconfigurable software platform for AV design, implementation, test, and analysis. AVstack solves the validation problem by enabling first-of-a-kind trade studies on datasets and physics-based simulators. AVstack solves the stagnation problem as a reconfigurable AV platform built on dozens of open-source AV components in a high-level programming language. We demonstrate the power of AVstack through longitudinal testing across multiple benchmark datasets and V2I-collaboration case studies that explore trade-offs of designing multi-sensor, multi-agent algorithms.

Due to the issue that existing wireless sensor network (WSN)-based anomaly detection methods only consider and analyze temporal features, in this paper, a self-supervised learning-based anomaly node detection method based on an autoencoder is designed. This method integrates temporal WSN data flow feature extraction, spatial position feature extraction and intermodal WSN correlation feature extraction into the design of the autoencoder to make full use of the spatial and temporal information of the WSN for anomaly detection. First, a fully connected network is used to extract the temporal features of nodes by considering a single mode from a local spatial perspective. Second, a graph neural network (GNN) is used to introduce the WSN topology from a global spatial perspective for anomaly detection and extract the spatial and temporal features of the data flows of nodes and their neighbors by considering a single mode. Then, the adaptive fusion method involving weighted summation is used to extract the relevant features between different models. In addition, this paper introduces a gated recurrent unit (GRU) to solve the long-term dependence problem of the time dimension. Eventually, the reconstructed output of the decoder and the hidden layer representation of the autoencoder are fed into a fully connected network to calculate the anomaly probability of the current system. Since the spatial feature extraction operation is advanced, the designed method can be applied to the task of large-scale network anomaly detection by adding a clustering operation. Experiments show that the designed method outperforms the baselines, and the F1 score reaches 90.6%, which is 5.2% higher than those of the existing anomaly detection methods based on unsupervised reconstruction and prediction. Code and model are available at https://github.com/GuetYe/anomaly_detection/GLSL

Three-dimensional (3D) technologies have been developing rapidly recent years, and have influenced industrial, medical, cultural, and many other fields. In this paper, we introduce an automatic 3D human head scanning-printing system, which provides a complete pipeline to scan, reconstruct, select, and finally print out physical 3D human heads. To enhance the accuracy of our system, we developed a consumer-grade composite sensor (including a gyroscope, an accelerometer, a digital compass, and a Kinect v2 depth sensor) as our sensing device. This sensing device is then mounted on a robot, which automatically rotates around the human subject with approximate 1-meter radius, to capture the full-view information. The data streams are further processed and fused into a 3D model of the subject using a tablet located on the robot. In addition, an automatic selection method, based on our specific system configurations, is proposed to select the head portion. We evaluated the accuracy of the proposed system by comparing our generated 3D head models, from both standard human head model and real human subjects, with the ones reconstructed from FastSCAN and Cyberware commercial laser scanning systems through computing and visualizing Hausdorff distances. Computational cost is also provided to further assess our proposed system.

One of the key challenges in deploying RL to real-world applications is to adapt to variations of unknown environment contexts, such as changing terrains in robotic tasks and fluctuated bandwidth in congestion control. Existing works on adaptation to unknown environment contexts either assume the contexts are the same for the whole episode or assume the context variables are Markovian. However, in many real-world applications, the environment context usually stays stable for a stochastic period and then changes in an abrupt and unpredictable manner within an episode, resulting in a segment structure, which existing works fail to address. To leverage the segment structure of piecewise stable context in real-world applications, in this paper, we propose a \textit{\textbf{Se}gmented \textbf{C}ontext \textbf{B}elief \textbf{A}ugmented \textbf{D}eep~(SeCBAD)} RL method. Our method can jointly infer the belief distribution over latent context with the posterior over segment length and perform more accurate belief context inference with observed data within the current context segment. The inferred belief context can be leveraged to augment the state, leading to a policy that can adapt to abrupt variations in context. We demonstrate empirically that SeCBAD can infer context segment length accurately and outperform existing methods on a toy grid world environment and Mujuco tasks with piecewise-stable context.

We address the task of open-world class-agnostic object detection, i.e., detecting every object in an image by learning from a limited number of base object classes. State-of-the-art RGB-based models suffer from overfitting the training classes and often fail at detecting novel-looking objects. This is because RGB-based models primarily rely on appearance similarity to detect novel objects and are also prone to overfitting short-cut cues such as textures and discriminative parts. To address these shortcomings of RGB-based object detectors, we propose incorporating geometric cues such as depth and normals, predicted by general-purpose monocular estimators. Specifically, we use the geometric cues to train an object proposal network for pseudo-labeling unannotated novel objects in the training set. Our resulting Geometry-guided Open-world Object Detector (GOOD) significantly improves detection recall for novel object categories and already performs well with only a few training classes. Using a single "person" class for training on the COCO dataset, GOOD surpasses SOTA methods by 5.0% AR@100, a relative improvement of 24%.