Feng Zhou

Zhijie Deng, Feng Zhou, Jun Zhu

Laplace approximation (LA) and its linearized variant (LLA) enable effortless adaptation of pretrained deep neural networks to Bayesian neural networks. The generalized Gauss-Newton (GGN) approximation is typically introduced to improve their tractability. However, LA and LLA are still confronted with non-trivial inefficiency issues and should rely on Kronecker-factored, diagonal, or even last-layer approximate GGN matrices in practical use. These approximations are likely to harm the fidelity of learning outcomes. To tackle this issue, inspired by the connections between LLA and neural tangent kernels (NTKs), we develop a Nystrom approximation to NTKs to accelerate LLA. Our method benefits from the capability of popular deep learning libraries for forward mode automatic differentiation, and enjoys reassuring theoretical guarantees. Extensive studies reflect the merits of the proposed method in aspects of both scalability and performance. Our method can even scale up to architectures like vision transformers. We also offer valuable ablation studies to diagnose our method. Code is available at \url{https://github.com/thudzj/ELLA}.

Li Pang, Xiangyong Cao, Datao Tang et al.

Hyperspectral image (HSI) plays a vital role in various fields such as agriculture and environmental monitoring. However, due to the expensive acquisition cost, the number of hyperspectral images is limited, degenerating the performance of downstream tasks. Although some recent studies have attempted to employ diffusion models to synthesize HSIs, they still struggle with the scarcity of HSIs, affecting the reliability and diversity of the generated images. Some studies propose to incorporate multi-modal data to enhance spatial diversity, but the spectral fidelity cannot be ensured. In addition, existing HSI synthesis models are typically uncontrollable or only support single-condition control, limiting their ability to generate accurate and reliable HSIs. To alleviate these issues, we propose HSIGene, a novel HSI generation foundation model which is based on latent diffusion and supports multi-condition control, allowing for more precise and reliable HSI generation. To enhance the spatial diversity of the training data while preserving spectral fidelity, we propose a new data augmentation method based on spatial super-resolution, in which HSIs are upscaled first, and thus abundant training patches could be obtained by cropping the high-resolution HSIs. In addition, to improve the perceptual quality of the augmented data, we introduce a novel two-stage HSI super-resolution framework, which first applies RGB bands super-resolution and then utilizes our proposed Rectangular Guided Attention Network (RGAN) for guided HSI super-resolution. Experiments demonstrate that the proposed model is capable of generating a vast quantity of realistic HSIs for downstream tasks such as denoising and super-resolution. The code and models are available at https://github.com/LiPang/HSIGene.

Qiang Meng, Feng Zhou

Most visual retrieval applications store feature vectors for downstream matching tasks. These vectors, from where user information can be spied out, will cause privacy leakage if not carefully protected. To mitigate privacy risks, current works primarily utilize non-invertible transformations or fully cryptographic algorithms. However, transformation-based methods usually fail to achieve satisfying matching performances while cryptosystems suffer from heavy computational overheads. In addition, secure levels of current methods should be improved to confront potential adversary attacks. To address these issues, this paper proposes a plug-in module called SecureVector that protects features by random permutations, 4L-DEC converting and existing homomorphic encryption techniques. For the first time, SecureVector achieves real-time and lossless feature matching among sanitized features, along with much higher security levels than current state-of-the-arts. Extensive experiments on face recognition, person re-identification, image retrieval, and privacy analyses demonstrate the effectiveness of our method. Given limited public projects in this field, codes of our method and implemented baselines are made open-source in https://github.com/IrvingMeng/SecureVector.

Yinghui Li, Haojing Huang, Shirong Ma et al.

Recently, the development and progress of Large Language Models (LLMs) have amazed the entire Artificial Intelligence community. Benefiting from their emergent abilities, LLMs have attracted more and more researchers to study their capabilities and performance on various downstream Natural Language Processing (NLP) tasks. While marveling at LLMs' incredible performance on all kinds of tasks, we notice that they also have excellent multilingual processing capabilities, such as Chinese. To explore the Chinese processing ability of LLMs, we focus on Chinese Text Correction, a fundamental and challenging Chinese NLP task. Specifically, we evaluate various representative LLMs on the Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC) tasks, which are two main Chinese Text Correction scenarios. Additionally, we also fine-tune LLMs for Chinese Text Correction to better observe the potential capabilities of LLMs. From extensive analyses and comparisons with previous state-of-the-art small models, we empirically find that the LLMs currently have both amazing performance and unsatisfactory behavior for Chinese Text Correction. We believe our findings will promote the landing and application of LLMs in the Chinese NLP community.

Feng Zhou, Jacqueline Meijer-Irons, Ambar Murillo

While Large Language Models (LLMs) offer a solution to the scale-versus-depth dilemma in qualitative analysis, the paradigm of maximizing automation is fundamentally at odds with the interpretive nature of qualitative inquiry. We argue that effective Human-AI collaboration is not an automation problem, but an interdependence problem. This paper reframes the design of "co-data" systems through the lens of Interdependence Theory, proposing a formal framework to structure human-AI productive interdependence. The framework guides the selection of an appropriate Level of Automation (LoA) for different stages of the qualitative analysis process by assessing task risk and the cost of validation. We present a case study where this framework led to a deliberately interdependent workflow, fostering the calibrated trust necessary for rigorous analysis. We conclude by presenting three design principles that instantiate this framework, demonstrating how to leverage AI as a powerful partner while preserving the human researcher's irreplaceable role in the transformation process of meaning-making.

Zhijie Deng, Feng Zhou, Jianfei Chen et al.

Deep Ensemble (DE) is an effective alternative to Bayesian neural networks for uncertainty quantification in deep learning. The uncertainty of DE is usually conveyed by the functional inconsistency among the ensemble members, say, the disagreement among their predictions. Yet, the functional inconsistency stems from unmanageable randomness and may easily collapse in specific cases. To render the uncertainty of DE reliable, we propose a refinement of DE where the functional inconsistency is explicitly characterized, and further tuned w.r.t. the training data and certain priori beliefs. Specifically, we describe the functional inconsistency with the empirical covariance of the functions dictated by ensemble members, which, along with the mean, define a Gaussian process (GP). Then, with specific priori uncertainty imposed, we maximize functional evidence lower bound to make the GP specified by DE approximate the Bayesian posterior. In this way, we relate DE to Bayesian inference to enjoy reliable Bayesian uncertainty. Moreover, we provide strategies to make the training efficient. Our approach consumes only marginally added training cost than the standard DE, but achieves better uncertainty quantification than DE and its variants across diverse scenarios.

Meng Chen, Jiawei Tu, Chao Qi et al.

The significant advancements in embodied vision navigation have raised concerns about its susceptibility to adversarial attacks exploiting deep neural networks. Investigating the adversarial robustness of embodied vision navigation is crucial, especially given the threat of 3D physical attacks that could pose risks to human safety. However, existing attack methods for embodied vision navigation often lack physical feasibility due to challenges in transferring digital perturbations into the physical world. Moreover, current physical attacks for object detection struggle to achieve both multi-view effectiveness and visual naturalness in navigation scenarios. To address this, we propose a practical attack method for embodied navigation by attaching adversarial patches to objects, where both opacity and textures are learnable. Specifically, to ensure effectiveness across varying viewpoints, we employ a multi-view optimization strategy based on object-aware sampling, which optimizes the patch's texture based on feedback from the vision-based perception model used in navigation. To make the patch inconspicuous to human observers, we introduce a two-stage opacity optimization mechanism, in which opacity is fine-tuned after texture optimization. Experimental results demonstrate that our adversarial patches decrease the navigation success rate by an average of 22.39%, outperforming previous methods in practicality, effectiveness, and naturalness. Code is available at: https://github.com/chen37058/Physical-Attacks-in-Embodied-Nav

Jiangfan Deng, Dewen Fan, Xiaosong Qiu et al.

Crowdedness caused by overlapping among similar objects is a ubiquitous challenge in the field of 2D visual object detection. In this paper, we first underline two main effects of the crowdedness issue: 1) IoU-confidence correlation disturbances (ICD) and 2) confused de-duplication (CDD). Then we explore a pathway of cracking these nuts from the perspective of data augmentation. Primarily, a particular copy-paste scheme is proposed towards making crowded scenes. Based on this operation, we first design a "consensus learning" method to further resist the ICD problem and then find out the pasting process naturally reveals a pseudo "depth" of object in the scene, which can be potentially used for alleviating CDD dilemma. Both methods are derived from magical using of the copy-pasting without extra cost for hand-labeling. Experiments show that our approach can easily improve the state-of-the-art detector in typical crowded detection task by more than 2% without any bells and whistles. Moreover, this work can outperform existing data augmentation strategies in crowded scenario.

Jackie Ayoub, Zifei Wang, Meitang Li et al.

Advanced driver assistance systems (ADAS) are designed to improve vehicle safety. However, it is difficult to achieve such benefits without understanding the causes and limitations of the current ADAS and their possible solutions. This study 1) investigated the limitations and solutions of ADAS through a literature review, 2) identified the causes and effects of ADAS through consumer complaints using natural language processing models, and 3) compared the major differences between the two. These two lines of research identified similar categories of ADAS causes, including human factors, environmental factors, and vehicle factors. However, academic research focused more on human factors of ADAS issues and proposed advanced algorithms to mitigate such issues while drivers complained more of vehicle factors of ADAS failures, which led to associated top consequences. The findings from these two sources tend to complement each other and provide important implications for the improvement of ADAS in the future.

Yuanzhen Luo, Qingyu Zhou, Feng Zhou

Keyphrase extraction (KPE) is an important task in Natural Language Processing for many scenarios, which aims to extract keyphrases that are present in a given document. Many existing supervised methods treat KPE as sequential labeling, span-level classification, or generative tasks. However, these methods lack the ability to utilize keyphrase information, which may result in biased results. In this study, we propose Diff-KPE, which leverages the supervised Variational Information Bottleneck (VIB) to guide the text diffusion process for generating enhanced keyphrase representations. Diff-KPE first generates the desired keyphrase embeddings conditioned on the entire document and then injects the generated keyphrase embeddings into each phrase representation. A ranking network and VIB are then optimized together with rank loss and classification loss, respectively. This design of Diff-KPE allows us to rank each candidate phrase by utilizing both the information of keyphrases and the document. Experiments show that Diff-KPE outperforms existing KPE methods on a large open domain keyphrase extraction benchmark, OpenKP, and a scientific domain dataset, KP20K.

Tianjun Ke, Haoqun Cao, Zenan Ling et al.

Meta-learning has demonstrated promising results in few-shot classification (FSC) by learning to solve new problems using prior knowledge. Bayesian methods are effective at characterizing uncertainty in FSC, which is crucial in high-risk fields. In this context, the logistic-softmax likelihood is often employed as an alternative to the softmax likelihood in multi-class Gaussian process classification due to its conditional conjugacy property. However, the theoretical property of logistic-softmax is not clear and previous research indicated that the inherent uncertainty of logistic-softmax leads to suboptimal performance. To mitigate these issues, we revisit and redesign the logistic-softmax likelihood, which enables control of the \textit{a priori} confidence level through a temperature parameter. Furthermore, we theoretically and empirically show that softmax can be viewed as a special case of logistic-softmax and logistic-softmax induces a larger family of data distribution than softmax. Utilizing modified logistic-softmax, we integrate the data augmentation technique into the deep kernel based Gaussian process meta-learning framework, and derive an analytical mean-field approximation for task-specific updates. Our approach yields well-calibrated uncertainty estimates and achieves comparable or superior results on standard benchmark datasets. Code is publicly available at \url{https://github.com/keanson/revisit-logistic-softmax}.

Haojing Huang, Jingheng Ye, Qingyu Zhou et al.

In recent years, Chinese Spelling Check (CSC) has been greatly improved by designing task-specific pre-training methods or introducing auxiliary tasks, which mostly solve this task in an end-to-end fashion. In this paper, we propose to decompose the CSC workflow into detection, reasoning, and searching subtasks so that the rich external knowledge about the Chinese language can be leveraged more directly and efficiently. Specifically, we design a plug-and-play detection-and-reasoning module that is compatible with existing SOTA non-autoregressive CSC models to further boost their performance. We find that the detection-and-reasoning module trained for one model can also benefit other models. We also study the primary interpretability provided by the task decomposition. Extensive experiments and detailed analyses demonstrate the effectiveness and competitiveness of the proposed module.

Feng Zhou, Quyu Kong, Zhijie Deng et al.

This paper presents a novel extension of multi-task Gaussian Cox processes for modeling multiple heterogeneous correlated tasks jointly, e.g., classification and regression, via multi-output Gaussian processes (MOGP). A MOGP prior over the parameters of the dedicated likelihoods for classification, regression and point process tasks can facilitate sharing of information between heterogeneous tasks, while allowing for nonparametric parameter estimation. To circumvent the non-conjugate Bayesian inference in the MOGP modulated heterogeneous multi-task framework, we employ the data augmentation technique and derive a mean-field approximation to realize closed-form iterative updates for estimating model parameters. We demonstrate the performance and inference on both 1D synthetic data as well as 2D urban data of Vancouver.

Han Meng, Xiaosong He, Zexing Chen et al.

Some Natural Language Generation (NLG) tasks require both faithfulness and diversity. The decoding strategy is intensively related to the quality of the generated text. Strategies such as beam search, greedy search, etc., perform with low diversity and high repetition. On the other hand, guided decoding, the solution towards diversity, may generate unfaithful expressions. To this end, this paper presents Information Filter upon Diversity-Improved Decoding (IFDID) to obtain the tradeoff between diversity and faithfulness. IFDID is a two-stage decoding strategy leveraging the proposed Enhance-Filter framework, which achieves the tradeoff by increasing the probabilities of some typical tokens being selected and subsequently filtering them by their information amount. To verify the effectiveness, we compare our method with other baselines on related CommonGEN, RocStories and AdGen benchmarks, which cover Chinese and English datasets. Our numerical experimental results and human evaluation outcomes verify the effectiveness of the proposed approach, as our approach achieves a 1.24 higher ROUGE score describing faithfulness as well as higher diversity represented by 62.5% higher upon Dist-2 than traditional approaches, demonstrating that IFDID is a novel SOTA decoding strategy for the tradeoff between diversity and faithfulness.

Feng Zhou, Jian Zhang, Yuhang Sun et al.

3D Gaussian Splatting (3DGS) has shown great potential in autonomous driving simulation and data generation, enabling photorealistic reconstruction and flexible scene manipulation. However, existing 3DGS scene editing methods have limited support for road geometry editing (e.g., inserting speed humps or sunken roads), and generally do not couple such edits with plausible vehicle-road interaction dynamics. Such editing is essential for generating training data under extreme driving scenarios or evaluating system reliability under these road irregularities. Moreover, many optimization-based methods require minutes of per-edit refinement, while existing efficient alternatives mainly focus on appearance-level or object-level manipulation rather than physics-aware road irregularity editing. To address these limitations, we propose RoVES, a Road-and-Vehicle Editing System for physics-aware 3D Gaussian editing in driving scenes. RoVES enables single-image-driven road geometry insertion and couples the edited road profile with a 4-DOF half-car vehicle dynamics model to achieve physics-aware vehicle pose correction in vertical displacement and pitch. RoVES inserts road elements in a one-shot, optimization-free pipeline (1.84s), and the full pipeline (including color transfer and vehicle-dynamics-based pose correction) completes in 6.24s; it edits dynamic vehicles via pose editing and corrects poses frame-by-frame to approximate dynamics-consistent vertical displacement and pitch responses. Experiments on the Waymo dataset show that RoVES provides practical efficiency and competitive visual consistency for physics-aware driving scene generation.

Feng Zhou, Antonio Cicone, Haomin Zhou

The decomposition of non-stationary signals is an important and challenging task in the field of signal time-frequency analysis. In the recent two decades, many signal decomposition methods led by the empirical mode decomposition, which was pioneered by Huang et al. in 1998, have been proposed by different research groups. However, they still have some limitations. For example, they are generally prone to boundary and mode mixing effects and are not very robust to noise. Inspired by the successful applications of deep learning in fields like image processing and natural language processing, and given the lack in the literature of works in which deep learning techniques are used directly to decompose non-stationary signals into simple oscillatory components, we use the convolutional neural network, residual structure and nonlinear activation function to compute in an innovative way the local average of the signal, and study a new non-stationary signal decomposition method under the framework of deep learning. We discuss the training process of the proposed model and study the convergence analysis of the learning algorithm. In the experiments, we evaluate the performance of the proposed model from two points of view: the calculation of the local average and the signal decomposition. Furthermore, we study the mode mixing, noise interference, and orthogonality properties of the decomposed components produced by the proposed method. All results show that the proposed model allows for better handling boundary effect, mode mixing effect, robustness, and the orthogonality of the decomposed components than existing methods.

Jucheng Hu, Zhangquan Chen, Yulin Chen et al.

Deciphering animal intent is a fundamental challenge in computational ethology, largely because of semantic aliasing, the phenomenon where identical external signals (e.g., a cat's purr) correspond to radically different internal states depending on physiological context. Existing Multimodal Large Language Models (MLLMs) are blind to high-frequency biological time-series data, restricting them to superficial behavioural pattern matching rather than genuine latent-state reasoning. To bridge this gap, we introduce Meow-Omni 1, the first open-source, quad-modal MLLM purpose-built for computational ethology. It natively fuses video, audio, and physiological time-series streams with textual reasoning. Through targeted architectural adaptation, we integrate specialized scientific encoders into a unified backbone and formalize intent inference via physiologically grounded cross-modal alignment. Evaluated on MeowBench, a novel, expert-verified quad-modal benchmark, Meow-Omni 1 achieves state-of-the-art intent-recognition accuracy (71.16%), substantially outperforming leading vision-language and omni-modal baselines. We release the complete open-source pipeline including model weights, training framework, and the Meow-10K dataset, to establish a scalable paradigm for inter-species intent understanding and to advance foundation models toward real-world veterinary diagnostics and wildlife conservation.

Zizhuo Meng, Boyu Li, Xuhui Fan et al.

The classical temporal point process (TPP) constructs an intensity function by taking the occurrence times into account. Nevertheless, occurrence time may not be the only relevant factor, other contextual data, termed covariates, may also impact the event evolution. Incorporating such covariates into the model is beneficial, while distinguishing their relevance to the event dynamics is of great practical significance. In this work, we propose a Transformer-based covariate temporal point process (TransFeat-TPP) model to improve the interpretability of deep covariate-TPPs while maintaining powerful expressiveness. TransFeat-TPP can effectively model complex relationships between events and covariates, and provide enhanced interpretability by discerning the importance of various covariates. Experimental results on synthetic and real datasets demonstrate improved prediction accuracy and consistently interpretable feature importance when compared to existing deep covariate-TPPs.

Pu Cao, Feng Zhou, Qing Song et al.

In the rapidly advancing realm of visual generation, diffusion models have revolutionized the landscape, marking a significant shift in capabilities with their impressive text-guided generative functions. However, relying solely on text for conditioning these models does not fully cater to the varied and complex requirements of different applications and scenarios. Acknowledging this shortfall, a variety of studies aim to control pre-trained text-to-image (T2I) models to support novel conditions. In this survey, we undertake a thorough review of the literature on controllable generation with T2I diffusion models, covering both the theoretical foundations and practical advancements in this domain. Our review begins with a brief introduction to the basics of denoising diffusion probabilistic models (DDPMs) and widely used T2I diffusion models. We then reveal the controlling mechanisms of diffusion models, theoretically analyzing how novel conditions are introduced into the denoising process for conditional generation. Additionally, we offer a detailed overview of research in this area, organizing it into distinct categories from the condition perspective: generation with specific conditions, generation with multiple conditions, and universal controllable generation. For an exhaustive list of the controllable generation literature surveyed, please refer to our curated repository at \url{https://github.com/PRIV-Creation/Awesome-Controllable-T2I-Diffusion-Models}.

Zeyue Zhang, Xiaoling Lu, Feng Zhou

The Bayesian two-step change point detection method is popular for the Hawkes process due to its simplicity and intuitiveness. However, the non-conjugacy between the point process likelihood and the prior requires most existing Bayesian two-step change point detection methods to rely on non-conjugate inference methods. These methods lack analytical expressions, leading to low computational efficiency and impeding timely change point detection. To address this issue, this work employs data augmentation to propose a conjugate Bayesian two-step change point detection method for the Hawkes process, which proves to be more accurate and efficient. Extensive experiments on both synthetic and real data demonstrate the superior effectiveness and efficiency of our method compared to baseline methods. Additionally, we conduct ablation studies to explore the robustness of our method concerning various hyperparameters. Our code is publicly available at https://github.com/Aurora2050/CoBay-CPD.

Feng Zhou, Yanjie Zhou, Longjie Wang et al.

Traditional one-shot medical image segmentation (MIS) methods use registration networks to propagate labels from a reference atlas or rely on comprehensive sampling strategies to generate synthetic labeled data for training. However, these methods often struggle with registration errors and low-quality synthetic images, leading to poor performance and generalization. To overcome this, we introduce a novel one-shot MIS framework based on knowledge distillation, which allows the network to directly 'see' real images through a distillation process guided by image reconstruction. It focuses on anatomical structures in a single labeled image and a few unlabeled ones. A registration-based data augmentation network creates realistic, labeled samples, while a feature distillation module helps the student network learn segmentation from these samples, guided by the teacher network. During inference, the streamlined student network accurately segments new images. Evaluations on three public datasets (OASIS for T1 brain MRI, BCV for abdomen CT, and VerSe for vertebrae CT) show superior segmentation performance and generalization across different medical image datasets and modalities compared to leading methods. Our code is available at https://github.com/NoviceFodder/OS-MedSeg.

Yixuan Zhang, Quyu Kong, Feng Zhou

In this study, we propose a novel deep spatio-temporal point process model, Deep Kernel Mixture Point Processes (DKMPP), that incorporates multimodal covariate information. DKMPP is an enhanced version of Deep Mixture Point Processes (DMPP), which uses a more flexible deep kernel to model complex relationships between events and covariate data, improving the model's expressiveness. To address the intractable training procedure of DKMPP due to the non-integrable deep kernel, we utilize an integration-free method based on score matching, and further improve efficiency by adopting a scalable denoising score matching method. Our experiments demonstrate that DKMPP and its corresponding score-based estimators outperform baseline models, showcasing the advantages of incorporating covariate information, utilizing a deep kernel, and employing score-based estimators.

Feng Zhou, Antonio Cicone, Haomin Zhou

Time-frequency analysis is an important and challenging task in many applications. Fourier and wavelet analysis are two classic methods that have achieved remarkable success in many fields. However, they also exhibit limitations when applied to nonlinear and non-stationary signals. To address this challenge, a series of nonlinear and adaptive methods, pioneered by the empirical mode decomposition method, have been proposed. The goal of these methods is to decompose a non-stationary signal into quasi-stationary components that enhance the clarity of features during time-frequency analysis. Recently, inspired by deep learning, we proposed a novel method called iterative residual convolutional neural network (IRCNN). IRCNN not only achieves more stable decomposition than existing methods but also handles batch processing of large-scale signals with low computational cost. Moreover, deep learning provides a unique perspective for non-stationary signal decomposition. In this study, we aim to further improve IRCNN with the help of several nimble techniques from deep learning and optimization to ameliorate the method and overcome some of the limitations of this technique.

Jiajun Fu, Yonghao Dang, Ruoqi Yin et al.

This technical report describes our first-place solution to the pose estimation challenge at ECCV 2022 Visual Perception for Navigation in Human Environments Workshop. In this challenge, we aim to estimate human poses from in-the-wild stitched panoramic images. Our method is built based on Faster R-CNN for human detection, and HRNet for human pose estimation. We describe technical details for the JRDB-Pose dataset, together with some experimental results. In the competition, we achieved 0.303 $\text{OSPA}_{\text{IOU}}$ and 64.047\% $\text{AP}_{\text{0.5}}$ on the test set of JRDB-Pose.

Yixuan Zhang, Feng Zhou, Zhidong Li et al.

Removing bias while keeping all task-relevant information is challenging for fair representation learning methods since they would yield random or degenerate representations w.r.t. labels when the sensitive attributes correlate with labels. Existing works proposed to inject the label information into the learning procedure to overcome such issues. However, the assumption that the observed labels are clean is not always met. In fact, label bias is acknowledged as the primary source inducing discrimination. In other words, the fair pre-processing methods ignore the discrimination encoded in the labels either during the learning procedure or the evaluation stage. This contradiction puts a question mark on the fairness of the learned representations. To circumvent this issue, we explore the following question: \emph{Can we learn fair representations predictable to latent ideal fair labels given only access to unreliable labels?} In this work, we propose a \textbf{D}e-\textbf{B}iased \textbf{R}epresentation Learning for \textbf{F}airness (DBRF) framework which disentangles the sensitive information from non-sensitive attributes whilst keeping the learned representations predictable to ideal fair labels rather than observed biased ones. We formulate the de-biased learning framework through information-theoretic concepts such as mutual information and information bottleneck. The core concept is that DBRF advocates not to use unreliable labels for supervision when sensitive information benefits the prediction of unreliable labels. Experiment results over both synthetic and real-world data demonstrate that DBRF effectively learns de-biased representations towards ideal labels.

Changming Xiao, Qi Yang, Feng Zhou et al.

Diffusion models have revolted the field of text-to-image generation recently. The unique way of fusing text and image information contributes to their remarkable capability of generating highly text-related images. From another perspective, these generative models imply clues about the precise correlation between words and pixels. In this work, a simple but effective method is proposed to utilize the attention mechanism in the denoising network of text-to-image diffusion models. Without re-training nor inference-time optimization, the semantic grounding of phrases can be attained directly. We evaluate our method on Pascal VOC 2012 and Microsoft COCO 2014 under weakly-supervised semantic segmentation setting and our method achieves superior performance to prior methods. In addition, the acquired word-pixel correlation is found to be generalizable for the learned text embedding of customized generation methods, requiring only a few modifications. To validate our discovery, we introduce a new practical task called "personalized referring image segmentation" with a new dataset. Experiments in various situations demonstrate the advantages of our method compared to strong baselines on this task. In summary, our work reveals a novel way to extract the rich multi-modal knowledge hidden in diffusion models for segmentation.

Lei Li, Sen Jia, Jianhao Wang et al.

This paper presents LLaMo (Large Language and Human Motion Assistant), a multimodal framework for human motion instruction tuning. In contrast to conventional instruction-tuning approaches that convert non-linguistic inputs, such as video or motion sequences, into language tokens, LLaMo retains motion in its native form for instruction tuning. This method preserves motion-specific details that are often diminished in tokenization, thereby improving the model's ability to interpret complex human behaviors. By processing both video and motion data alongside textual inputs, LLaMo enables a flexible, human-centric analysis. Experimental evaluations across high-complexity domains, including human behaviors and professional activities, indicate that LLaMo effectively captures domain-specific knowledge, enhancing comprehension and prediction in motion-intensive scenarios. We hope LLaMo offers a foundation for future multimodal AI systems with broad applications, from sports analytics to behavioral prediction. Our code and models are available on the project website: https://github.com/ILGLJ/LLaMo.

Qingyun Jin, Xiaohui Song, Feng Zhou et al.

Large language models (LLMs) have demonstrated exceptional performance across diverse natural language processing tasks. However, as the model size and the input sequence's length increase, the linearly increasing key-value (KV) cache significantly degrades inference throughput. Therefore, grouped-query attention (GQA), as an alternative to multi-head attention (MHA), has been widely introduced into LLMs. In this work, we propose a cost-effective method for converting MHA into GQA with any compression ratio of KV heads. The key point of our method lies in the application of Procrustes analysis to the attention heads, which enhances the similarity among attention heads while preserving computational invariance, thereby improving the model's post-training performance. Subsequently, we employ $\mathit{L_0}$ regularization to prune redundant parameters. The model after pruning can be adapted to the standard GQA framework. Experimental results show that our strategy can compress up to 87.5\% KV heads of LLaMA2-7B model and 75\% KV heads of Sheared-LLaMA-1.3B with acceptable performance degradation. Our code is released at https://github.com/fpcsong/mha2gqa.

Tengxuan Liu, Shiyao Li, Jiayi Yang et al. · tsinghua

Recently, significant progress has been made in developing reasoning-capable Large Language Models (LLMs) through long Chain-of-Thought (CoT) techniques. However, this long-CoT reasoning process imposes substantial memory overhead due to the large Key-Value (KV) Cache memory overhead. Post-training KV Cache quantization has emerged as a promising compression technique and has been extensively studied in short-context scenarios. However, directly applying existing methods to long-CoT LLMs causes significant performance degradation due to the following two reasons: (1) Large cumulative error: Existing methods fail to adequately leverage available memory, and they directly quantize the KV Cache during each decoding step, leading to large cumulative quantization error. (2) Short-context calibration: Due to Rotary Positional Embedding (RoPE), the use of short-context data during calibration fails to account for the distribution of less frequent channels in the Key Cache, resulting in performance loss. We propose Progressive Mixed-Precision KV Cache Quantization (PM-KVQ) for long-CoT LLMs to address the above issues in two folds: (1) To reduce cumulative error, we design a progressive quantization strategy to gradually lower the bit-width of KV Cache in each block. Then, we propose block-wise memory allocation to assign a higher bit-width to more sensitive transformer blocks. (2) To increase the calibration length without additional overhead, we propose a new calibration strategy with positional interpolation that leverages short calibration data with positional interpolation to approximate the data distribution of long-context data. Extensive experiments on 7B-70B long-CoT LLMs show that PM-KVQ improves reasoning benchmark performance by up to 8% over SOTA baselines under the same memory budget. Our code is available at https://github.com/thu-nics/PM-KVQ.

Guoliang Xu, Jianqin Yin, Feng Zhou et al.

Previous methods usually only extract the image modality's information to recognize group activity. However, mining image information is approaching saturation, making it difficult to extract richer information. Therefore, extracting complementary information from other modalities to supplement image information has become increasingly important. In fact, action labels provide clear text information to express the action's semantics, which existing methods often overlook. Thus, we propose ActivityCLIP, a plug-and-play method for mining the text information contained in the action labels to supplement the image information for enhancing group activity recognition. ActivityCLIP consists of text and image branches, where the text branch is plugged into the image branch (The off-the-shelf image-based method). The text branch includes Image2Text and relation modeling modules. Specifically, we propose the knowledge transfer module, Image2Text, which adapts image information into text information extracted by CLIP via knowledge distillation. Further, to keep our method convenient, we add fewer trainable parameters based on the relation module of the image branch to model interaction relation in the text branch. To show our method's generality, we replicate three representative methods by ActivityCLIP, which adds only limited trainable parameters, achieving favorable performance improvements for each method. We also conduct extensive ablation studies and compare our method with state-of-the-art methods to demonstrate the effectiveness of ActivityCLIP.

Lizhi Wang, Feng Zhou, Bo yu et al.

Recent advancements in 3D reconstruction technologies have paved the way for high-quality and real-time rendering of complex 3D scenes. Despite these achievements, a notable challenge persists: it is difficult to precisely reconstruct specific objects from large scenes. Current scene reconstruction techniques frequently result in the loss of object detail textures and are unable to reconstruct object portions that are occluded or unseen in views. To address this challenge, we delve into the meticulous 3D reconstruction of specific objects within large scenes and propose a framework termed OMEGAS: Object Mesh Extraction from Large Scenes Guided by Gaussian Segmentation. Specifically, we proposed a novel 3D target segmentation technique based on 2D Gaussian Splatting, which segments 3D consistent target masks in multi-view scene images and generates a preliminary target model. Moreover, to reconstruct the unseen portions of the target, we propose a novel target replenishment technique driven by large-scale generative diffusion priors. We demonstrate that our method can accurately reconstruct specific targets from large scenes, both quantitatively and qualitatively. Our experiments show that OMEGAS significantly outperforms existing reconstruction methods across various scenarios. Our project page is at: https://github.com/CrystalWlz/OMEGAS

Pan Yin, Kaiyu Li, Xiangyong Cao et al.

Recently, road graph extraction has garnered increasing attention due to its crucial role in autonomous driving, navigation, etc. However, accurately and efficiently extracting road graphs remains a persistent challenge, primarily due to the severe scarcity of labeled data. To address this limitation, we collect a global-scale satellite road graph extraction dataset, i.e. Global-Scale dataset. Specifically, the Global-Scale dataset is $\sim20 \times$ larger than the largest existing public road extraction dataset and spans over 13,800 $km^2$ globally. Additionally, we develop a novel road graph extraction model, i.e. SAM-Road++, which adopts a node-guided resampling method to alleviate the mismatch issue between training and inference in SAM-Road, a pioneering state-of-the-art road graph extraction model. Furthermore, we propose a simple yet effective ``extended-line'' strategy in SAM-Road++ to mitigate the occlusion issue on the road. Extensive experiments demonstrate the validity of the collected Global-Scale dataset and the proposed SAM-Road++ method, particularly highlighting its superior predictive power in unseen regions. The dataset and code are available at \url{https://github.com/earth-insights/samroadplus}.

Lilit Avetisyan, Chengxin Zhang, Sue Bai et al.

Micro-mobility is promising to contribute to sustainable cities in the future with its efficiency and low cost. To better design such a sustainable future, it is necessary to understand the trends and challenges. Thus, we examined people's opinions on micro-mobility in the US and the EU using Tweets. We used topic modeling based on advanced natural language processing techniques and categorized the data into seven topics: promotion and service, mobility, technical features, acceptance, recreation, infrastructure and regulations. Furthermore, using sentiment analysis, we investigated people's positive and negative attitudes towards specific aspects of these topics and compared the patterns of the trends and challenges in the US and the EU. We found that 1) promotion and service included the majority of Twitter discussions in the both regions, 2) the EU had more positive opinions than the US, 3) micro-mobility devices were more widely used for utilitarian mobility and recreational purposes in the EU than in the US, and 4) compared to the EU, people in the US had many more concerns related to infrastructure and regulation issues. These findings help us understand the trends and challenges and prioritize different aspects in micro-mobility to improve their safety and experience across the two areas for designing a more sustainable micro-mobility future.

Junliang Lyu, Yixuan Zhang, Xiaoling Lu et al.

This work addresses a key limitation in current federated learning approaches, which predominantly focus on homogeneous tasks, neglecting the task diversity on local devices. We propose a principled integration of multi-task learning using multi-output Gaussian processes (MOGP) at the local level and federated learning at the global level. MOGP handles correlated classification and regression tasks, offering a Bayesian non-parametric approach that naturally quantifies uncertainty. The central server aggregates the posteriors from local devices, updating a global MOGP prior redistributed for training local models until convergence. Challenges in performing posterior inference on local devices are addressed through the Pólya-Gamma augmentation technique and mean-field variational inference, enhancing computational efficiency and convergence rate. Experimental results on both synthetic and real data demonstrate superior predictive performance, OOD detection, uncertainty calibration and convergence rate, highlighting the method's potential in diverse applications. Our code is publicly available at https://github.com/JunliangLv/task_diversity_BFL.

Junchao Lin, Zenan Ling, Zhanbo Feng et al.

Implicit graph neural networks (IGNNs), which exhibit strong expressive power with a single layer, have recently demonstrated remarkable performance in capturing long-range dependencies (LRD) in underlying graphs while effectively mitigating the over-smoothing problem. However, IGNNs rely on computationally expensive fixed-point iterations, which lead to significant speed and scalability limitations, hindering their application to large-scale graphs. To achieve fast fixed-point solving for IGNNs, we propose a novel graph neural solver, IGNN-Solver, which leverages the generalized Anderson Acceleration method, parameterized by a tiny GNN, and learns iterative updates as a graph-dependent temporal process. To improve effectiveness on large-scale graph tasks, we further integrate sparsification and storage compression methods, specifically tailored for the IGNN-Solver, into its design. Extensive experiments demonstrate that the IGNN-Solver significantly accelerates inference on both small- and large-scale tasks, achieving a $1.5\times$ to $8\times$ speedup without sacrificing accuracy. This advantage becomes more pronounced as the graph scale grows, facilitating its large-scale deployment in real-world applications. The code to reproduce our results is available at https://github.com/landrarwolf/IGNN-Solver.

Yixuan Zhang, Jiabin Luo, Zhenggang Wang et al.

Fairness concerns are increasingly critical as machine learning models are deployed in high-stakes applications. While existing fairness-aware methods typically intervene at the model level, they often suffer from high computational costs, limited scalability, and poor generalization. To address these challenges, we propose a Bayesian data selection framework that ensures fairness by aligning group-specific posterior distributions of model parameters and sample weights with a shared central distribution. Our framework supports flexible alignment via various distributional discrepancy measures, including Wasserstein distance, maximum mean discrepancy, and $f$-divergence, allowing geometry-aware control without imposing explicit fairness constraints. This data-centric approach mitigates group-specific biases in training data and improves fairness in downstream tasks, with theoretical guarantees. Experiments on benchmark datasets show that our method consistently outperforms existing data selection and model-based fairness methods in both fairness and accuracy.

Linyan Gu, Lihua Yang, Feng Zhou

Transformer networks have achieved remarkable empirical success across a wide range of applications, yet their theoretical expressive power remains insufficiently understood. In this paper, we study the expressive capabilities of Transformer architectures. We first establish an explicit approximation of maxout networks by Transformer networks while preserving comparable model complexity. As a consequence, Transformers inherit the universal approximation capability of ReLU networks under similar complexity constraints. Building on this connection, we develop a framework to analyze the approximation of continuous piecewise linear functions by Transformers and quantitatively characterize their expressivity via the number of linear regions, which grows exponentially with depth. Our analysis establishes a theoretical bridge between approximation theory for standard feedforward neural networks and Transformer architectures. It also yields structural insights into Transformers: self-attention layers implement max-type operations, while feedforward layers realize token-wise affine transformations.

Yiyang Ma, Feng Zhou, Xuedan Yin et al.

Leveraging pre-trained Diffusion Transformers (DiTs) for high-resolution (HR) image synthesis often leads to spatial layout collapse and degraded texture fidelity. Prior work mitigates these issues with complex pipelines that first perform a base-resolution (i.e., training-resolution) denoising process to guide HR generation. We instead explore the intrinsic generative mechanisms of DiTs and propose ResDiT, a training-free method that scales resolution efficiently. We identify the core factor governing spatial layout, position embeddings (PEs), and show that the original PEs encode incorrect positional information when extrapolated to HR, which triggers layout collapse. To address this, we introduce a PE scaling technique that rectifies positional encoding under resolution changes. To further remedy low-fidelity details, we develop a local-enhancement mechanism grounded in base-resolution local attention. We design a patch-level fusion module that aggregates global and local cues, together with a Gaussian-weighted splicing strategy that eliminates grid artifacts. Comprehensive evaluations demonstrate that ResDiT consistently delivers high-fidelity, high-resolution image synthesis and integrates seamlessly with downstream tasks, including spatially controlled generation.

Haoqun Cao, Yixuan Zhang, Feng Zhou

Score matching estimators have garnered significant attention in recent years because they eliminate the need to compute normalizing constants, thereby mitigating the computational challenges associated with maximum likelihood estimation (MLE).While several studies have proposed score matching estimators for point processes, this work highlights the limitations of these existing methods, which stem primarily from the lack of a mathematically rigorous analysis of how score matching behaves on finite point processes -- special random configurations on bounded spaces where many of the usual assumptions and properties of score matching no longer hold. To this end, we develop a formal framework for score matching on finite point processes via Janossy measures and, within this framework, introduce an (autoregressive) weighted score-matching estimator, whose statistical properties we analyze in classical parametric settings. For general nonparametric (e.g., deep) point process models, we show that score matching alone does not uniquely identify the ground-truth distribution due to subtle normalization issues, and we propose a simple survival-classification augmentation that yields a complete, integration-free training objective for any intensity-based point process model for spatio-temporal case. Experiments on synthetic and real-world temporal and spatio-temporal datasets, demonstrate that our method accurately recovers intensities and achieves performance comparable to MLE with better efficiency.

Yue Jiang, Jichu Li, Yang Liu et al.

We introduce DanmakuTPPBench, a comprehensive benchmark designed to advance multi-modal Temporal Point Process (TPP) modeling in the era of Large Language Models (LLMs). While TPPs have been widely studied for modeling temporal event sequences, existing datasets are predominantly unimodal, hindering progress in models that require joint reasoning over temporal, textual, and visual information. To address this gap, DanmakuTPPBench comprises two complementary components: (1) DanmakuTPP-Events, a novel dataset derived from the Bilibili video platform, where user-generated bullet comments (Danmaku) naturally form multi-modal events annotated with precise timestamps, rich textual content, and corresponding video frames; (2) DanmakuTPP-QA, a challenging question-answering dataset constructed via a novel multi-agent pipeline powered by state-of-the-art LLMs and multi-modal LLMs (MLLMs), targeting complex temporal-textual-visual reasoning. We conduct extensive evaluations using both classical TPP models and recent MLLMs, revealing significant performance gaps and limitations in current methods' ability to model multi-modal event dynamics. Our benchmark establishes strong baselines and calls for further integration of TPP modeling into the multi-modal language modeling landscape. Project page: https://github.com/FRENKIE-CHIANG/DanmakuTPPBench

Tianjun Ke, Haoqun Cao, Feng Zhou

Bayesian few-shot classification has been a focal point in the field of few-shot learning. This paper seamlessly integrates mirror descent-based variational inference into Gaussian process-based few-shot classification, addressing the challenge of non-conjugate inference. By leveraging non-Euclidean geometry, mirror descent achieves accelerated convergence by providing the steepest descent direction along the corresponding manifold. It also exhibits the parameterization invariance property concerning the variational distribution. Experimental results demonstrate competitive classification accuracy, improved uncertainty quantification, and faster convergence compared to baseline models. Additionally, we investigate the impact of hyperparameters and components. Code is publicly available at https://github.com/keanson/MD-BSFC.

Yishuo Wang, Feng Zhou, Qicheng Meng et al.

Existing ocean front detection methods--including histogram-based variance analysis, Lyapunov exponent, gradient thresholding, and machine learning--suffer from critical limitations: discontinuous outputs, over-detection, reliance on single-threshold decisions, and lack of open-source implementations. To address these challenges, this paper proposes the Bayesian Front Detection and Tracking framework with Metric Space Analysis (BFDT-MSA). The framework introduces three innovations: (1) a Bayesian decision mechanism that integrates gradient priors and field operators to eliminate manual threshold sensitivity; (2) morphological refinement algorithms for merging fragmented fronts, deleting spurious rings, and thinning frontal zones to pixel-level accuracy; and (3) a novel metric space definition for temporal front tracking, enabling systematic analysis of front evolution. Validated on global SST data (2022--2024), BFDT-MSA reduces over-detection by $73\%$ compared to histogram-based methods while achieving superior intensity ($0.16^\circ$C/km), continuity, and spatiotemporal coherence. The open-source release bridges a critical gap in reproducible oceanographic research.

Youjin Wang, Jiaqiao Zhao, Rong Fu et al.

Balancing fine-grained local modeling with long-range dependency capture under computational constraints remains a central challenge in sequence modeling. While Transformers provide strong token mixing, they suffer from quadratic complexity, whereas Mamba-style selective state-space models (SSMs) scale linearly but often struggle to capture high-rank and synchronous global interactions. We present a consistency boundary analysis that characterizes when diagonal short-memory SSMs can approximate causal attention and identifies structural gaps that remain. Motivated by this analysis, we propose InfoMamba, an attention-free hybrid architecture. InfoMamba replaces token-level self-attention with a concept bottleneck linear filtering layer that serves as a minimal-bandwidth global interface and integrates it with a selective recurrent stream through information-maximizing fusion (IMF). IMF dynamically injects global context into the SSM dynamics and encourages complementary information usage through a mutual-information-inspired objective. Extensive experiments on classification, dense prediction, and non-vision tasks show that InfoMamba consistently outperforms strong Transformer and SSM baselines, achieving competitive accuracy-efficiency trade-offs while maintaining near-linear scaling.

Jiayu Bai, Danchen Yu, Zhenyu Liao et al.

Kronecker adapters have emerged as a promising approach for fine-tuning large-scale models, enabling high-rank updates through tunable component structures. However, existing work largely treats the component structure as a fixed or heuristic design choice, leaving the dimensions and number of Kronecker components underexplored. In this paper, we identify component structure as a key factor governing the capacity of Kronecker adapters. We perform a fine-grained analysis of both the dimensions and number of Kronecker components. In particular, we show that the alignment between Kronecker adapters and full fine-tuning depends on component configurations. Guided by these insights, we propose Component Designed Kronecker Adapters (CDKA). We further provide parameter-budget-aware configuration guidelines and a tailored training stabilization strategy for practical deployment. Experiments across various natural language processing tasks demonstrate the effectiveness of CDKA. Code is available at https://github.com/rainstonee/CDKA.

Pu Cao, Yiyang Ma, Feng Zhou et al.

In latent diffusion models, the autoencoder (AE) is typically expected to balance two capabilities: faithful reconstruction and a generation-friendly latent space (e.g., low gFID). In recent ImageNet-scale AE studies, we observe a systematic bias toward generative metrics in handling this trade-off: reconstruction metrics are increasingly under-reported, and ablation-based AE selection often favors the best-gFID configuration even when reconstruction fidelity degrades. We theoretically analyze why this gFID-dominant preference can appear unproblematic for ImageNet generation, yet becomes risky when scaling to controllable diffusion: AEs can induce condition drift, which limits achievable condition alignment. Meanwhile, we find that reconstruction fidelity, especially instance-level measures, better indicates controllability. We empirically validate the impact of tilted autoencoder evaluation on controllability by studying several recent ImageNet AEs. Using a multi-dimensional condition-drift evaluation protocol reflecting controllable generation tasks, we find that gFID is only weakly predictive of condition preservation, whereas reconstruction-oriented metrics are substantially more aligned. ControlNet experiments further confirm that controllability tracks condition preservation rather than gFID. Overall, our results expose a gap between ImageNet-centric AE evaluation and the requirements of scalable controllable diffusion, offering practical guidance for more reliable benchmarking and model selection.

Feng Zhou, Wenkai Guo, Pu Cao et al.

Sparse-view 3D Gaussian Splatting (3DGS) often overfits to the training views, leading to artifacts like blurring in novel view rendering. Prior work addresses it either by enhancing the initialization (\emph{i.e.}, the point cloud from Structure-from-Motion (SfM)) or by adding training-time constraints (regularization) to the 3DGS optimization. Yet our controlled ablations reveal that initialization is the decisive factor: it determines the attainable performance band in sparse-view 3DGS, while training-time constraints yield only modest within-band improvements at extra cost. Given initialization's primacy, we focus our design there. Although SfM performs poorly under sparse views due to its reliance on feature matching, it still provides reliable seed points. Thus, building on SfM, our effort aims to supplement the regions it fails to cover as comprehensively as possible. Specifically, we design: (i) frequency-aware SfM that improves low-texture coverage via low-frequency view augmentation and relaxed multi-view correspondences; (ii) 3DGS self-initialization that lifts photometric supervision into additional points, compensating SfM-sparse regions with learned Gaussian centers; and (iii) point-cloud regularization that enforces multi-view consistency and uniform spatial coverage through simple geometric/visibility priors, yielding a clean and reliable point cloud. Our experiments on LLFF and Mip-NeRF360 demonstrate consistent gains in sparse-view settings, establishing our approach as a stronger initialization strategy. Code is available at https://github.com/zss171999645/ItG-GS.

Hao Li, Ju Dai, Feng Zhou et al.

While 3D facial animation has made impressive progress, challenges still exist in realizing fine-grained stylized 3D facial expression manipulation due to the lack of appropriate datasets. In this paper, we introduce the AUBlendSet, a 3D facial dataset based on AU-Blendshape representation for fine-grained facial expression manipulation across identities. AUBlendSet is a blendshape data collection based on 32 standard facial action units (AUs) across 500 identities, along with an additional set of facial postures annotated with detailed AUs. Based on AUBlendSet, we propose AUBlendNet to learn AU-Blendshape basis vectors for different character styles. AUBlendNet predicts, in parallel, the AU-Blendshape basis vectors of the corresponding style for a given identity mesh, thereby achieving stylized 3D emotional facial manipulation. We comprehensively validate the effectiveness of AUBlendSet and AUBlendNet through tasks such as stylized facial expression manipulation, speech-driven emotional facial animation, and emotion recognition data augmentation. Through a series of qualitative and quantitative experiments, we demonstrate the potential and importance of AUBlendSet and AUBlendNet in 3D facial animation tasks. To the best of our knowledge, AUBlendSet is the first dataset, and AUBlendNet is the first network for continuous 3D facial expression manipulation for any identity through facial AUs. Our source code is available at https://github.com/wslh852/AUBlendNet.git.

Hao Li, Ju Dai, Xin Zhao et al.

In 3D speech-driven facial animation generation, existing methods commonly employ pre-trained self-supervised audio models as encoders. However, due to the prevalence of phonetically similar syllables with distinct lip shapes in language, these near-homophone syllables tend to exhibit significant coupling in self-supervised audio feature spaces, leading to the averaging effect in subsequent lip motion generation. To address this issue, this paper proposes a plug-and-play semantic decorrelation module-Wav2Sem. This module extracts semantic features corresponding to the entire audio sequence, leveraging the added semantic information to decorrelate audio encodings within the feature space, thereby achieving more expressive audio features. Extensive experiments across multiple Speech-driven models indicate that the Wav2Sem module effectively decouples audio features, significantly alleviating the averaging effect of phonetically similar syllables in lip shape generation, thereby enhancing the precision and naturalness of facial animations. Our source code is available at https://github.com/wslh852/Wav2Sem.git.

Qiang Meng, Chixiang Zhang, Xiaoqiang Xu et al.

Achieving backward compatibility when rolling out new models can highly reduce costs or even bypass feature re-encoding of existing gallery images for in-production visual retrieval systems. Previous related works usually leverage losses used in knowledge distillation which can cause performance degradations or not guarantee compatibility. To address these issues, we propose a general framework called Learning Compatible Embeddings (LCE) which is applicable for both cross model compatibility and compatible training in direct/forward/backward manners. Our compatibility is achieved by aligning class centers between models directly or via a transformation, and restricting more compact intra-class distributions for the new model. Experiments are conducted in extensive scenarios such as changes of training dataset, loss functions, network architectures as well as feature dimensions, and demonstrate that LCE efficiently enables model compatibility with marginal sacrifices of accuracies. The code will be available at https://github.com/IrvingMeng/LCE.

Feng Zhou, X. Jessie Yang, Joost de Winter

Situation awareness (SA) is critical to improving takeover performance during the transition period from automated driving to manual driving. Although many studies measured SA during or after the driving task, few studies have attempted to predict SA in real time in automated driving. In this work, we propose to predict SA during the takeover transition period in conditionally automated driving using eye-tracking and self-reported data. First, a tree ensemble machine learning model, named LightGBM (Light Gradient Boosting Machine), was used to predict SA. Second, in order to understand what factors influenced SA and how, SHAP (SHapley Additive exPlanations) values of individual predictor variables in the LightGBM model were calculated. These SHAP values explained the prediction model by identifying the most important factors and their effects on SA, which further improved the model performance of LightGBM through feature selection. We standardized SA between 0 and 1 by aggregating three performance measures (i.e., placement, distance, and speed estimation of vehicles with regard to the ego-vehicle) of SA in recreating simulated driving scenarios, after 33 participants viewed 32 videos with six lengths between 1 and 20 s. Using only eye-tracking data, our proposed model outperformed other selected machine learning models, having a root-mean-squared error (RMSE) of 0.121, a mean absolute error (MAE) of 0.096, and a 0.719 correlation coefficient between the predicted SA and the ground truth. The code is available at https://github.com/refengchou/Situation-awareness-prediction. Our proposed model provided important implications on how to monitor and predict SA in real time in automated driving using eye-tracking data.