Tao Zhu

Lei Cai, Yuli Fu, Wanliang Huo et al.

Multi-scale architectures and attention modules have shown effectiveness in many deep learning-based image de-raining methods. However, manually designing and integrating these two components into a neural network requires a bulk of labor and extensive expertise. In this article, a high-performance multi-scale attentive neural architecture search (MANAS) framework is technically developed for image deraining. The proposed method formulates a new multi-scale attention search space with multiple flexible modules that are favorite to the image de-raining task. Under the search space, multi-scale attentive cells are built, which are further used to construct a powerful image de-raining network. The internal multiscale attentive architecture of the de-raining network is searched automatically through a gradient-based search algorithm, which avoids the daunting procedure of the manual design to some extent. Moreover, in order to obtain a robust image de-raining model, a practical and effective multi-to-one training strategy is also presented to allow the de-raining network to get sufficient background information from multiple rainy images with the same background scene, and meanwhile, multiple loss functions including external loss, internal loss, architecture regularization loss, and model complexity loss are jointly optimized to achieve robust de-raining performance and controllable model complexity. Extensive experimental results on both synthetic and realistic rainy images, as well as the down-stream vision applications (i.e., objection detection and segmentation) consistently demonstrate the superiority of our proposed method. The code is publicly available at https://github.com/lcai-gz/MANAS.

Shen Yan, Tao Zhu, Zirui Wang et al. · cmu, deepmind

We explore an efficient approach to establish a foundational video-text model. We present VideoCoCa that maximally reuses a pretrained image-text contrastive captioner (CoCa) model and adapt it to video-text tasks with minimal extra training. While previous works adapt image-text models with various cross-frame fusion modules, we find that the generative attentional pooling and contrastive attentional pooling layers in CoCa are instantly adaptable to flattened frame embeddings, yielding state-of-the-art results on zero-shot video classification and zero-shot text-to-video retrieval. Furthermore, we explore lightweight finetuning on top of VideoCoCa, and achieve strong results on video question-answering and video captioning.

Chenglin Yang, Siyuan Qiao, Yuan Cao et al. · deepmind

Generative training has been demonstrated to be powerful for building visual-language models. However, on zero-shot discriminative benchmarks, there is still a performance gap between models trained with generative and discriminative objectives. In this paper, we aim to narrow this gap by improving the efficacy of generative training on classification tasks, without any finetuning processes or additional modules. Specifically, we focus on narrowing the gap between the generative captioner and the CLIP classifier. We begin by analysing the predictions made by the captioner and classifier and observe that the caption generation inherits the distribution bias from the language model trained with pure text modality, making it less grounded on the visual signal. To tackle this problem, we redesign the scoring objective for the captioner to alleviate the distributional bias and focus on measuring the gain of information brought by the visual inputs. We further design a generative training objective to match the evaluation objective. We name our model trained and evaluated from the novel procedures as Information Gain (IG) captioner. We pretrain the models on the public Laion-5B dataset and perform a series of discriminative evaluations. For the zero-shot classification on ImageNet, IG captioner achieves $> 18\%$ improvements over the standard captioner, achieving comparable performances with the CLIP classifier. IG captioner also demonstrated strong performance on zero-shot image-text retrieval tasks on MSCOCO and Flickr30K. We hope this paper inspires further research towards unifying generative and discriminative training procedures for visual-language models.

Jinqiang Wang, Tao Zhu, Liming Chen et al.

Contrastive learning has been applied to Human Activity Recognition (HAR) based on sensor data owing to its ability to achieve performance comparable to supervised learning with a large amount of unlabeled data and a small amount of labeled data. The pre-training task for contrastive learning is generally instance discrimination, which specifies that each instance belongs to a single class, but this will consider the same class of samples as negative examples. Such a pre-training task is not conducive to human activity recognition tasks, which are mainly classification tasks. To address this problem, we follow SimCLR to propose a new contrastive learning framework that negative selection by clustering in HAR, which is called ClusterCLHAR. Compared with SimCLR, it redefines the negative pairs in the contrastive loss function by using unsupervised clustering methods to generate soft labels that mask other samples of the same cluster to avoid regarding them as negative samples. We evaluate ClusterCLHAR on three benchmark datasets, USC-HAD, MotionSense, and UCI-HAR, using mean F1-score as the evaluation metric. The experiment results show that it outperforms all the state-of-the-art methods applied to HAR in self-supervised learning and semi-supervised learning.

Furong Duan, Tao Zhu, Jinqiang Wang et al.

Sensor-based human activity segmentation and recognition are two important and challenging problems in many real-world applications and they have drawn increasing attention from the deep learning community in recent years. Most of the existing deep learning works were designed based on pre-segmented sensor streams and they have treated activity segmentation and recognition as two separate tasks. In practice, performing data stream segmentation is very challenging. We believe that both activity segmentation and recognition may convey unique information which can complement each other to improve the performance of the two tasks. In this paper, we firstly proposes a new multitask deep neural network to solve the two tasks simultaneously. The proposed neural network adopts selective convolution and features multiscale windows to segment activities of long or short time durations. First, multiple windows of different scales are generated to center on each unit of the feature sequence. Then, the model is trained to predict, for each window, the activity class and the offset to the true activity boundaries. Finally, overlapping windows are filtered out by non-maximum suppression, and adjacent windows of the same activity are concatenated to complete the segmentation task. Extensive experiments were conducted on eight popular benchmarking datasets, and the results show that our proposed method outperforms the state-of-the-art methods both for activity recognition and segmentation.

Zhe Zhao, Yudong Li, Cheng Hou et al.

Recently, the success of pre-training in text domain has been fully extended to vision, audio, and cross-modal scenarios. The proposed pre-training models of different modalities are showing a rising trend of homogeneity in their model structures, which brings the opportunity to implement different pre-training models within a uniform framework. In this paper, we present TencentPretrain, a toolkit supporting pre-training models of different modalities. The core feature of TencentPretrain is the modular design. The toolkit uniformly divides pre-training models into 5 components: embedding, encoder, target embedding, decoder, and target. As almost all of common modules are provided in each component, users can choose the desired modules from different components to build a complete pre-training model. The modular design enables users to efficiently reproduce existing pre-training models or build brand-new one. We test the toolkit on text, vision, and audio benchmarks and show that it can match the performance of the original implementations.

Kunbo Ding, Weijie Liu, Yuejian Fang et al.

Existing zero-shot cross-lingual transfer methods rely on parallel corpora or bilingual dictionaries, which are expensive and impractical for low-resource languages. To disengage from these dependencies, researchers have explored training multilingual models on English-only resources and transferring them to low-resource languages. However, its effect is limited by the gap between embedding clusters of different languages. To address this issue, we propose Embedding-Push, Attention-Pull, and Robust targets to transfer English embeddings to virtual multilingual embeddings without semantic loss, thereby improving cross-lingual transferability. Experimental results on mBERT and XLM-R demonstrate that our method significantly outperforms previous works on the zero-shot cross-lingual text classification task and can obtain a better multilingual alignment.

Jinqiang Wang, Huansheng Ning, Jianguo Ding et al.

Proactively predicting a users next utterance in human-machine dialogue can streamline interaction and improve user experience. Existing commercial API-based solutions are subject to privacy concerns while deploying general-purpose LLMs locally remains computationally expensive. As such, training a compact, task-specific LLM provides a practical alternative. Although user simulator methods can predict a user's next utterance, they mainly imitate their speaking style rather than advancing the dialogue. Preference data synthesis has been investigated to generate data for proactive next utterance prediction and help align LLMs with user preferences. Yet existing methods lack the ability to explicitly model the intent reasoning that leads to the user's next utterance and to define and synthesize preference and non-preference reasoning processes for predicting the user's next utterance.To address these challenges, we propose ProUtt, an LLM-driven preference data synthesis method for proactive next utterance prediction. ProUtt converts dialogue history into an intent tree and explicitly models intent reasoning trajectories by predicting the next plausible path from both exploitation and exploration perspectives. It then constructs preference and non-preference reasoning processes by perturbing or revising intent tree paths at different future turns. Extensive evaluations using LLM-as-a-judge and human judgments demonstrate that ProUtt consistently outperforms existing data synthesis methods, user simulators, and commercial LLM APIs across four benchmark datasets. We release both the code and the synthesized datasets to facilitate future research.

Gemini Team, Petko Georgiev, Ving Ian Lei et al. · deepmind, mila

In this report, we introduce the Gemini 1.5 family of models, representing the next generation of highly compute-efficient multimodal models capable of recalling and reasoning over fine-grained information from millions of tokens of context, including multiple long documents and hours of video and audio. The family includes two new models: (1) an updated Gemini 1.5 Pro, which exceeds the February version on the great majority of capabilities and benchmarks; (2) Gemini 1.5 Flash, a more lightweight variant designed for efficiency with minimal regression in quality. Gemini 1.5 models achieve near-perfect recall on long-context retrieval tasks across modalities, improve the state-of-the-art in long-document QA, long-video QA and long-context ASR, and match or surpass Gemini 1.0 Ultra's state-of-the-art performance across a broad set of benchmarks. Studying the limits of Gemini 1.5's long-context ability, we find continued improvement in next-token prediction and near-perfect retrieval (>99%) up to at least 10M tokens, a generational leap over existing models such as Claude 3.0 (200k) and GPT-4 Turbo (128k). Finally, we highlight real-world use cases, such as Gemini 1.5 collaborating with professionals on completing their tasks achieving 26 to 75% time savings across 10 different job categories, as well as surprising new capabilities of large language models at the frontier; when given a grammar manual for Kalamang, a language with fewer than 200 speakers worldwide, the model learns to translate English to Kalamang at a similar level to a person who learned from the same content.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Yuntian Liu, Tao Zhu, Xiaoyang Liu et al.

Statement autoformalization, the automated translation of statements from natural language into formal languages, has become a subject of extensive research, yet the development of robust automated evaluation metrics remains limited. Existing evaluation methods often lack semantic understanding, face challenges with high computational costs, and are constrained by the current progress of automated theorem proving. To address these issues, we propose GTED (Generalized Tree Edit Distance), a novel evaluation framework that first standardizes formal statements and converts them into operator trees, then determines the semantic similarity using the eponymous GTED metric. Across the miniF2F and ProofNet benchmarks, GTED consistently ranks as a top-performing metric, achieving the highest accuracy and Kappa on miniF2F and the joint-highest accuracy on ProofNet. This strong overall performance provides the community with a computationally lightweight and more faithful metric for automated evaluation. The code and experimental results are available at https://github.com/XiaoyangLiu-sjtu/GTED.

Tao Zhu, Qi Yu, Xinru Dong et al.

Weakly-supervised video anomaly detection (WS-VAD) using Multiple Instance Learning (MIL) suffers from label ambiguity, hindering discriminative feature learning. We propose ProDisc-VAD, an efficient framework tackling this via two synergistic components. The Prototype Interaction Layer (PIL) provides controlled normality modeling using a small set of learnable prototypes, establishing a robust baseline without being overwhelmed by dominant normal data. The Pseudo-Instance Discriminative Enhancement (PIDE) loss boosts separability by applying targeted contrastive learning exclusively to the most reliable extreme-scoring instances (highest/lowest scores). ProDisc-VAD achieves strong AUCs (97.98% ShanghaiTech, 87.12% UCF-Crime) using only 0.4M parameters, over 800x fewer than recent ViT-based methods like VadCLIP. Code is available at https://github.com/modadundun/ProDisc-VAD.

Jinqiang Wang, Huansheng Ning, Yi Peng et al.

Large Language Models (LLMs) have demonstrated surprising performance across various natural language processing tasks. Recently, medical LLMs enhanced with domain-specific knowledge have exhibited excellent capabilities in medical consultation and diagnosis. These models can smoothly simulate doctor-patient dialogues and provide professional medical advice. Most medical LLMs are developed through continued training of open-source general LLMs, which require significantly fewer computational resources than training LLMs from scratch. Additionally, this approach offers better patient privacy protection than API-based solutions. Given the above advantages, this survey systematically summarizes how to train medical LLMs based on open-source general LLMs from a more fine-grained perspective. It covers (a) how to acquire training corpus and construct customized medical training sets, (b) how to choose an appropriate training paradigm, (c) how to choose a suitable evaluation benchmark, and (d) existing challenges and promising research directions are discussed. This survey can provide guidance for the development of LLMs focused on various medical applications, such as medical education, diagnostic planning, and clinical assistants. Related resources and supplemental information can be found on the GitHub repository.

Qi Qiu, Tao Zhu, Helin Gong et al.

Limited by the complexity of basis function (B-spline) calculations, Kolmogorov-Arnold Networks (KAN) suffer from restricted parallel computing capability on GPUs. This paper proposes a novel ReLU-KAN implementation that inherits the core idea of KAN. By adopting ReLU (Rectified Linear Unit) and point-wise multiplication, we simplify the design of KAN's basis function and optimize the computation process for efficient CUDA computing. The proposed ReLU-KAN architecture can be readily implemented on existing deep learning frameworks (e.g., PyTorch) for both inference and training. Experimental results demonstrate that ReLU-KAN achieves a 20x speedup compared to traditional KAN with 4-layer networks. Furthermore, ReLU-KAN exhibits a more stable training process with superior fitting ability while preserving the "catastrophic forgetting avoidance" property of KAN. You can get the code in https://github.com/quiqi/relu_kan

Weijie Liu, Tao Zhu, Weiquan Mao et al.

In this paper, we pay attention to the issue which is usually overlooked, i.e., \textit{similarity should be determined from different perspectives}. To explore this issue, we release a Multi-Perspective Text Similarity (MPTS) dataset, in which sentence similarities are labeled from twelve perspectives. Furthermore, we conduct a series of experimental analysis on this task by retrofitting some famous text matching models. Finally, we obtain several conclusions and baseline models, laying the foundation for the following investigation of this issue. The dataset and code are publicly available at Github\footnote{\url{https://github.com/autoliuweijie/MPTS}

Lijie Fan, Luming Tang, Siyang Qin et al. · deepmind

We present UniFluid, a unified autoregressive framework for joint visual generation and understanding leveraging continuous visual tokens. Our unified autoregressive architecture processes multimodal image and text inputs, generating discrete tokens for text and continuous tokens for image. We find though there is an inherent trade-off between the image generation and understanding task, a carefully tuned training recipe enables them to improve each other. By selecting an appropriate loss balance weight, the unified model achieves results comparable to or exceeding those of single-task baselines on both tasks. Furthermore, we demonstrate that employing stronger pre-trained LLMs and random-order generation during training is important to achieve high-fidelity image generation within this unified framework. Built upon the Gemma model series, UniFluid exhibits competitive performance across both image generation and understanding, demonstrating strong transferability to various downstream tasks, including image editing for generation, as well as visual captioning and question answering for understanding.

Shuangjian Li, Tao Zhu, Furong Duan et al.

Wearable sensor-based human activity recognition (HAR) is a critical research domain in activity perception. However, achieving high efficiency and long sequence recognition remains a challenge. Despite the extensive investigation of temporal deep learning models, such as CNNs, RNNs, and transformers, their extensive parameters often pose significant computational and memory constraints, rendering them less suitable for resource-constrained mobile health applications. This study introduces HARMamba, an innovative light-weight and versatile HAR architecture that combines selective bidirectional State Spaces Model and hardware-aware design. To optimize real-time resource consumption in practical scenarios, HARMamba employs linear recursive mechanisms and parameter discretization, allowing it to selectively focus on relevant input sequences while efficiently fusing scan and recompute operations. The model employs independent channels to process sensor data streams, dividing each channel into patches and appending classification tokens to the end of the sequence. It utilizes position embedding to represent the sequence order. The patch sequence is subsequently processed by HARMamba Block, and the classification head finally outputs the activity category. The HARMamba Block serves as the fundamental component of the HARMamba architecture, enabling the effective capture of more discriminative activity sequence features. HARMamba outperforms contemporary state-of-the-art frameworks, delivering comparable or better accuracy with significantly reducing computational and memory demands. It's effectiveness has been extensively validated on 4 publically available datasets namely PAMAP2, WISDM, UNIMIB SHAR and UCI. The F1 scores of HARMamba on the four datasets are 99.74%, 99.20%, 88.23% and 97.01%, respectively.

Lulu Ban, Tao Zhu, Xiangqing Lu et al.

Cross-dataset Human Activity Recognition (HAR) suffers from limited model generalization, hindering its practical deployment. To address this critical challenge, inspired by the success of DoReMi in Large Language Models (LLMs), we introduce a data mixture optimization strategy for pre-training HAR models, aiming to improve the recognition performance across heterogeneous datasets. However, directly applying DoReMi to the HAR field encounters new challenges due to the continuous, multi-channel and intrinsic heterogeneous characteristics of IMU sensor data. To overcome these limitations, we propose a novel framework HAR-DoReMi, which introduces a masked reconstruction task based on Mean Squared Error (MSE) loss. By raplacing the discrete language sequence prediction task, which relies on the Negative Log-Likelihood (NLL) loss, in the original DoReMi framework, the proposed framework is inherently more appropriate for handling the continuous and multi-channel characteristics of IMU data. In addition, HAR-DoReMi integrates the Mahony fusion algorithm into the self-supervised HAR pre-training, aiming to mitigate the heterogeneity of varying sensor orientation. This is achieved by estimating the sensor orientation within each dataset and facilitating alignment with a unified coordinate system, thereby improving the cross-dataset generalization ability of the HAR model. Experimental evaluation on multiple cross-dataset HAR transfer tasks demonstrates that HAR-DoReMi improves the accuracy by an average of 6.51%, compared to the current state-of-the-art method with only approximately 30% to 50% of the data usage. These results confirm the effectiveness of HAR-DoReMi in improving the generalization and data efficiency of pre-training HAR models, underscoring its significant potential to facilitate the practical deployment of HAR technology.

Guifeng Wang, Yuanfeng Song, Meng Yang et al.

Text-to-SQL, a pivotal natural language processing (NLP) task that converts textual queries into executable SQL, has seen substantial progress in recent years. However, existing evaluation and reward mechanisms used to train and assess the text-to-SQL models remain a critical bottleneck. Current approaches heavily rely on manually annotated gold SQL queries, which are costly to produce and impractical for large-scale evaluation. More importantly, most reinforcement learning (RL) methods in text-to-SQL leverage only the final binary execution outcome as the reward signal, a coarse-grained supervision that overlooks detailed structural and semantic errors from the perspective of rubrics. To address these challenges, we propose RuCo-C, a novel generative judge model for fine-grained, query-specific automatic evaluation using interpretable critiques without human intervention. Our framework first automatically generates query-specific evaluation rubrics for human-free annotation, linking them to interpretable critiques. Subsequently, it integrates densified reward feedback through a "progressive exploration" strategy during the RL training process, which dynamically adjusts the rewards to enhance the model's performance. Comprehensive experiments demonstrate that RuCo-C outperforms existing methods in text-to-SQL evaluation, yielding significant performance gains.

Tao Zhu, Yinfeng Yu, Liejun Wang et al.

Diffusion models have demonstrated remarkable performance in speech synthesis, but typically require multi-step sampling, resulting in low inference efficiency. Recent studies address this issue by distilling diffusion models into consistency models, enabling efficient one-step generation. However, these approaches introduce additional training costs and rely heavily on the performance of pre-trained teacher models. In this paper, we propose ECTSpeech, a simple and effective one-step speech synthesis framework that, for the first time, incorporates the Easy Consistency Tuning (ECT) strategy into speech synthesis. By progressively tightening consistency constraints on a pre-trained diffusion model, ECTSpeech achieves high-quality one-step generation while significantly reducing training complexity. In addition, we design a multi-scale gate module (MSGate) to enhance the denoiser's ability to fuse features at different scales. Experimental results on the LJSpeech dataset demonstrate that ECTSpeech achieves audio quality comparable to state-of-the-art methods under single-step sampling, while substantially reducing the model's training cost and complexity.

Xiaoyang Liu, Tao Zhu, Zineng Dong et al.

Statement autoformalization, the automated translation of statements from natural language into formal languages, has seen significant advancements, yet the development of automated evaluation metrics remains limited. Existing metrics for formal statement similarity often fail to balance semantic and structural information. String-based approaches capture syntactic structure but ignore semantic meaning, whereas proof-based methods validate semantic equivalence but disregard structural nuances and, critically, provide no graded similarity score in the event of proof failure. To address these issues, we introduce ASSESS (A Semantic and Structural Evaluation Framework for Statement Similarity), which comprehensively integrates semantic and structural information to provide a continuous similarity score. Our framework first transforms formal statements into Operator Trees to capture their syntactic structure and then computes a similarity score using our novel TransTED (Transformation Tree Edit Distance) Similarity metric, which enhances traditional Tree Edit Distance by incorporating semantic awareness through transformations. For rigorous validation, we present EPLA (Evaluating Provability and Likeness for Autoformalization), a new benchmark of 524 expert-annotated formal statement pairs derived from miniF2F and ProofNet, with labels for both semantic provability and structural likeness. Experiments on EPLA demonstrate that TransTED Similarity outperforms existing methods, achieving state-of-the-art accuracy and the highest Kappa coefficient. The benchmark, and implementation code will be made public soon.

Jinqiang Wang, Huansheng Ning, Tao Zhu et al.

In the tourism domain, Large Language Models (LLMs) often struggle to mine implicit user intentions from tourists' ambiguous inquiries and lack the capacity to proactively guide users toward clarifying their needs. A critical bottleneck is the scarcity of high-quality training datasets that facilitate proactive questioning and implicit intention mining. While recent advances leverage LLM-driven data synthesis to generate such datasets and transfer specialized knowledge to downstream models, existing approaches suffer from several shortcomings: (1) lack of adaptation to the tourism domain, (2) skewed distributions of detail levels in initial inquiries, (3) contextual redundancy in the implicit intention mining module, and (4) lack of explicit thinking about tourists' emotions and intention values. Therefore, we propose SynPT (A Data Synthesis Method Driven by LLMs for Proactive Mining of Implicit User Intentions in the Tourism), which constructs an LLM-driven user agent and assistant agent to simulate dialogues based on seed data collected from Chinese tourism websites. This approach addresses the aforementioned limitations and generates SynPT-Dialog, a training dataset containing explicit reasoning. The dataset is utilized to fine-tune a general LLM, enabling it to proactively mine implicit user intentions. Experimental evaluations, conducted from both human and LLM perspectives, demonstrate the superiority of SynPT compared to existing methods. Furthermore, we analyze key hyperparameters and present case studies to illustrate the practical applicability of our method, including discussions on its adaptability to English-language scenarios. All code and data are publicly available.

Yuanlong Wu, Mingxing Nie, Tao Zhu et al.

Class-incremental learning (CIL) for time series data faces critical challenges in balancing stability against catastrophic forgetting and plasticity for new knowledge acquisition, particularly under real-world constraints where historical data access is restricted. While pre-trained models (PTMs) have shown promise in CIL for vision and NLP domains, their potential in time series class-incremental learning (TSCIL) remains underexplored due to the scarcity of large-scale time series pre-trained models. Prompted by the recent emergence of large-scale pre-trained models (PTMs) for time series data, we present the first exploration of PTM-based Time Series Class-Incremental Learning (TSCIL). Our approach leverages frozen PTM backbones coupled with incrementally tuning the shared adapter, preserving generalization capabilities while mitigating feature drift through knowledge distillation. Furthermore, we introduce a Feature Drift Compensation Network (DCN), designed with a novel two-stage training strategy to precisely model feature space transformations across incremental tasks. This allows for accurate projection of old class prototypes into the new feature space. By employing DCN-corrected prototypes, we effectively enhance the unified classifier retraining, mitigating model feature drift and alleviating catastrophic forgetting. Extensive experiments on five real-world datasets demonstrate state-of-the-art performance, with our method yielding final accuracy gains of 1.4%-6.1% across all datasets compared to existing PTM-based approaches. Our work establishes a new paradigm for TSCIL, providing insights into stability-plasticity optimization for continual learning systems.

Qi Qiu, Tao Zhu, Furong Duan et al.

Inertial Measurement Unit (IMU) sensors are widely employed for Human Activity Recognition (HAR) due to their portability, energy efficiency, and growing research interest. However, a significant challenge for IMU-HAR models is achieving robust generalization performance across diverse users. This limitation stems from substantial variations in data distribution among individual users. One primary reason for this distribution disparity lies in the representation of IMU sensor data in the local coordinate system, which is susceptible to subtle user variations during IMU wearing. To address this issue, we propose a novel approach that extracts a global view representation based on the characteristics of IMU data, effectively alleviating the data distribution discrepancies induced by wearing styles. To validate the efficacy of the global view representation, we fed both global and local view data into model for experiments. The results demonstrate that global view data significantly outperforms local view data in cross-user experiments. Furthermore, we propose a Multi-view Supervised Network (MVFNet) based on Shuffling to effectively fuse local view and global view data. It supervises the feature extraction of each view through view division and view shuffling, so as to avoid the model ignoring important features as much as possible. Extensive experiments conducted on OPPORTUNITY and PAMAP2 datasets demonstrate that the proposed algorithm outperforms the current state-of-the-art methods in cross-user HAR.

Shuangjian Li, Tao Zhu, Mingxing Nie et al.

Traditional deep learning methods struggle to simultaneously segment, recognize, and forecast human activities from sensor data. This limits their usefulness in many fields such as healthcare and assisted living, where real-time understanding of ongoing and upcoming activities is crucial. This paper introduces P2LHAP, a novel Patch-to-Label Seq2Seq framework that tackles all three tasks in a efficient single-task model. P2LHAP divides sensor data streams into a sequence of "patches", served as input tokens, and outputs a sequence of patch-level activity labels including the predicted future activities. A unique smoothing technique based on surrounding patch labels, is proposed to identify activity boundaries accurately. Additionally, P2LHAP learns patch-level representation by sensor signal channel-independent Transformer encoders and decoders. All channels share embedding and Transformer weights across all sequences. Evaluated on three public datasets, P2LHAP significantly outperforms the state-of-the-art in all three tasks, demonstrating its effectiveness and potential for real-world applications.

Gemini Team, Rohan Anil, Sebastian Borgeaud et al.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of the Gemini family in cross-modal reasoning and language understanding will enable a wide variety of use cases. We discuss our approach toward post-training and deploying Gemini models responsibly to users through services including Gemini, Gemini Advanced, Google AI Studio, and Cloud Vertex AI.

Jinqiang Wang, Tao Zhu, Jingyuan Gan et al.

While deep learning has contributed to the advancement of sensor-based Human Activity Recognition (HAR), it is usually a costly and challenging supervised task with the needs of a large amount of labeled data. To alleviate this issue, contrastive learning has been applied for sensor-based HAR. Data augmentation is an essential part of contrastive learning and has a significant impact on the performance of downstream tasks. However, current popular augmentation methods do not achieve competitive performance in contrastive learning for sensor-based HAR. Motivated by this issue, we propose a new sensor data augmentation method by resampling, which simulates more realistic activity data by varying the sampling frequency to maximize the coverage of the sampling space. In addition, we extend MoCo, a popular contrastive learning framework, to MoCoHAR for HAR. The resampling augmentation method will be evaluated on two contrastive learning frameworks, SimCLRHAR and MoCoHAR, using UCI-HAR, MotionSensor, and USC-HAD datasets. The experiment results show that the resampling augmentation method outperforms all state-of-the-art methods under a small amount of labeled data, on SimCLRHAR and MoCoHAR, with mean F1-score as the evaluation metric. The results also demonstrate that not all data augmentation methods have positive effects in the contrastive learning framework.

Andrew Cotter, Maya Gupta, Heinrich Jiang et al.

We propose learning flexible but interpretable functions that aggregate a variable-length set of permutation-invariant feature vectors to predict a label. We use a deep lattice network model so we can architect the model structure to enhance interpretability, and add monotonicity constraints between inputs-and-outputs. We then use the proposed set function to automate the engineering of dense, interpretable features from sparse categorical features, which we call semantic feature engine. Experiments on real-world data show the achieved accuracy is similar to deep sets or deep neural networks, and is easier to debug and understand.

Tao Zhu, David Phipps, Adam Pridgen et al.

Weibo and other popular Chinese microblogging sites are well known for exercising internal censorship, to comply with Chinese government requirements. This research seeks to quantify the mechanisms of this censorship: how fast and how comprehensively posts are deleted.Our analysis considered 2.38 million posts gathered over roughly two months in 2012, with our attention focused on repeatedly visiting "sensitive" users. This gives us a view of censorship events within minutes of their occurrence, albeit at a cost of our data no longer representing a random sample of the general Weibo population. We also have a larger 470 million post sampling from Weibo's public timeline, taken over a longer time period, that is more representative of a random sample. We found that deletions happen most heavily in the first hour after a post has been submitted. Focusing on original posts, not reposts/retweets, we observed that nearly 30% of the total deletion events occur within 5- 30 minutes. Nearly 90% of the deletions happen within the first 24 hours. Leveraging our data, we also considered a variety of hypotheses about the mechanisms used by Weibo for censorship, such as the extent to which Weibo's censors use retrospective keyword-based censorship, and how repost/retweet popularity interacts with censorship. We also used natural language processing techniques to analyze which topics were more likely to be censored.