Search

Few-shot classification (FSC) is a fundamental yet challenging task in computer vision that involves recognizing novel classes from limited data. While previous methods have focused on enhancing visual features or incorporating additional modalities, Large Vision Language Models (LVLMs) offer a promising alternative due to their rich knowledge and strong visual perception. However, LVLMs risk learning specific response formats rather than effectively extracting useful information from support data in FSC tasks. In this paper, we investigate LVLMs' performance in FSC and identify key issues such as insufficient learning and the presence of severe positional biases. To tackle the above challenges, we adopt the meta-learning strategy to teach models "learn to learn". By constructing a rich set of meta-tasks for instruction fine-tuning, LVLMs enhance the ability to extract information from few-shot support data for classification. Additionally, we further boost LVLM's few-shot learning capabilities through label augmentation and candidate selection in the fine-tuning and inference stage, respectively. Label augmentation is implemented via a character perturbation strategy to ensure the model focuses on support information. Candidate selection leverages attribute descriptions to filter out unreliable candidates and simplify the task. Extensive experiments demonstrate that our approach achieves superior performance on both general and fine-grained datasets. Furthermore, our candidate selection strategy has been proven beneficial for training-free LVLMs.

Associating driver attention with driving scene across two fields of views (FOVs) is a hard cross-domain perception problem, which requires comprehensive consideration of cross-view mapping, dynamic driving scene analysis, and driver status tracking. Previous methods typically focus on a single view or map attention to the scene via estimated gaze, failing to exploit the implicit connection between them. Moreover, simple fusion modules are insufficient for modeling the complex relationships between the two views, making information integration challenging. To address these issues, we propose a novel method for end-to-end scene-associated driver attention estimation, called EraW-Net. This method enhances the most discriminative dynamic cues, refines feature representations, and facilitates semantically aligned cross-domain integration through a W-shaped architecture, termed W-Net. Specifically, a Dynamic Adaptive Filter Module (DAF-Module) is proposed to address the challenges of frequently changing driving environments by extracting vital regions. It suppresses the indiscriminately recorded dynamics and highlights crucial ones by innovative joint frequency-spatial analysis, enhancing the model's ability to parse complex dynamics. Additionally, to track driver states during non-fixed facial poses, we propose a Global Context Sharing Module (GCS-Module) to construct refined feature representations by capturing hierarchical features that adapt to various scales of head and eye movements. Finally, W-Net achieves systematic cross-view information integration through its "Encoding-Independent Partial Decoding-Fusion Decoding" structure, addressing semantic misalignment in heterogeneous data integration. Experiments demonstrate that the proposed method robustly and accurately estimates the mapping of driver attention in scene on large public datasets., Comment: 13pages, 9 figures

Fine-grained sentiment analysis involves extracting and organizing sentiment elements from textual data. However, existing approaches often overlook issues of category semantic inclusion and overlap, as well as inherent structural patterns within the target sequence. This study introduces a generative sentiment analysis model. To address the challenges related to category semantic inclusion and overlap, a latent category distribution variable is introduced. By reconstructing the input of a variational autoencoder, the model learns the intensity of the relationship between categories and text, thereby improving sequence generation. Additionally, a trie data structure and constrained decoding strategy are utilized to exploit structural patterns, which in turn reduces the search space and regularizes the generation process. Experimental results on the Restaurant-ACOS and Laptop-ACOS datasets demonstrate a significant performance improvement compared to baseline models. Ablation experiments further confirm the effectiveness of latent category distribution and constrained decoding strategy.

Accurate workload forecasting is critical for efficient resource management in cloud computing systems, enabling effective scheduling and autoscaling. Despite recent advances with transformer-based forecasting models, challenges remain due to the non-stationary, nonlinear characteristics of workload time series and the long-term dependencies. In particular, inconsistent performance between long-term history and near-term forecasts hinders long-range predictions. This paper proposes a novel framework leveraging self-supervised multiscale representation learning to capture both long-term and near-term workload patterns. The long-term history is encoded through multiscale representations while the near-term observations are modeled via temporal flow fusion. These representations of different scales are fused using an attention mechanism and characterized with normalizing flows to handle non-Gaussian/non-linear distributions of time series. Extensive experiments on 9 benchmarks demonstrate superiority over existing methods., Comment: Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (CIKM '24), October 21--25, 2024, Boise, ID, USA

Phase engineering strategies in two-dimensional transition metal dichalcogenides (2D-TMDs) have garnered significant attention due to their potential applications in electronics, optoelectronics, and energy storage. Various methods, including direct synthesis, pressure control, and chemical doping, have been employed to manipulate structural transitions in 2D-TMDs. Metal intercalation emerges as an effective technique to modulate phase transition dynamics by inserting external atoms or ions between the layers of 2D-TMDs, altering their electronic structure and physical properties. Here, we investigate the significant structural phase transitions in Pb(Ta1+xSe2)2 single crystals induced by Ta intercalation using a combination of Raman spectroscopy and first-principles calculations. The results highlight the pivotal role of Ta atoms in driving these transitions and elucidate the interplay between intercalation, phase transitions, and resulting electronic and vibrational properties in 2D-TMDs. By focusing on Pb(Ta1+xSe2)2 as an ideal case study and investigating like metal intercalation, this study advances understanding in the field and paves the way for the development of novel applications for 2D-TMDs, offering insights into the potential of these materials for future technological advancements.

Document-level event extraction aims to extract structured event information from unstructured text. However, a single document often contains limited event information and the roles of different event arguments may be biased due to the influence of the information source. This paper addresses the limitations of traditional document-level event extraction by proposing the task of cross-document event extraction (CDEE) to integrate event information from multiple documents and provide a comprehensive perspective on events. We construct a novel cross-document event extraction dataset, namely CLES, which contains 20,059 documents and 37,688 mention-level events, where over 70% of them are cross-document. To build a benchmark, we propose a CDEE pipeline that includes 5 steps, namely event extraction, coreference resolution, entity normalization, role normalization and entity-role resolution. Our CDEE pipeline achieves about 72% F1 in end-to-end cross-document event extraction, suggesting the challenge of this task. Our work builds a new line of information extraction research and will attract new research attention., Comment: ACL2024(Findings)

Existing cross-document event coreference resolution models, which either compute mention similarity directly or enhance mention representation by extracting event arguments (such as location, time, agent, and patient), lacking the ability to utilize document-level information. As a result, they struggle to capture long-distance dependencies. This shortcoming leads to their underwhelming performance in determining coreference for the events where their argument information relies on long-distance dependencies. In light of these limitations, we propose the construction of document-level Rhetorical Structure Theory (RST) trees and cross-document Lexical Chains to model the structural and semantic information of documents. Subsequently, cross-document heterogeneous graphs are constructed and GAT is utilized to learn the representations of events. Finally, a pair scorer calculates the similarity between each pair of events and co-referred events can be recognized using standard clustering algorithm. Additionally, as the existing cross-document event coreference datasets are limited to English, we have developed a large-scale Chinese cross-document event coreference dataset to fill this gap, which comprises 53,066 event mentions and 4,476 clusters. After applying our model on the English and Chinese datasets respectively, it outperforms all baselines by large margins.

A comprehensive guidance view for cardiac interventional surgery can be provided by the real-time fusion of the intraoperative 2D images and preoperative 3D volume based on the ultrasound frame-to-volume registration. However, cardiac ultrasound images are characterized by a low signal-to-noise ratio and small differences between adjacent frames, coupled with significant dimension variations between 2D frames and 3D volumes to be registered, resulting in real-time and accurate cardiac ultrasound frame-to-volume registration being a very challenging task. This paper introduces a lightweight end-to-end Cardiac Ultrasound frame-to-volume Registration network, termed CU-Reg. Specifically, the proposed model leverages epicardium prompt-guided anatomical clues to reinforce the interaction of 2D sparse and 3D dense features, followed by a voxel-wise local-global aggregation of enhanced features, thereby boosting the cross-dimensional matching effectiveness of low-quality ultrasound modalities. We further embed an inter-frame discriminative regularization term within the hybrid supervised learning to increase the distinction between adjacent slices in the same ultrasound volume to ensure registration stability. Experimental results on the reprocessed CAMUS dataset demonstrate that our CU-Reg surpasses existing methods in terms of registration accuracy and efficiency, meeting the guidance requirements of clinical cardiac interventional surgery., Comment: This paper has been accepted by MICCAI 2024

Improving the performance of large language models (LLMs) in complex question-answering (QA) scenarios has always been a research focal point. Recent studies have attempted to enhance LLMs' performance by combining step-wise planning with external retrieval. While effective for advanced models like GPT-3.5, smaller LLMs face challenges in decomposing complex questions, necessitating supervised fine-tuning. Previous work has relied on manual annotation and knowledge distillation from teacher LLMs, which are time-consuming and not accurate enough. In this paper, we introduce a novel framework for enhancing LLMs' planning capabilities by using planning data derived from knowledge graphs (KGs). LLMs fine-tuned with this data have improved planning capabilities, better equipping them to handle complex QA tasks that involve retrieval. Evaluations on multiple datasets, including our newly proposed benchmark, highlight the effectiveness of our framework and the benefits of KG-derived planning data., Comment: Work in progress

Time series forecasts of different temporal granularity are widely used in real-world applications, e.g., sales prediction in days and weeks for making different inventory plans. However, these tasks are usually solved separately without ensuring coherence, which is crucial for aligning downstream decisions. Previous works mainly focus on ensuring coherence with some straightforward methods, e.g., aggregation from the forecasts of fine granularity to the coarse ones, and allocation from the coarse granularity to the fine ones. These methods merely take the temporal hierarchical structure to maintain coherence without improving the forecasting accuracy. In this paper, we propose a novel granularity message-passing mechanism (GMP) that leverages temporal hierarchy information to improve forecasting performance and also utilizes an adaptive reconciliation (AR) strategy to maintain coherence without performance loss. Furthermore, we introduce an optimization module to achieve task-based targets while adhering to more real-world constraints. Experiments on real-world datasets demonstrate that our framework (GMP-AR) achieves superior performances on temporal hierarchical forecasting tasks compared to state-of-the-art methods. In addition, our framework has been successfully applied to a real-world task of payment traffic management in Alipay by integrating with the task-based optimization module.

Fine-tuning large language models (LLMs) can cause them to lose their general capabilities. However, the intrinsic mechanisms behind such forgetting remain unexplored. In this paper, we begin by examining this phenomenon by focusing on knowledge understanding and instruction following, with the latter identified as the main contributor to forgetting during fine-tuning. Consequently, we propose the Instruction Vector (IV) framework to capture model representations highly related to specific instruction-following capabilities, thereby making it possible to understand model-intrinsic forgetting. Through the analysis of IV dynamics pre and post-training, we suggest that fine-tuning mostly adds specialized reasoning patterns instead of erasing previous skills, which may appear as forgetting. Building on this insight, we develop IV-guided training, which aims to preserve original computation graph, thereby mitigating catastrophic forgetting. Empirical tests on three benchmarks confirm the efficacy of this new approach, supporting the relationship between IVs and forgetting. Our code will be made available soon.

In recent years, deep learning-based point cloud normal estimation has made great progress. However, existing methods mainly rely on the PCPNet dataset, leading to overfitting. In addition, the correlation between point clouds with different noise scales remains unexplored, resulting in poor performance in cross-domain scenarios. In this paper, we explore the consistency of intrinsic features learned from clean and noisy point clouds using an Asymmetric Siamese Network architecture. By applying reasonable constraints between features extracted from different branches, we enhance the quality of normal estimation. Moreover, we introduce a novel multi-view normal estimation dataset that includes a larger variety of shapes with different noise levels. Evaluation of existing methods on this new dataset reveals their inability to adapt to different types of shapes, indicating a degree of overfitting. Extensive experiments show that the proposed dataset poses significant challenges for point cloud normal estimation and that our feature constraint mechanism effectively improves upon existing methods and reduces overfitting in current architectures.

The causal dependence in data is often characterized by Directed Acyclic Graphical (DAG) models, widely used in many areas. Causal discovery aims to recover the DAG structure using observational data. This paper focuses on causal discovery with multi-variate count data. We are motivated by real-world web visit data, recording individual user visits to multiple websites. Building a causal diagram can help understand user behavior in transitioning between websites, inspiring operational strategy. A challenge in modeling is user heterogeneity, as users with different backgrounds exhibit varied behaviors. Additionally, social network connections can result in similar behaviors among friends. We introduce personalized Binomial DAG models to address heterogeneity and network dependency between observations, which are common in real-world applications. To learn the proposed DAG model, we develop an algorithm that embeds the network structure into a dimension-reduced covariate, learns each node's neighborhood to reduce the DAG search space, and explores the variance-mean relation to determine the ordering. Simulations show our algorithm outperforms state-of-the-art competitors in heterogeneous data. We demonstrate its practical usefulness on a real-world web visit dataset.

We previously developed a FLASH planning framework for streamlined pin-ridge-filter (pin-RF) design, demonstrating its feasibility for single-energy proton FLASH planning. In this study, we refined the pin-RF design for easy assembly using reusable modules, focusing on its application in liver SABR. This framework generates an intermediate IMPT plan and translates it into step widths and thicknesses of pin-RFs for a single-energy FLASH plan. Parameters like energy spacing, monitor unit limit, and spot quantity were adjusted during IMPT planning, resulting in pin-RFs assembled using predefined modules with widths from 1 to 6 mm, each with a WET of 5 mm. This approach was validated on three liver SABR cases. FLASH doses, quantified using the FLASH effectiveness model at 1 to 5 Gy thresholds, were compared to conventional IMPT (IMPT-CONV) doses to assess clinical benefits. The highest demand for 6 mm width modules, moderate for 2-4 mm, and minimal for 1- and 5-mm modules were shown across all cases. At lower dose thresholds, the two-beam case showed significant dose reductions (>23%), while the other two three-beam cases showed moderate reductions (up to 14.7%), indicating the need for higher fractional beam doses for an enhanced FLASH effect. Positive clinical benefits were seen only in the two-beam case at the 5 Gy threshold. At the 1 Gy threshold, the FLASH plan of the two-beam case outperformed its IMPT-CONV plan, reducing dose indicators by up to 28.3%. However, the three-beam cases showed negative clinical benefits at the 1 Gy threshold, with some dose indicators increasing by up to 16% due to lower fractional beam doses and closer beam arrangements. This study evaluated the feasibility of modularizing streamlined pin-RFs in single-energy proton FLASH planning for liver SABR, offering guidance on optimal module composition and strategies to enhance FLASH planning.

In recent years, large language models (LLMs) have made remarkable achievements in various domains. However, the untimeliness and cost of knowledge updates coupled with hallucination issues of LLMs have curtailed their applications in knowledge intensive tasks, where retrieval augmented generation (RAG) can be of help. Nevertheless, existing retrieval augmented models typically use similarity as a bridge between queries and documents and follow a retrieve then read procedure. In this work, we argue that similarity is not always the panacea and totally relying on similarity would sometimes degrade the performance of retrieval augmented generation. To this end, we propose MetRag, a Multi layEred Thoughts enhanced Retrieval Augmented Generation framework. To begin with, beyond existing similarity oriented thought, we embrace a small scale utility model that draws supervision from an LLM for utility oriented thought and further come up with a smarter model by comprehensively combining the similarity and utility oriented thoughts. Furthermore, given the fact that the retrieved document set tends to be huge and using them in isolation makes it difficult to capture the commonalities and characteristics among them, we propose to make an LLM as a task adaptive summarizer to endow retrieval augmented generation with compactness-oriented thought. Finally, with multi layered thoughts from the precedent stages, an LLM is called for knowledge augmented generation. Extensive experiments on knowledge-intensive tasks have demonstrated the superiority of MetRag., Comment: 12 pages

In recent years, State Space Models (SSMs) with efficient hardware-aware designs, known as the Mamba deep learning models, have made significant progress in modeling long sequences such as language understanding. Therefore, building efficient and general-purpose visual backbones based on SSMs is a promising direction. Compared to traditional convolutional neural networks (CNNs) and Vision Transformers (ViTs), the performance of Vision Mamba (ViM) methods is not yet fully competitive. To enable SSMs to process image data, ViMs typically flatten 2D images into 1D sequences, inevitably ignoring some 2D local dependencies, thereby weakening the model's ability to interpret spatial relationships from a global perspective. We use Fast Fourier Transform (FFT) to obtain the spectrum of the feature map and add it to the original feature map, enabling ViM to model a unified visual representation in both frequency and spatial domains. The introduction of frequency domain information enables ViM to have a global receptive field during scanning. We propose a novel model called Vim-F, which employs pure Mamba encoders and scans in both the frequency and spatial domains. Moreover, we question the necessity of position embedding in ViM and remove it accordingly in Vim-F, which helps to fully utilize the efficient long-sequence modeling capability of ViM. Finally, we redesign a patch embedding for Vim-F, leveraging a convolutional stem to capture more local correlations, further improving the performance of Vim-F. Code is available at: \url{https://github.com/yws-wxs/Vim-F}.

Online service platforms offering a wide range of services through miniapps have become crucial for users who visit these platforms with clear intentions to find services they are interested in. Aiming at effective content delivery, cross-domain recommendation are introduced to learn high-quality representations by transferring behaviors from data-rich scenarios. However, these methods overlook the impact of the decision path that users take when conduct behaviors, that is, users ultimately exhibit different behaviors based on various intents. To this end, we propose HIER, a novel Hierarchical decIsion path Enhanced Representation learning for cross-domain recommendation. With the help of graph neural networks for high-order topological information of the knowledge graph between multi-source behaviors, we further adaptively learn decision paths through well-designed exemplar-level and information bottleneck based contrastive learning. Extensive experiments in online and offline environments show the superiority of HIER.

Chain-of-thought distillation is a powerful technique for transferring reasoning abilities from large language models (LLMs) to smaller student models. Previous methods typically require the student to mimic the step-by-step rationale produced by LLMs, often facing the following challenges: (i) Tokens within a rationale vary in significance, and treating them equally may fail to accurately mimic keypoint tokens, leading to reasoning errors. (ii) They usually distill knowledge by consistently predicting all the steps in a rationale, which falls short in distinguishing the learning order of step generation. This diverges from the human cognitive progression of starting with easy tasks and advancing to harder ones, resulting in sub-optimal outcomes. To this end, we propose a unified framework, called KPOD, to address these issues. Specifically, we propose a token weighting module utilizing mask learning to encourage accurate mimicry of keypoint tokens by the student during distillation. Besides, we develop an in-rationale progressive distillation strategy, starting with training the student to generate the final reasoning steps and gradually extending to cover the entire rationale. To accomplish this, a weighted token generation loss is proposed to assess step reasoning difficulty, and a value function is devised to schedule the progressive distillation by considering both step difficulty and question diversity. Extensive experiments on four reasoning benchmarks illustrate our KPOD outperforms previous methods by a large margin., Comment: Accepted by ICML 2024

Detecting objects from Unmanned Aerial Vehicles (UAV) is often hindered by a large number of small objects, resulting in low detection accuracy. To address this issue, mainstream approaches typically utilize multi-stage inferences. Despite their remarkable detecting accuracies, real-time efficiency is sacrificed, making them less practical to handle real applications. To this end, we propose to improve the single-stage inference accuracy through learning scale-invariant features. Specifically, a Scale-Invariant Feature Disentangling module is designed to disentangle scale-related and scale-invariant features. Then an Adversarial Feature Learning scheme is employed to enhance disentanglement. Finally, scale-invariant features are leveraged for robust UAV-based object detection. Furthermore, we construct a multi-modal UAV object detection dataset, State-Air, which incorporates annotated UAV state parameters. We apply our approach to three state-of-the-art lightweight detection frameworks on three benchmark datasets, including State-Air. Extensive experiments demonstrate that our approach can effectively improve model accuracy. Our code and dataset are provided in Supplementary Materials and will be publicly available once the paper is accepted.

Recently, point cloud processing and analysis have made great progress due to the development of 3D Transformers. However, existing 3D Transformer methods usually are computationally expensive and inefficient due to their huge and redundant attention maps. They also tend to be slow due to requiring time-consuming point cloud sampling and grouping processes. To address these issues, we propose an efficient point TransFormer with Dynamic Token Aggregating (DTA-Former) for point cloud representation and processing. Firstly, we propose an efficient Learnable Token Sparsification (LTS) block, which considers both local and global semantic information for the adaptive selection of key tokens. Secondly, to achieve the feature aggregation for sparsified tokens, we present the first Dynamic Token Aggregating (DTA) block in the 3D Transformer paradigm, providing our model with strong aggregated features while preventing information loss. After that, a dual-attention Transformer-based Global Feature Enhancement (GFE) block is used to improve the representation capability of the model. Equipped with LTS, DTA, and GFE blocks, DTA-Former achieves excellent classification results via hierarchical feature learning. Lastly, a novel Iterative Token Reconstruction (ITR) block is introduced for dense prediction whereby the semantic features of tokens and their semantic relationships are gradually optimized during iterative reconstruction. Based on ITR, we propose a new W-net architecture, which is more suitable for Transformer-based feature learning than the common U-net design. Extensive experiments demonstrate the superiority of our method. It achieves SOTA performance with up to 30$\times$ faster than prior point Transformers on ModelNet40, ShapeNet, and airborne MultiSpectral LiDAR (MS-LiDAR) datasets., Comment: 16 pages, 12 figures, 7 tables

3D Transformers have achieved great success in point cloud understanding and representation. However, there is still considerable scope for further development in effective and efficient Transformers for large-scale LiDAR point cloud scene segmentation. This paper proposes a novel 3D Transformer framework, named 3D Learnable Supertoken Transformer (3DLST). The key contributions are summarized as follows. Firstly, we introduce the first Dynamic Supertoken Optimization (DSO) block for efficient token clustering and aggregating, where the learnable supertoken definition avoids the time-consuming pre-processing of traditional superpoint generation. Since the learnable supertokens can be dynamically optimized by multi-level deep features during network learning, they are tailored to the semantic homogeneity-aware token clustering. Secondly, an efficient Cross-Attention-guided Upsampling (CAU) block is proposed for token reconstruction from optimized supertokens. Thirdly, the 3DLST is equipped with a novel W-net architecture instead of the common U-net design, which is more suitable for Transformer-based feature learning. The SOTA performance on three challenging LiDAR datasets (airborne MultiSpectral LiDAR (MS-LiDAR) (89.3% of the average F1 score), DALES (80.2% of mIoU), and Toronto-3D dataset (80.4% of mIoU)) demonstrate the superiority of 3DLST and its strong adaptability to various LiDAR point cloud data (airborne MS-LiDAR, aerial LiDAR, and vehicle-mounted LiDAR data). Furthermore, 3DLST also achieves satisfactory results in terms of algorithm efficiency, which is up to 5x faster than previous best-performing methods., Comment: 13 pages, 10 figures, 7 tables

Time series forecasting is widely used in extensive applications, such as traffic planning and weather forecasting. However, real-world time series usually present intricate temporal variations, making forecasting extremely challenging. Going beyond the mainstream paradigms of plain decomposition and multiperiodicity analysis, we analyze temporal variations in a novel view of multiscale-mixing, which is based on an intuitive but important observation that time series present distinct patterns in different sampling scales. The microscopic and the macroscopic information are reflected in fine and coarse scales respectively, and thereby complex variations can be inherently disentangled. Based on this observation, we propose TimeMixer as a fully MLP-based architecture with Past-Decomposable-Mixing (PDM) and Future-Multipredictor-Mixing (FMM) blocks to take full advantage of disentangled multiscale series in both past extraction and future prediction phases. Concretely, PDM applies the decomposition to multiscale series and further mixes the decomposed seasonal and trend components in fine-to-coarse and coarse-to-fine directions separately, which successively aggregates the microscopic seasonal and macroscopic trend information. FMM further ensembles multiple predictors to utilize complementary forecasting capabilities in multiscale observations. Consequently, TimeMixer is able to achieve consistent state-of-the-art performances in both long-term and short-term forecasting tasks with favorable run-time efficiency.

In recent years, point cloud normal estimation, as a classical and foundational algorithm, has garnered extensive attention in the field of 3D geometric processing. Despite the remarkable performance achieved by current Neural Network-based methods, their robustness is still influenced by the quality of training data and the models' performance. In this study, we designed a fundamental framework for normal estimation, enhancing existing model through the incorporation of global information and various constraint mechanisms. Additionally, we employed a confidence-based strategy to select the reasonable samples for fair and robust network training. The introduced sample confidence can be integrated into the loss function to balance the influence of different samples on model training. Finally, we utilized existing orientation methods to correct estimated non-oriented normals, achieving state-of-the-art performance in both oriented and non-oriented tasks. Extensive experimental results demonstrate that our method works well on the widely used benchmarks.

Preference datasets are essential for incorporating human preferences into pre-trained language models, playing a key role in the success of Reinforcement Learning from Human Feedback. However, these datasets often demonstrate conflicting alignment objectives, leading to increased vulnerability to jailbreak attacks and challenges in adapting downstream tasks to prioritize specific alignment objectives without negatively impacting others. In this work, we introduce a novel statistical metric, Alignment Dimension Conflict, to quantify the degree of conflict within preference datasets. We then present \texttt{Hummer} and its fine-grained variant, \texttt{Hummer-F}, as innovative pairwise preference datasets with reduced-conflict alignment objectives. \texttt{Hummer} is built based on UltraFeedback and is enhanced by AI feedback from GPT-4, marking as the first preference dataset aimed at reducing the competition between alignment objectives. Furthermore, we develop reward models, HummerRM and HummerRM-F, which employ a hybrid sampling approach to balance diverse alignment objectives effectively. This sampling method positions HummerRM as an ideal model for domain-specific further fine-tuning and reducing vulnerabilities to attacks.

This study aims to develop a digital twin (DT) framework to enhance adaptive proton stereotactic body radiation therapy (SBRT) for prostate cancer. Prostate SBRT has emerged as a leading option for external beam radiotherapy due to its effectiveness and reduced treatment duration. However, interfractional anatomy variations can impact treatment outcomes. This study seeks to address these uncertainties using DT concept, with the goal of improving treatment quality, potentially revolutionizing prostate radiotherapy to offer personalized treatment solutions. Our study presented a pioneering approach that leverages DT technology to enhance adaptive proton SBRT. The framework improves treatment plans by utilizing patient-specific CTV setup uncertainty, which is usually smaller than conventional clinical setups. This research contributes to the ongoing efforts to enhance the efficiency and efficacy of prostate radiotherapy, with ultimate goals of improving patient outcomes and life quality.

In recent years, convolutional neural networks (CNNs) with channel-wise feature refining mechanisms have brought noticeable benefits to modelling channel dependencies. However, current attention paradigms fail to infer an optimal channel descriptor capable of simultaneously exploiting statistical and spatial relationships among feature maps. In this paper, to overcome this shortcoming, we present a novel channel-wise spatially autocorrelated (CSA) attention mechanism. Inspired by geographical analysis, the proposed CSA exploits the spatial relationships between channels of feature maps to produce an effective channel descriptor. To the best of our knowledge, this is the f irst time that the concept of geographical spatial analysis is utilized in deep CNNs. The proposed CSA imposes negligible learning parameters and light computational overhead to the deep model, making it a powerful yet efficient attention module of choice. We validate the effectiveness of the proposed CSA networks (CSA-Nets) through extensive experiments and analysis on ImageNet, and MS COCO benchmark datasets for image classification, object detection, and instance segmentation. The experimental results demonstrate that CSA-Nets are able to consistently achieve competitive performance and superior generalization than several state-of-the-art attention-based CNNs over different benchmark tasks and datasets.

This paper aims to create a deep learning framework that can estimate the deformation vector field (DVF) for directly registering abdominal MRI-CT images. The proposed method assumed a diffeomorphic deformation. By using topology-preserved deformation features extracted from the probabilistic diffeomorphic registration model, abdominal motion can be accurately obtained and utilized for DVF estimation. The model integrated Swin transformers, which have demonstrated superior performance in motion tracking, into the convolutional neural network (CNN) for deformation feature extraction. The model was optimized using a cross-modality image similarity loss and a surface matching loss. To compute the image loss, a modality-independent neighborhood descriptor (MIND) was used between the deformed MRI and CT images. The surface matching loss was determined by measuring the distance between the warped coordinates of the surfaces of contoured structures on the MRI and CT images. The deformed MRI image was assessed against the CT image using the target registration error (TRE), Dice similarity coefficient (DSC), and mean surface distance (MSD) between the deformed contours of the MRI image and manual contours of the CT image. When compared to only rigid registration, DIR with the proposed method resulted in an increase of the mean DSC values of the liver and portal vein from 0.850 and 0.628 to 0.903 and 0.763, a decrease of the mean MSD of the liver from 7.216 mm to 3.232 mm, and a decrease of the TRE from 26.238 mm to 8.492 mm. The proposed deformable image registration method based on a diffeomorphic transformer provides an effective and efficient way to generate an accurate DVF from an MRI-CT image pair of the abdomen. It could be utilized in the current treatment planning workflow for liver radiotherapy., Comment: 18 pages and 4 figures

Denoising is a crucial preprocessing step for hyperspectral images (HSIs) due to noise arising from intra-imaging mechanisms and environmental factors. Long-range spatial-spectral correlation modeling is beneficial for HSI denoising but often comes with high computational complexity. Based on the state space model (SSM), Mamba is known for its remarkable long-range dependency modeling capabilities and computational efficiency. Building on this, we introduce a memory-efficient spatial-spectral UMamba (SSUMamba) for HSI denoising, with the spatial-spectral continuous scan (SSCS) Mamba being the core component. SSCS Mamba alternates the row, column, and band in six different orders to generate the sequence and uses the bidirectional SSM to exploit long-range spatial-spectral dependencies. In each order, the images are rearranged between adjacent scans to ensure spatial-spectral continuity. Additionally, 3D convolutions are embedded into the SSCS Mamba to enhance local spatial-spectral modeling. Experiments demonstrate that SSUMamba achieves superior denoising results with lower memory consumption per batch compared to transformer-based methods. The source code is available at https://github.com/lronkitty/SSUMamba.

Diffusion-based technologies have made significant strides, particularly in personalized and customized facialgeneration. However, existing methods face challenges in achieving high-fidelity and detailed identity (ID)consistency, primarily due to insufficient fine-grained control over facial areas and the lack of a comprehensive strategy for ID preservation by fully considering intricate facial details and the overall face. To address these limitations, we introduce ConsistentID, an innovative method crafted for diverseidentity-preserving portrait generation under fine-grained multimodal facial prompts, utilizing only a single reference image. ConsistentID comprises two key components: a multimodal facial prompt generator that combines facial features, corresponding facial descriptions and the overall facial context to enhance precision in facial details, and an ID-preservation network optimized through the facial attention localization strategy, aimed at preserving ID consistency in facial regions. Together, these components significantly enhance the accuracy of ID preservation by introducing fine-grained multimodal ID information from facial regions. To facilitate training of ConsistentID, we present a fine-grained portrait dataset, FGID, with over 500,000 facial images, offering greater diversity and comprehensiveness than existing public facial datasets. % such as LAION-Face, CelebA, FFHQ, and SFHQ. Experimental results substantiate that our ConsistentID achieves exceptional precision and diversity in personalized facial generation, surpassing existing methods in the MyStyle dataset. Furthermore, while ConsistentID introduces more multimodal ID information, it maintains a fast inference speed during generation., Comment: Project page: https://ssugarwh.github.io/consistentid.github.io/

Bayesian flow networks (BFNs) iteratively refine the parameters, instead of the samples in diffusion models (DMs), of distributions at various noise levels through Bayesian inference. Owing to its differentiable nature, BFNs are promising in modeling both continuous and discrete data, while simultaneously maintaining fast sampling capabilities. This paper aims to understand and enhance BFNs by connecting them with DMs through stochastic differential equations (SDEs). We identify the linear SDEs corresponding to the noise-addition processes in BFNs, demonstrate that BFN's regression losses are aligned with denoise score matching, and validate the sampler in BFN as a first-order solver for the respective reverse-time SDE. Based on these findings and existing recipes of fast sampling in DMs, we propose specialized solvers for BFNs that markedly surpass the original BFN sampler in terms of sample quality with a limited number of function evaluations (e.g., 10) on both image and text datasets. Notably, our best sampler achieves an increase in speed of 5~20 times for free. Our code is available at https://github.com/ML-GSAI/BFN-Solver., Comment: Published as a conference paper at ICML 2024

Radiance fields have demonstrated impressive performance in synthesizing lifelike 3D talking heads. However, due to the difficulty in fitting steep appearance changes, the prevailing paradigm that presents facial motions by directly modifying point appearance may lead to distortions in dynamic regions. To tackle this challenge, we introduce TalkingGaussian, a deformation-based radiance fields framework for high-fidelity talking head synthesis. Leveraging the point-based Gaussian Splatting, facial motions can be represented in our method by applying smooth and continuous deformations to persistent Gaussian primitives, without requiring to learn the difficult appearance change like previous methods. Due to this simplification, precise facial motions can be synthesized while keeping a highly intact facial feature. Under such a deformation paradigm, we further identify a face-mouth motion inconsistency that would affect the learning of detailed speaking motions. To address this conflict, we decompose the model into two branches separately for the face and inside mouth areas, therefore simplifying the learning tasks to help reconstruct more accurate motion and structure of the mouth region. Extensive experiments demonstrate that our method renders high-quality lip-synchronized talking head videos, with better facial fidelity and higher efficiency compared with previous methods., Comment: Accepted at ECCV 2024. Project page: https://fictionarry.github.io/TalkingGaussian/

This paper develops an innovative method that enables neural networks to generate and utilize knowledge graphs, which describe their concept-level knowledge and optimize network parameters through alignment with human-provided knowledge. This research addresses a gap where traditionally, network-generated knowledge has been limited to applications in downstream symbolic analysis or enhancing network transparency. By integrating a novel autoencoder design with the Vector Symbolic Architecture (VSA), we have introduced auxiliary tasks that support end-to-end training. Our approach eschews traditional dependencies on ontologies or word embedding models, mining concepts from neural networks and directly aligning them with human knowledge. Experiments show that our method consistently captures network-generated concepts that align closely with human knowledge and can even uncover new, useful concepts not previously identified by humans. This plug-and-play strategy not only enhances the interpretability of neural networks but also facilitates the integration of symbolic logical reasoning within these systems.

Ultrarelativistic vortex leptons with intrinsic orbital angular momenta (OAM) have important applications in high energy particle physics, nuclear physics, astrophysics, etc. However, unfortunately, their generation still poses a great challenge. Here, we put forward a novel method for generating ultrarelativistic vortex positrons and electrons through nonlinear Breit-Wheeler (NBW) scattering of vortex $\gamma$ photons. For the first time, a complete angular momentum-resolved scattering theory has been formulated, introducing the angular momentum of laser photons and vortex particles into the conventional NBW scattering framework. We find that vortex positron (electron) can be produced when the outgoing electron (positron) is generated along the collision axis. By unveiling the angular momentum transfer mechanism, we clarify that OAM of the $\gamma$ photon and angular momenta of multiple laser photons are entirely transferred to the generated pairs, leading to the production of ultrarelativistic vortex positrons or electrons with large OAM. Furthermore, we find that the cone opening angle and superposition state of the vortex $\gamma$ photon, distinct characteristics aside from its intrinsic OAM, can be determined via the angular distribution of created pairs in NBW processes. Our method paves the way for investigating strong-field quantum electrodynamics processes concerning the generation and detection of vortex particle beams in intense lasers., Comment: 5

Offline batch inference is a common task in the industry for deep learning applications, but it can be challenging to ensure stability and performance when dealing with large amounts of data and complicated inference pipelines. This paper demonstrated AntBatchInfer, an elastic batch inference framework, which is specially optimized for the non-dedicated cluster. AntBatchInfer addresses these challenges by providing multi-level fault-tolerant capabilities, enabling the stable execution of versatile and long-running inference tasks. It also improves inference efficiency by pipelining, intra-node, and inter-node scaling. It further optimizes the performance in complicated multiple-model batch inference scenarios. Through extensive experiments and real-world statistics, we demonstrate the superiority of our framework in terms of stability and efficiency. In the experiment, it outperforms the baseline by at least $2\times$ and $6\times$ in the single-model or multiple-model batch inference. Also, it is widely used at Ant Group, with thousands of daily jobs from various scenarios, including DLRM, CV, and NLP, which proves its practicability in the industry.

Many distributed training techniques like Parameter Server and AllReduce have been proposed to take advantage of the increasingly large data and rich features. However, stragglers frequently occur in distributed training due to resource contention and hardware heterogeneity, which significantly hampers the training efficiency. Previous works only address part of the stragglers and could not adaptively solve various stragglers in practice. Additionally, it is challenging to use a systematic framework to address all stragglers because different stragglers require diverse data allocation and fault-tolerance mechanisms. Therefore, this paper proposes a unified distributed training framework called AntDT (Ant Distributed Training Framework) to adaptively solve the straggler problems. Firstly, the framework consists of four components, including the Stateful Dynamic Data Sharding service, Monitor, Controller, and Agent. These components work collaboratively to efficiently distribute workloads and provide a range of pre-defined straggler mitigation methods with fault tolerance, thereby hiding messy details of data allocation and fault handling. Secondly, the framework provides a high degree of flexibility, allowing for the customization of straggler mitigation solutions based on the specific circumstances of the cluster. Leveraging this flexibility, we introduce two straggler mitigation solutions, namely AntDT-ND for non-dedicated clusters and AntDT-DD for dedicated clusters, as practical examples to resolve various types of stragglers at Ant Group. Justified by our comprehensive experiments and industrial deployment statistics, AntDT outperforms other SOTA methods more than 3x in terms of training efficiency. Additionally, in Alipay's homepage recommendation scenario, using AntDT reduces the training duration of the ranking model from 27.8 hours to just 5.4 hours.