Large language models (LLMs) have become an important semantic infrastructure for modern recommender systems. A prevailing paradigm integrates LLM-derived semantic embeddings with collaborative representations via representation alignment, implicitly assuming that the two views encode a shared latent entity and that stronger alignment yields better results. We formalize this assumption as the global low-complexity alignment hypothesis and argue that it is stronger than necessary and often structurally mismatched with real-world recommendation settings. We propose a complementary perspective in which semantic and collaborative representations are treated as partially shared yet fundamentally heterogeneous views, each containing both shared and view-specific factors. Under this shared-plus-private latent structure, enforcing global geometric alignment may distort local structure, suppress view-specific signals, and reduce informational diversity. To support this perspective, we develop complementarity-aware diagnostics that quantify overlap, unique-hit contribution, and theoretical fusion upper bounds. Empirical analyses on sparse recommendation benchmarks reveal low item-level agreement between semantic and collaborative views and substantial oracle fusion gains, indicating strong complementarity. Furthermore, controlled alignment probes show that low-capacity mappings capture only shared components and fail to recover full collaborative geometry, especially under distribution shift. These findings suggest that alignment should not be treated as the default integration principle. We advocate a shift from alignment-centric modeling to complementarity fusion-centric, complementarity-aware design, where shared factors are selectively integrated while private signals are preserved. This reframing provides a principled foundation for the next generation of LLM-enhanced recommender systems.

The escalating complexity of industrial recommendation systems, characterized by diverse user behaviors and cross-domain application scenarios, necessitates advanced multi-task and multi-domain learning paradigms. Existing methods often struggle with efficient knowledge transfer across tasks and domains due to semantic gaps and distribution shifts. To address these challenges, we propose the Perception-Aware Multi-Task Transformer Network for Personalized Multi-Domain Adaptation (PMTA), a unified framework that integrates three key innovations: First, the Task Prompt Encoding (TPE) module dynamically generates prompts by synthesizing personalized user data with task-specific information. Second, the Transformer-based Multi-Task Perception (TMPN) network enables adaptive cross-task knowledge transfer through attention mechanisms. Third, the Multi-Domain Adaptation (MDAN) component captures domain-specific behavior patterns via learnable prior information. Experimental results demonstrate PMTA's effectiveness, achieving 0.168% increase in watch time and significant improvements in engagement metrics (AAD: +0.0113%, AAH: +0.0608%). Deployed on Douyin and Douyin Lite, it significantly improves recommendation quality and drives commercial success.

Drug-target binding affinity prediction is a fundamental task for drug discovery. It has been extensively explored in literature and promising results are reported. However, in this paper, we demonstrate that the results may be misleading and cannot be well generalized to real practice. The core observation is that the canonical randomized split of a test set in conventional evaluation leaves the test set dominated by samples with high similarity to the training set. The performance of models is severely degraded on samples with lower similarity to the training set but the drawback is highly overlooked in current evaluation. As a result, the performance can hardly be trusted when the model meets low-similarity samples in real practice. To address this problem, we propose a framework of similarity aware evaluation in which a novel split methodology is proposed to adapt to any desired distribution. This is achieved by a formulation of optimization problems which are approximately and efficiently solved by gradient descent. We perform extensive experiments across five representative methods in four datasets for two typical target evaluations and compare them with various counterpart methods. Results demonstrate that the proposed split methodology can significantly better fit desired distributions and guide the development of models.

Sequential recommendation, a critical task in recommendation systems, predicts the next user action based on the understanding of the user's historical behaviors. Conventional studies mainly focus on cross-behavior modeling with self-attention based methods while neglecting comprehensive user interest modeling for more dimensions. In this study, we propose a novel sequential recommendation model, Pyramid Mixer, which leverages the MLP-Mixer architecture to achieve efficient and complete modeling of user interests. Our method learns comprehensive user interests via cross-behavior and cross-feature user sequence modeling. The mixer layers are stacked in a pyramid way for cross-period user temporal interest learning. Through extensive offline and online experiments, we demonstrate the effectiveness and efficiency of our method, and we obtain a +0.106% improvement in user stay duration and a +0.0113% increase in user active days in the online A/B test. The Pyramid Mixer has been successfully deployed on the industrial platform, demonstrating its scalability and impact in real-world applications.

In large-scale recommendation systems, modeling long-term user interests is progressively gaining attention among researchers and practitioners. Existing work, such as SIM and TWIN, typically employs a two-stage approach to model long-term user behavior sequences for efficiency concerns. The first stage rapidly retrieves a subset of sequences related to the target item from a long sequence using a search-based mechanism namely the General Search Unit (GSU), while the second stage calculates the interest scores using the Exact Search Unit (ESU) on the retrieved results. Given the extensive length of user behavior sequences spanning the entire life cycle, potentially reaching up to 10^6 in scale, there is currently no effective solution for fully modeling such expansive user interests. To overcome this issue, we introduced TWIN-V2, an enhancement of TWIN, where a divide-and-conquer approach is applied to compress life-cycle behaviors and uncover more accurate and diverse user interests. Specifically, a hierarchical clustering method groups items with similar characteristics in life-cycle behaviors into a single cluster during the offline phase. By limiting the size of clusters, we can compress behavior sequences well beyond the magnitude of 10^5 to a length manageable for online inference in GSU retrieval. Cluster-aware target attention extracts comprehensive and multi-faceted long-term interests of users, thereby making the final recommendation results more accurate and diverse. Extensive offline experiments on a multi-billion-scale industrial dataset and online A/B tests have demonstrated the effectiveness of TWIN-V2. Under an efficient deployment framework, TWIN-V2 has been successfully deployed to the primary traffic that serves hundreds of millions of daily active users at Kuaishou.

Domain shift is a prevalent issue for medical imaging in both the cross-modality and cross-site manner. As a result, the research of domain generalization meets a significant need that requires no access to any samples in novel target domains. In this paper, we propose BACON, a boundary-aware contrastive learning framework for single-source domain generalization, which enjoys the advantages of both the data-driven methods (e.g. augmentation) and inductive bias methods (e.g. normalization and whitening). BACON designs a novel data augmentation by integrating the bidirectional CutMix with the RandConv and introduces inductive bias through an innovative contrastive learning loss function. Experiments on both the cross-modality and cross-site scenario demonstrate the effectiveness, where BACON achieves a substantial improvement of 33.51% and 33.47% Dice scores over the baseline in the BraTS 2020 dataset and sets the new state of the art on the Prostate Cross-site dataset with an average Dice score of 72.51%.

Previous works on video object segmentation (VOS) are trained on densely annotated videos. Nevertheless, acquiring annotations in pixel level is expensive and time-consuming. In this work, we demonstrate the feasibility of training a satisfactory VOS model on sparsely annotated videos—we merely require two labeled frames per training video while the performance is sustained. We term this novel training paradigm as two-shot video object segmentation, or two-shot VOS for short. The underlying idea is to generate pseudo labels for unlabeled frames during training and to optimize the model on the combination of labeled and pseudo-labeled data. Our approach is extremely simple and can be applied to a majority of existing frameworks. We first pre-train a VOS model on sparsely annotated videos in a semi-supervised manner, with the first frame always being a labeled one. Then, we adopt the pretrained VOS model to generate pseudo labels for all unlabeled frames, which are subsequently stored in a pseudo-label bank. Finally, we retrain a VOS model on both labeled and pseudo-labeled data without any restrictions on the first frame. For the first time, we present a general way to train VOS models on two-shot VOS datasets. By using 7.3% and 2.9% labeled data of YouTube-VOS and DAVIS benchmarks, our approach achieves comparable results in contrast to the counterparts trained on fully labeled set. Code and models are available at https://github.com/yk-pku/Two-shot-Video-Object-Segmentation.

Segmentation of magnetic resonance images is an essential way of measuring the volume of tissues and lesions, which can improve the efficiency of diagnosis. The mainstream image segmentation methods are based on deep learning, which requires a large amount of labeled data. However, labeling magnetic resonance images is expensive and time-consuming. Therefore, we propose a consistent teacher-student model for magnetic resonance image segmentation, which is abbreviated as CTSSeg. Specifically, the CTSSeg includes a student network and a teacher network, where the student network learns supervised from labeled data, while the teacher network utilizes unlabeled data to improve the student network via contrastive learning and pseudo-label learning. We evaluate the proposed CTSSeg on the Atrial Segmentation Challenge dataset and a local clinical dataset. The experimental results show that our method can make full use of both labeled and unlabeled data and yield state-of-the-art performance.

Life-long user behavior modeling, i.e., extracting a user's hidden interests from rich historical behaviors in months or even years, plays a central role in modern CTR prediction systems. Conventional algorithms mostly follow two cascading stages: a simple General Search Unit (GSU) for fast and coarse search over tens of thousands of long-term behaviors and an Exact Search Unit (ESU) for effective Target Attention (TA) over the small number of finalists from GSU. Although efficient, existing algorithms mostly suffer from a crucial limitation: the inconsistent target-behavior relevance metrics between GSU and ESU. As a result, their GSU usually misses highly relevant behaviors but retrieves ones considered irrelevant by ESU. In such case, the TA in ESU, no matter how attention is allocated, mostly deviates from the real user interests and thus degrades the overall CTR prediction accuracy. To address such inconsistency, we propose TWo-stage Interest Network (TWIN), where our Consistency-Preserved GSU (CP-GSU) adopts the identical target-behavior relevance metric as the TA in ESU, making the two stages twins. Specifically, to break TA's computational bottleneck and extend it from ESU to GSU, or namely from behavior length 102 to length 104 - 105, we build a novel attention mechanism by behavior feature splitting. For the video inherent features of a behavior, we calculate their linear projection by efficient pre-computing & caching strategies. And for the user-item cross features, we compress each into a one-dimentional bias term in the attention score calculation to save the computational cost. The consistency between two stages, together with the effective TA-based relevance metric in CP-GSU, contributes to significant performance gain in CTR prediction. Offline experiments on a 46 billion scale real production dataset from Kuaishou and an Online A/B test show that TWIN outperforms all compared SOTA algorithms. With optimized online infrastructure, we reduce the computational bottleneck by 99.3%, which contributes to the successful deployment of TWIN on Kuaishou, serving the main traffic of hundreds of millions of active users everyday.

With the increase of content pages and interactive buttons in online services such as online-shopping and video-watching websites, industrial-scale recommender systems face challenges in multi-domain and multi-task recommendations. The core of multi-task and multi-domain recommendation is to accurately capture user interests in multiple scenarios given multiple user behaviors. In this paper, we propose a plug-and-play Parameter and Embedding Personalized Network (PEPNet) for multi-domain and multi-task recommendation. PEPNet takes personalized prior information as input and dynamically scales the bottom-level Embedding and top-level DNN hidden units through gate mechanisms. Embedding Personalized Network (EPNet) performs personalized selection on Embedding to fuse features with different importance for different users in multiple domains. Parameter Personalized Network (PPNet) executes personalized modification on DNN parameters to balance targets with different sparsity for different users in multiple tasks. We have made a series of special engineering optimizations combining the Kuaishou training framework and the online deployment environment. By infusing personalized selection of Embedding and personalized modification of DNN parameters, PEPNet tailored to the interests of each individual obtains significant performance gains, with online improvements exceeding 1% in multiple task metrics across multiple domains. We have deployed PEPNet in Kuaishou apps, serving over 300 million users every day.

Drug-target binding affinity prediction is a fundamental task for drug discovery and has been studied for decades. Most methods follow the canonical paradigm that processes the inputs of the protein (target) and the ligand (drug) separately and then combines them together. In this study we demonstrate, surprisingly, that a model is able to achieve even superior performance without access to any protein-sequence-related information. Instead, a protein is characterized completely by the ligands that it interacts. Specifically, we treat different proteins separately, which are jointly trained in a multi-head manner, so as to learn a robust and universal representation of ligands that is generalizable across proteins. Empirical evidences show that the novel paradigm outperforms its competitive sequence-based counterpart, with the Mean Squared Error (MSE) of 0.4261 versus 0.7612 and the R-Square of 0.7984 versus 0.6570 compared with DeepAffinity. We also investigate the transfer learning scenario where unseen proteins are encountered after the initial training, and the cross-dataset evaluation for prospective studies. The results reveals the robustness of the proposed model in generalizing to unseen proteins as well as in predicting future data. Source codes and data are available at https://github.com/huzqatpku/SAM-DTA.

Multi-label few-shot image classification (ML-FSIC) is the task of assigning descriptive labels to previously unseen images, based on a small number of training examples. A key feature of the multi-label setting is that images often have multiple labels, which typically refer to different regions of the image. When estimating prototypes, in a metric-based setting, it is thus important to determine which regions are relevant for which labels, but the limited amount of training data makes this highly challenging. As a solution, in this paper we propose to use word embeddings as a form of prior knowledge about the meaning of the labels. In particular, visual prototypes are obtained by aggregating the local feature maps of the support images, using an attention mechanism that relies on the label embeddings. As an important advantage, our model can infer prototypes for unseen labels without the need for fine-tuning any model parameters, which demonstrates its strong generalization abilities. Experiments on COCO and PASCAL VOC furthermore show that our model substantially improves the current state-of-the-art.

The prosperity of deep neural networks in various computer vision applications heavily depends on large-scale high-quality annotated datasets. However, inaccurate annotations are often inevitable when building hand-labeled datasets, which can lead to imperfectly supervision problems. Most existing methods to address these problems mainly use noisy labels to learn latent representations, which may be corrupted when the noise rate is high, thus limiting the predictions for clean test data set. In this paper, we propose a flexible coarse-to-fine representation learning framework to improve the quality of latent representations by exploiting the reliable geometry information from raw features. This framework includes: (i) an anchored representation learner, which is a coarse mechanism for learning reasonable feature-similarity to avoid feature corruption, (ii) a confident representation learner, which is used to learn confident data pairs, and (iii) an exploratory representation learner, which is proposed for learning from feasible unconfident pairs and performing label refurbishment. Comprehensive experiments have demonstrated the effectiveness of our proposed method, especially when the noise level is high and the task is relatively difficult.

Multimodal Machine Comprehension ($\rm M^3C$) has been a challenging task that requires understanding both language and vision, as well as their integration and interaction. For example, the RecipeQA challenge, which provides several $\rm M^3C$ tasks, requires deep neural models to understand textual instructions, images of different steps, as well as the logic orders of food cooking. To address this challenge, we propose a Multi-Level Multi-Modal Transformer (MLMM-Trans) framework to integrate and understand multiple textual instructions and multiple images. Our model can conduct intensive attention mechanism at multiple levels of objects (e.g., step level and passage-image level) for sequences of different modalities. Experiments have shown that our model can achieve the state-of-the-art results on the three multimodal tasks of RecipeQA.

Multi-Choice Reading Comprehension~(MCRC) is an essential task where a machine selects the correct answer from multiple choices given a context document and a corresponding question. Existing methods usually make predictions based on a single-round reasoning process with the attention mechanism, however, this may be insufficient for tasks that require a more complex reasoning process. To effectively comprehend the context and select the correct answer from different perspectives, we propose the Read-Attend-Exclude (RAE) model which is motivated by what human readers do for MCRC in multi-rounds reasoning process. Specifically, the RAE model includes four components: the Scan Reading Module, the Attended Intensive Reading Module, the Answer Exclusion Module, and the Gated Fusion Module that makes the final decisions collectively based on the aforementioned three modules. Extensive experiments demonstrate the strong results of the proposed model on the DREAM dataset and the effectiveness of all proposed modules.

Response selection plays an important role in fully automated dialogue systems. Given the dialogue context, the goal of response selection is to identify the best-matched next utterance (i.e., response) from multiple candidates. Despite the efforts of many previous useful models, this task remains challenging due to the huge semantic gap and also the large size of candidate set. To address these issues, we propose a Spatio-Temporal Matching network (STM) for response selection. In detail, soft alignment is first used to obtain the local relevance between the context and the response. And then, we construct spatio-temporal features by aggregating attention images in time dimension and make use of 3D convolution and pooling operations to extract matching information. Evaluation on two large-scale multi-turn response selection tasks has demonstrated that our proposed model significantly outperforms the state-of-the-art model. Particularly, visualization analysis shows that the spatio-temporal features enables matching information in segment pairs and time sequences, and have good interpretability for multi-turn text matching.

In this paper, we study the machine reading comprehension of temporal order in text. Given a document of instruction sequences, a model aims to find out the most coherent sequences of activities matching the document among all answer candidates. To tackle the task, we proposeOrdMatch model, which is able to match each activity in a sequence to the corresponding instruction in the document and regularizes the partial order of activities to match the order of instructions. We evaluate the task using the RecipeQA dataset, which includes step-by-step instructions of cooking recipes. Our model outperforms the state-of-the-art models with a wide margin. The experimental results demonstrate the effectiveness of our novel ordering regularizer. Our code will be made available at \hrefhttps://github.com/Aolius/OrdMatch https://github.com/Aolius/OrdMatch.

Authentication is a task aiming to confirm the truth between data instances and personal identities. Typical authentication applications include face recognition, person re-identification, authentication based on mobile devices and so on. The recently-emerging data-driven authentication process may encounter undesired biases, i.e., the models are often trained in one domain (e.g., for people wearing spring outfits) while required to apply in other domains (e.g., they change the clothes to summer outfits). To address this issue, we propose a novel two-stage method that disentangles the class/identity from domain-differences, and we consider multiple types of domain-difference. In the first stage, we learn disentangled representations by a one-versus-rest disentangle learning (OVRDL) mechanism. In the second stage, we improve the disentanglement by an additive adversarial learning (AAL) mechanism. Moreover, we discuss the necessity to avoid a learning dilemma due to disentangling causally related types of domain-difference. Comprehensive evaluation results demonstrate the effectiveness and superiority of the proposed method.

Restricted Boltzmann Machines are important and useful generative models which learn a probability distribution from a set of vector inputs. Despite their success in a number of applications, standard RBMs designed for vectorized inputs are incapable of dealing with high-order data, since vectorization of high-order data may cause both modes collapsing and explosive parameter growth. To address this issue, we formulate a new tensor-input RBM model, which employs the tensor-ring (TR) decomposition structure to naturally represent the high-order relationship between the visual layer and the hidden layer. For convenience, we name the proposed model as TR-RBM. In particular, the tensor ring decomposition enjoys many good properties, such as the rank stableness, leading to better generalization performance compared with other low-rank decomposition methods. Moreover, TR-RBM can also reduce the complexity of RBM by reshaping of both visible and hidden layers into the tensor forms, leading a significant drop of parameter size. Experimental results in comparison with the classical RBMs and the Matrix-Product-Operator RBM have shown the promising performance of the proposed method in the tasks of feature extraction and denoising.