Industrial large-scale recommendation models (LRMs) face the challenge of jointly modeling long-range user behavior sequences and heterogeneous non-sequential features under strict efficiency constraints. However, most existing architectures employ a decoupled pipeline: long sequences are first compressed with a query-token based sequence compressor like LONGER, followed by fusion with dense features through token-mixing modules like RankMixer, which thereby limits both the representation capacity and the interaction flexibility. This paper presents HyFormer, a unified hybrid transformer architecture that tightly integrates long-sequence modeling and feature interaction into a single backbone. From the perspective of sequence modeling, we revisit and redesign query tokens in LRMs, and frame the LRM modeling task as an alternating optimization process that integrates two core components: Query Decoding which expands non-sequential features into Global Tokens and performs long sequence decoding over layer-wise key-value representations of long behavioral sequences; and Query Boosting which enhances cross-query and cross-sequence heterogeneous interactions via efficient token mixing. The two complementary mechanisms are performed iteratively to refine semantic representations across layers. Extensive experiments on billion-scale industrial datasets demonstrate that HyFormer consistently outperforms strong LONGER and RankMixer baselines under comparable parameter and FLOPs budgets, while exhibiting superior scaling behavior with increasing parameters and FLOPs. Large-scale online A/B tests in high-traffic production systems further validate its effectiveness, showing significant gains over deployed state-of-the-art models. These results highlight the practicality and scalability of HyFormer as a unified modeling framework for industrial LRMs.

Short video has witnessed rapid growth in the past few years in multimedia platforms. To ensure the freshness of the videos, platforms receive a large number of user-uploaded videos every day, making collaborative filtering-based recommender methods suffer from the item cold-start problem (e.g., the new-coming videos are difficult to compete with existing videos). Consequently, increasing efforts tackle the cold-start issue from the content perspective, focusing on modeling the multi-modal preferences of users, a fair way to compete with new-coming and existing videos. However, recent studies ignore the existing gap between multi-modal embedding extraction and user interest modeling as well as the discrepant intensities of user preferences for different modalities. In this paper, we propose M3CSR, a multi-modal modeling framework for cold-start short video recommendation. Specifically, we preprocess content-oriented multi-modal features for items and obtain trainable category IDs by performing clustering. In each modality, we combine modality-specific cluster ID embedding and the mapped original modality feature as modality-specific representation of the item to address the gap. Meanwhile, M3CSR measures the user modality-specific intensity based on the correlation between modality-specific interest and behavioral interest and employs pairwise loss to further decouple user multi-modal interests. Extensive experiments on four real-world datasets demonstrate the superiority of our proposed model. The framework has been deployed on a billion-user scale short video application and has shown improvements in various commercial metrics within cold-start scenarios.

The item cold-start problem is crucial for online recommender systems, as the success of the cold-start phase determines whether items can transition into popular ones. Prompt learning, a powerful technique used in natural language processing (NLP) to address zero- or few-shot problems, has been adapted for recommender systems to tackle similar challenges. However, existing methods typically rely on content-based properties or text descriptions for prompting, which we argue may be suboptimal for cold-start recommendations due to 1) semantic gaps with recommender tasks, 2) model bias caused by warm-up items contribute most of the positive feedback to the model, which is the core of the cold-start problem that hinders the recommender quality on cold-start items. We propose to leverage high-value positive feedback, termed pinnacle feedback as prompt information, to simultaneously resolve the above two problems. We experimentally prove that compared to the content description proposed in existing works, the positive feedback is more suitable to serve as prompt information by bridging the semantic gaps. Besides, we propose item-wise personalized prompt networks to encode pinnaclce feedback to relieve the model bias by the positive feedback dominance problem. Extensive experiments on four real-world datasets demonstrate the superiority of our model over state-of-the-art methods. Moreover, PROMO has been successfully deployed on a popular short-video sharing platform, a billion-user scale commercial short-video application, achieving remarkable performance gains across various commercial metrics within cold-start scenarios.

Multi-view Graph Learning is popular in recommendations due to its ability to capture relationships and connections across multiple views. Existing multi-view graph learning methods generally involve constructing graphs of views and performing information aggregation on view representations. Despite their effectiveness, they face two data limitations: Multi-focal Multi-source data noise and multi-source Data Sparsity. The former arises from the combination of noise from individual views and conflicting edges between views when information from all views is combined. The latter occurs because multi-view learning exacerbate the negative influence of data sparsity because these methods require more model parameters to learn more view information. Motivated by these issues, we propose MMGCL, a meta knowledge-enhanced multi-view graph contrastive learning framework for recommendations. To tackle the data noise issue, MMGCL extract meta knowledge to preserve important information from all views to form a meta view representation. It then rectifies every view in multi-learning frameworks, thus simultaneously removing the view-private noisy edges and conflicting edges across different views. To address the data sparsity issue, MMGCL performs meta knowledge transfer contrastive learning optimization on all views to reduce the searching space for model parameters and add more supervised signal. Besides, we have deployed MMGCL in a real industrial recommender system in China, and we further evaluate it on three benchmark datasets and a practical industry online application. Extensive experiments on these datasets demonstrate the state-of-the-art recommendation performance of MMGCL.

Parameter shielding refers to the situation that one test parameter disables others in test execution.The quality of test case generation techniques is limited by the wide existence of parameter shielding.It is challenging to automatically find out conditions that cause the parameter shielding.This paper presents a novel approach for exploring the shielding conditions of test parameters.Our approach executes test inputs and collects runtime information of execution as features of test inputs.Then, a clustering algorithm is used to group test inputs with similar runtime information while a decision tree algorithm is built to extract the conditions in the groups.Finally, our approach identifies the shielding conditions based on the decision tree.Experiments on seven programs show that our approach can effectively detect the parameter shielding and the related conditions.