Transformers are the cornerstone of modern large language models, but their quadratic computational complexity limits efficiency in long-sequence processing. Recent advancements in Mamba, a state space model (SSM) with linear complexity, offer promising efficiency gains but suffer from unstable contextual learning and multitask generalization. Some works conduct layer-level hybrid structures that combine Transformer and Mamba layers, aiming to make full use of both advantages. This paper proposes TransMamba, a novel sequence-level hybrid framework that unifies Transformer and Mamba through shared parameter matrices (QKV and CBx), and thus could dynamically switch between attention and SSM mechanisms at different token lengths and layers. We design the Memory Converter to bridge Transformer and Mamba by converting attention outputs into SSM-compatible states, ensuring seamless information flow at TransPoints where the transformation happens. The TransPoint scheduling is also thoroughly explored for balancing effectiveness and efficiency. We conducted extensive experiments demonstrating that TransMamba achieves superior training efficiency and performance compared to single and hybrid baselines, and validated the deeper consistency between Transformer and Mamba paradigms at sequence level, offering a scalable solution for next-generation language modeling.

Mixture-of-Experts (MoE) models rely on an external router to assign tokens to experts.This design inherently separates the routing decision from each expert's internal capabilities, leading to suboptimal performance.In this work, we address this limitation with Union-of-Experts (UoE), an MoE variant that performs "expertautonomous routing".The core mechanism of UoE is to pre-designate a minute fraction of neurons within each expert as routing neurons.Experts autonomously select relevant tokens by comparing the activation intensity of these neurons, aligning routing decisions with each expert's functional profile.To prevent the waste of activations from unselected experts, we aggregate all routing neuron outputs and sum them into the final layer output.This aggregation acts as a novel virtual shared expert whose parameters are distributed across the individual experts, and improves overall parameter efficiency.We pre-train UoE models with up to 3B parameters, demonstrating that they outperform traditional MoEs with matched efficiency.Furthermore, our analysis of the routing neurons provides valuable insights into expert-autonomous selection and advances the understanding of MoE routing.

Hengyuan Zhang, Shiping Yang, Xiao Liang, Chenming Shang, Yuxuan Jiang, Chaofan Tao, Jing Xiong, Hayden Kwok-Hay So, Ruobing Xie, Angel X. Chang, Ngai Wong. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Question and answer (Q&A) platforms usually recommend question-answer pairs to meet users' knowledge acquisition needs, unlike traditional recommendations that recommend only one item. This makes user behaviors more complex, and presents two challenges for Q&A recommendation, including: the collaborative information entanglement, which means user feedback is influenced by either the question or the answer; and the semantic information entanglement, where questions are correlated with their corresponding answers, and correlations also exist among different question-answer pairs. Traditional recommendation methods treat the question-answer pair as a whole or only consider the answer as a single item, which overlooks the two challenges and cannot effectively model user interests. To address these challenges, we introduce a graph neural network model named Question & Answer Graph Collaborative Filtering (QAGCF). QAGCF creates graphs separately from collaborative and semantic views to disentangle the collaborative and semantic information of question-answer pairs. The collaborative view disentangles questions and answers to individually model collaborative information, while the semantic view captures the semantic information both within and between question-answer pairs. These views are further merged into a global graph to integrate the collaborative and semantic information. Polynomial-based graph filters are used to address the high heterophily issues of the global graph. Additionally, contrastive learning is utilized to obtain robust embeddings during training. Extensive experiments on industrial and public datasets demonstrate that QAGCF consistently outperforms baselines and achieves state-of-the-art results.

Recommendations usually focus on immediate accuracy metrics like Click-Through Rate (CTR), ignoring user long-term metrics. User retention, which reflects the percentage of today’s users who will return to the system in the next few days, should be paid more attention to. However, most existing methods did not focus on user retention, since their complexity and uncertainty make it extremely hard to discover why a user will or will not return to a system. Recently, a few pioneers have optimized user retention, focusing solely on accuracy without delving into its underlying rationale. This is primarily due to the absence of explicit supervised signals. In this work, we design a Behavior-wise Contrastive Multi-Instance Learning (BCMIL) module, which jointly models clicked and impressed items to capture interpretable user retention. Specifically, we conduct in-depth analyses in real-world scenarios to discover implicit retention-related supervised signals. To model these signals, we design a Forward Supervised Signals Extractor (FSSE) that utilizes a heterogeneous graph, enhancing the reliability of user retention. To mitigate randomness and uncertainty, we propose a Backward Supervised Signals Stabilizer (BSSS) that utilizes overlooked label-part behaviors within each training window to retrospectively guide the training process. Offline and online evaluations of an industrial system verify the effectiveness of our methods.

Recommender system (RS) aims to capture personalized preferences from massive user behaviors, making them pivotal in the era of information explosion. However, the presence of “information cocoons,” interaction sparsity, cold-start problem, and feedback loops inherent in RS make users interact with a limited number of items. Conventional recommendation algorithms typically focus on the positive historical behaviors, while neglecting the essential role of negative feedback in user preference understanding. As a promising but easy-to-ignored area, negative sampling is proficient in revealing the genuine negative aspect inherent in user behaviors, emerging as an inescapable procedure in RS. In this survey, we first discuss existing user feedback, the critical role of negative sampling and the optimization objectives in RS, and thoroughly analyze challenges that consistently impede its progress. Then, we conduct an extensive literature review on the existing negative sampling strategies in RS and classify them into five categories with their discrepant techniques. Finally, we detail the insights of the tailored negative sampling strategies in diverse RS scenarios and outline an overview of the prospective research directions toward which the community may engage and benefit.

This is an erratum for the article “Recommendation as Instruction Following: A Large Language Model Empowered Recommendation Approach” published in ACM Trans. Inf. Syst. 43, 5, Article 114 (July 2025), 37 pages.

Contrastive learning is a prevalent technique in self-supervised vision representation learning, typically generating positive pairs by applying two data augmentations to the same image. Designing effective data augmentation strategies is crucial for the success of contrastive learning. Inspired by the story of the blind men and the elephant, we introduce JointCrop and JointBlur. These methods generate more challenging positive pairs by leveraging the joint distribution of the two augmentation parameters, thereby enabling contrastive learning to acquire more effective feature representations. To the best of our knowledge, this is the first effort to explicitly incorporate the joint distribution of two data augmentation parameters into contrastive learning. As a plug-and-play framework without additional computational overhead, JointCrop and JointBlur enhance the performance of SimCLR, BYOL, MoCo v1, MoCo v2, MoCo v3, SimSiam, and Dino baselines with notable improvements.

Multimodal recommendation (MMRec) aims to integrate multimodal information of items to address the inherent data sparsity issue in collaborative-based recommendation. Traditional MMRec methods typically capture the structure-level item representations from the observed user behaviors within the multimodal graph, overlooking the potential impact of negative instances for personalized preference understanding. In light of the outstanding generative ability and step-by-step inference characteristic of Diffusion Models (DMs), we propose a Curriculum Conditioned Diffusion framework for Multimodal Recommendation (CCDRec), which precisely excavates the modality-aware distribution-level correlation among multi-modalities and elegantly integrates the reverse phase of DMs into negative sampling to highlight the most suitable instances in a curricular manner. Specifically, CCDRec proposes the Diffusion-controlled Multimodal Aligning module (DMA) to align multimodal knowledge with collaborative signals by capturing the fine-grained relationships among multi-modalities in the probabilistic distribution space. Furthermore, CCDRec designs the Negative-sensitive Diffusive Inferring module (NDI) to progressively synthesize the negative sample pool with diverse hardness to support the following knowledge-aware negative sampling. To gradually ramp up the training complexity, CCDRec further introduces a Curricular Negative Sampler (CNS) to tally the curriculum learning paradigm with the reverse phase of DMA, thereby adaptively sampling the gold-standard negative instances to enhance optimization. Extensive experiments on three datasets with four diverse backbones demonstrate the effectiveness and robustness of our CCDRec. The visualization analyses also clarify the underlying mechanism of our DMA in multimodal representation alignment and CNS in curricular negative discovery. The code and the corresponding dataset will be uploaded in the Appendix.

Ning Ding, Yulin Chen, Ganqu Cui, Xingtai Lv, Weilin Zhao, Kaiyan Zhang, Ruobing Xie, Bowen Zhou, Zhiyuan Liu, Maosong Sun. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Catastrophic forgetting remains a formidable obstacle to building an omniscient model in large language models (LLMs).Despite the pioneering research on task-level forgetting in LLM fine-tuning, there is scant focus on forgetting during pre-training.We systematically explored the existence and measurement of forgetting in pre-training, questioning traditional metrics such as perplexity (PPL) and introducing new metrics to better detect entity memory retention.Based on our revised assessment of forgetting metrics, we explored low-cost, straightforward methods to mitigate forgetting during the pre-training phase.In addition, we carefully analyzed the learning curves, offering insights into the dynamics of forgetting.Extensive evaluations and analyses on forgetting of pre-training could facilitate future research on LLMs.

Dwell time (DT) is a critical post-click metric for evaluating user preference in recommender systems, complementing the traditional click-through rate (CTR). Although multi-task learning is widely adopted to jointly optimize DT and CTR, we observe that multi-task models systematically collapse their DT predictions to the shortest and longest bins, under-predicting the moderate durations. We attribute this moderate-duration bin under-representation to over-reliance on the CTR-DT spurious correlation, and propose ORCA to address it with causal-decoupling. Specifically, ORCA explicitly models and subtracts CTR's negative transfer while preserving its positive transfer. We further introduce (i) feature-level counterfactual intervention, and (ii) a task-interaction module with instance inverse-weighting, weakening CTR-mediated effect and restoring direct DT semantics. ORCA is model-agnostic and easy to deploy. Experiments show an average 10.6% lift in DT metrics without harming CTR. Code is available at https://github.com/Chrissie-Law/ORCA-Mitigating-Over-Reliance-for-Multi-Task-Dwell-Time-Prediction-with-Causal-Decoupling.

Multimodal Large Language Models (MLLMs) hallucinate, resulting in an emerging topic of visual hallucination evaluation (VHE). This paper contributes a ChatGPT-Prompted visual hallucination evaluation Dataset (PhD) for objective VHE at a large scale. The essence of VHE is to ask an MLLM questions about specific images to assess its susceptibility to hallucination. Depending on what to ask (objects, attributes, sentiment, etc.) and how the questions are asked, we structure PhD along two dimensions, i.e. task and mode. Five visual recognition tasks, ranging from low-level (object/attribute recognition) to middle-level (sentiment/position recognition and counting), are considered. Besides a normal visual QA mode, which we term PhD-base, PhD also asks questions with specious context (PhD-sec) or with incorrect context (PhD-icc), or with AI-generated counter common sense images (PhD-ccs). We construct PhD by a ChatGPT-assisted semi-automated pipeline, encompassing four pivotal modules: task-specific hallucinatory item (hitem) selection, hitem-embedded question generation, specious/incorrect context generation, and counter-common-sense (CCS) image generation. With over 14k daily images, 750 CCS images and 102k VQA triplets in total, PhD reveals considerable variability in MLLMs’ performance across various modes and tasks, offering valuable insights into the nature of hallucination. As such, PhD stands as a potent tool not only for VHE but may also play a significant role in the refinement of MLLMs.

An Wang, Xingwu Sun, Ruobing Xie, Shuaipeng Li, Jiaqi Zhu, Zhen Yang, Pinxue Zhao, Weidong Han, Zhanhui Kang, Di Wang, Naoaki Okazaki, Cheng-zhong Xu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

The goal of transferable sequential recommendation (TSR) is to improve the performance of sequential recommenders in multiple target domains leveraging knowledge transferred from source domains. Most existing transferable sequential recommenders rely on item modality information but pay insufficient attention to the significant differences across domains (i.e., distinctive user behavior characteristics) in recommender systems. Besides, they are typically built on specific pre-trained model structures, limiting their model generalization to various different downstream domains and sequential recommenders in practical applications. This paper aims to address these issues by exploring the potential of mixture-of-LoRAs based domain adaption and an ensemble of sequential recommenders for learning generalizable and transferable domain-adaptive item representations. We propose MoL-TSR, a pre-trained plug-in mixture of low-rank adaption model for TSR. Specifically, we first devise a mixture-ofLoRAs based domain projector to produce domain-adaptive item representations based on both item modality and user behavior information for recommendation, enhancing the item discrimination and domain transfer ability. Then, we pre-train an ensemble of sequential recommendation models with contrastive learning to enhance the model-aspect generalization. As a general pretrained model, the mixture-of-LoRAs based domain projector in our MoL-TSR is a plug-in item semantic representation learning module, which is convenient to migrate into other transferable sequential recommenders to enhance item semantics better suited for recommendation. Extensive experiments demonstrate the effectiveness and model generalization of MoL-TSR, promising a practical solution for learning generalizable and transferable item semantic representations in recommendation. Our code is available at https://github.com/RUCAIBox/MoL-TSR.

The Text-to-Video Retrieval (T2VR) task aims to retrieve unlabeled videos by textual queries with the same semantic meanings. Recent CLIP-based approaches have explored two frameworks: Two-Tower versus Single-Tower framework, yet the former suffers from low effectiveness, while the latter suffers from low efficiency. In this study, we explore a new Hybrid-Tower framework that can hybridize the advantages of the Two-Tower and Single-Tower framework, achieving high effectiveness and efficiency simultaneously. We propose a novel hybrid method, Fine-grained Pseudo-query Interaction and Generation for T2VR, ie, PIG, which includes a new pseudo-query generator designed to generate a pseudo-query for each video. This enables the video feature and the textual features of pseudo-query to interact in a fine-grained manner, similar to the Single-Tower approaches to hold high effectiveness, even before the real textual query is received. Simultaneously, our method introduces no additional storage or computational overhead compared to the Two-Tower framework during the inference stage, thus maintaining high efficiency. Extensive experiments on five commonly used text-video retrieval benchmarks demonstrate that our method achieves a significant improvement over the baseline, with an increase of $1.6\% \sim 3.9\%$ in R@1. Furthermore, our method matches the efficiency of Two-Tower models while achieving near state-of-the-art performance, highlighting the advantages of the Hybrid-Tower framework.

The rapid advancement of large language models (LLMs) has paved the way for the development of highly capable autonomous agents. However, existing multi-agent frameworks often struggle with integrating diverse capable third-party agents due to reliance on agents defined within their own ecosystems. They also face challenges in simulating distributed environments, as most frameworks are limited to single-device setups. Furthermore, these frameworks often rely on hard-coded communication pipelines, limiting their adaptability to dynamic task requirements. Inspired by the concept of the Internet, we propose the Internet of Agents (IoA), a novel framework that addresses these limitations by providing a flexible and scalable platform for LLM-based multi-agent collaboration. IoA introduces an agent integration protocol, an instant-messaging-like architecture design, and dynamic mechanisms for agent teaming and conversation flow control. Through extensive experiments on general assistant tasks, embodied AI tasks, and retrieval-augmented generation benchmarks, we demonstrate that IoA consistently outperforms state-of-the-art baselines, showcasing its ability to facilitate effective collaboration among heterogeneous agents. IoA represents a step towards linking diverse agents in an Internet-like environment, where agents can seamlessly collaborate to achieve greater intelligence and capabilities. We will release our code to facilitate further research.

We introduce EURUS, a suite of large language models (LLMs) optimized for reasoning. Finetuned from Mistral-7B, Llama-3-8B, and Mixtral-8x22B, EURUS models achieve state-of-the-art results among open-source models on a diverse set of benchmarks covering mathematics, code generation, and logical reasoning problems. Notably, EURUX-8X22B outperforms GPT-3.5 Turbo in reasoning through a comprehensive benchmarking across 12 test sets covering five tasks. The strong performance of EURUS can be primarily attributed to ULTRAINTERACT, our newly-curated large-scale, high-quality training data dataset specifically designed for complex reasoning tasks. ULTRAINTERACT can be used in both supervised fine-tuning, preference learning, and reward modeling. It pairs each instruction with a preference tree consisting of (1) reasoning chains with diverse planning strategies in a unified format, (2) multi-turn interaction trajectories with the environment and the critique, and (3) pairwise positive and negative responses to facilitate preference learning. ULTRAINTERACT allows us to conduct an in-depth exploration of preference learning for reasoning tasks. Our investigation reveals that some well-established preference learning algorithms may be less suitable for reasoning tasks compared to their effectiveness in general conversations. The hypothesis is that in reasoning tasks, the space of correct answers is much smaller than that of incorrect ones, so it is necessary to explicitly increase the reward of chosen data. Therefore, in addition to increasing the reward margin as many preference learning algorithms do, the absolute values of positive responses’ rewards should be positive and may serve as a proxy for performance. Inspired by this, we derive a novel reward modeling objective and empirically that it leads to a stable reward modeling curve and better performance. Together with ULTRAINTERACT, we obtain a strong reward model.

Mixture-of-Experts (MoE) models mostly use a router to assign tokens to specific expert modules, activating only partial parameters and often outperforming dense models. We argue that the separation between the router’s decision-making and the experts’ execution is a critical yet overlooked issue, leading to suboptimal expert selection and learning. To address this, we propose Autonomy-of-Expert (AoE), a novel MoE paradigm in which experts autonomously select themselves to process inputs. AoE is based on the insight that an expert is aware of its own capacity to effectively process a token, an awareness reflected in the scale of its internal activations. In AoE, routers are removed; instead, experts pre-compute internal activations for inputs and are ranked based on their activation norms. Only the top-ranking experts proceed with the forward pass, while the others abort. The overhead of pre-computing activations is reduced through a low-rank weight factorization. This self-evaluating-then-partner-comparing approach ensures improved expert selection and effective learning. We pre-train language models having 700M up to 4B parameters, demonstrating that AoE outperforms traditional MoE models with comparable efficiency.

Low-precision training is considered an effective strategy for reducing both training and downstream inference costs. Previous scaling laws for precision mainly focus on integer quantization, which pay less attention to the constituents in floating-point (FP) quantization, and thus cannot well fit the LLM losses in this scenario. In contrast, while FP quantization training is more commonly implemented in production, it's research has been relatively superficial. In this paper, we thoroughly explore the effects of FP quantization targets, exponent bits, mantissa bits, and the calculation granularity of the scaling factor in FP quantization training performance of LLM models. In addition to an accurate FP quantization unified scaling law, we also provide valuable suggestions for the community: (1) Exponent bits contribute slightly more to the model performance than mantissa bits. We provide the optimal exponent-mantissa bit ratio for different bit numbers, which is available for future reference by hardware manufacturers; (2) We discover the formation of the critical data size in low-precision LLM training. Too much training data exceeding the critical data size will inversely bring in degradation of LLM performance; (3) The optimal FP quantization precision is directly proportional to the computational power, but within a wide computational power range. We estimate that the best cost-performance precision should lie between 4-8 bits.

In retrieval-augmented generation (RAG) systems, Large Language Models (LLMs) have been shown to be effective for re-ranking. However, existing research often prioritizes passage relevance over reliability, which can result in the incorporation of conflicting information and the generation of ambiguous responses. This issue becomes particularly pronounced when addressing inter-context knowledge conflicts, where candidate documents present contradictory information that may mislead the model. To mitigate this problem, we propose a novel cross-validation re-ranking technique designed specifically to resolve inter-context knowledge conflicts during the retrieval process. We also develop a new dataset, ContraPRT, to evaluate the ability of models to rank passages containing conflicting knowledge. Experimental results using GPT-4 and LlaMA3-70B demonstrate that our approach not only effectively filters out conflicting information but also ensures accurate passage rankings, thereby providing reliable supplementary knowledge for the generation module.

Personalized product search (PPS) aims to retrieve products relevant to the given query considering user preferences within their purchase histories. Since large language models (LLM) exhibit impressive potential in content understanding and reasoning, current methods explore to leverage LLM to comprehend the complicated relationships among user, query and product to improve the search performance of PPS. Despite the progress, LLM-based PPS solutions merely take textual contents into consideration, neglecting multimodal contents which play a critical role for product search. Motivated by this, we propose a novel framework, HMPPS, for Harnessing Multimodal large language models (MLLM) to deal with Personalized Product Search based on multimodal contents. Nevertheless, the redundancy and noise in PPS input stand for a great challenge to apply MLLM for PPS, which not only misleads MLLM to generate inaccurate search results but also increases the computation expense of MLLM. To deal with this problem, we additionally design two query-aware refinement modules for HMPPS: 1) a perspective-guided summarization module that generates refined product descriptions around core perspectives relevant to search query, reducing noise and redundancy within textual contents; and 2) a two-stage training paradigm that introduces search query for user history filtering based on multimodal representations, capturing precise user preferences and decreasing the inference cost. Extensive experiments are conducted on four public datasets to demonstrate the effectiveness of HMPPS. Furthermore, HMPPS is deployed on an online search system with billion-level daily active users and achieves an evident gain in A/B testing.

Recent advances in large vision-language models (LVLMs) have showcased their remarkable capabilities across a wide range of multimodal vision-language tasks. However, these models remain vulnerable to visual adversarial attacks, which can substantially compromise their performance. In this paper, we introduce F3, a novel adversarial purification framework that employs a counterintuitive ''fighting fire with fire'' strategy: intentionally introducing simple perturbations to adversarial examples to mitigate their harmful effects. Specifically, F3 leverages cross-modal attentions derived from randomly perturbed adversary examples as reference targets. By injecting noise into these adversarial examples, F3 effectively refines their attention, resulting in cleaner and more reliable model outputs. Remarkably, this seemingly paradoxical approach of employing noise to counteract adversarial attacks yields impressive purification results. Furthermore, F3 offers several distinct advantages: it is training-free and straightforward to implement, and exhibits significant computational efficiency improvements compared to existing purification methods. These attributes render F3 particularly suitable for large-scale industrial applications where both robust performance and operational efficiency are critical priorities. The code is available at https://github.com/btzyd/F3.

Large vision-language models (LVLMs) have demonstrated exceptional performance on complex multimodal tasks. However, they continue to suffer from significant hallucination issues, including object, attribute, and relational hallucinations. To accurately detect these hallucinations, we investigated the variations in cross-modal attention patterns between hallucination and non-hallucination states. Leveraging these distinctions, we developed a lightweight detector capable of identifying hallucinations. Our proposed method, Detecting Hallucinations by Cross-modal Attention Patterns (DHCP), is straightforward and does not require additional LVLM training or extra LVLM inference steps. Experimental results show that DHCP achieves remarkable performance in hallucination detection. By offering novel insights into the identification and analysis of hallucinations in LVLMs, DHCP contributes to advancing the reliability and trustworthiness of these models. The code is available at https://github.com/btzyd/DHCP.

Yudong Zhang, Ruobing Xie, Jiansheng Chen, Xingwu Sun, Zhanhui Kang, Yu Wang. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Realignment becomes necessary when a language model (LM) fails to meet expected performance. We propose a flexible realignment framework that supports quantitative control of alignment degree during training and inference. This framework incorporates **Training-time Realignment (TrRa)**, which efficiently realigns the reference model by leveraging the controllable fusion of logits from both the reference and already aligned models. For example, TrRa reduces token usage by **54.63%** on DeepSeek-R1-Distill-Qwen-1.5B without any performance degradation, outperforming DeepScaleR-1.5B’s **33.86%**. To complement TrRa during inference, we introduce a layer adapter that enables **smooth Inference-time Realignment (InRa)**. This adapter is initialized to perform an identity transformation at the bottom layer and is inserted preceding the original layers. During inference, input embeddings are simultaneously processed by the adapter and the original layer, followed by the remaining layers, and then controllably interpolated at the logit level. We upgraded DeepSeek-R1-Distill-Qwen-7B from a slow-thinking model to one that supports both fast and slow thinking, allowing flexible alignment control even **during inference**. By encouraging deeper reasoning, it even surpassed its original performance.

Large Language Models (LLMs) have emerged as a new paradigm for recommendation by converting interacted item history into language modeling. However, constrained by the limited context length of LLMs, existing approaches have to truncate item history in the prompt, focusing only on recent interactions and sacrificing the ability to model long-term history. To enable LLMs to model long histories, we pursue a concise embedding representation for items and sessions. In the LLM embedding space, we construct an item's embedding by aggregating its textual token embeddings; similarly, we construct a session's embedding by aggregating its item embeddings. While efficient, this way poses two challenges since it ignores the temporal significance of user interactions and LLMs do not natively interpret our custom embeddings. To overcome these, we propose PatchRec, a multi-grained patch training method consisting of two stages: (1) Patch Pre-training, which familiarizes LLMs with aggregated embeddings -- patches, and (2) Patch Fine-tuning, which enables LLMs to capture time-aware significance in interaction history. Extensive experiments show that PatchRec effectively models longer behavior histories with improved efficiency. This work facilitates the practical use of LLMs for modeling long behavior histories.

Negative sampling plays a crucial role for cross-domain recommendation as it provides contrastive signals to learn user preference. Existing methods usually select items with high predicted scores or popularity as hard negative samples to improve model training. However, such methods suffer from choosing false negative samples since items with high predicted scores or popularity could also indicate potential positive user preference. Although several studies devoted to discovering true negative samples, few of them leverage user cross-domain behaviors to alleviate the false negative issue. How to effectively mine and utilize hard negative samples to improve cross-domain recommendation remains an open question.

Recently, many efforts have been devoted to building effective sequential recommenders.Despite their effectiveness, these methods typically develop a single model to serve all users.However, our empirical studies reveal that different sequential encoders have intrinsic architectural biases and tend to focus on specific behavioral patterns, i.e., particular frequency range of user behavior sequences.For example, the Self-Attention module is essentially a low-pass filter, focusing on low-frequency information while neglecting the high-frequency details.This evidently limits their ability to capture diverse user patterns, leading to suboptimal recommendations.To tackle this problem, we present FamouSRec, a Frequency-Augmented Mixture-of-Heterogeneous-Experts Framework for personalized Recommendations.Our approach builds an MoEbased recommender system, integrating the strengths of various experts to achieve diversified user modeling.For developing the MoE framework, as the key to our approach, we instantiate experts with various model architectures, aiming to leverage their inherent architectural biases and capture diverse behavioral patterns.For selecting appropriate experts to serve individuals, we introduce a frequency-augmented router.It first identifies frequency components in user behavior sequences that are suited for expert encoding, and then conducts customized routing based on the informativeness of these components.Building on this framework, we further propose two novel contrastive tasks to enhance expert

As trustworthy AI continues to advance, the fairness issue in recommendations has received increasing attention. A recommender system is considered unfair when it produces unequal outcomes for different user groups based on user-sensitive attributes (e.g., age, gender). Some researchers have proposed data augmentation-based methods aiming at alleviating user-level unfairness by altering the skewed distribution of training data among various user groups. Despite yielding promising results, they often rely on fairness-related assumptions that may not align with reality, potentially reducing the data quality and negatively affecting model effectiveness. To tackle this issue, in this paper, we study how to implement high-quality data augmentation to improve recommendation fairness. Specifically, we proposeFairDgcl, a dynamic graph adversarial contrastive learning framework aiming at improving fairness in recommender system. First, FairDgcl develops an adversarial contrastive network with a view generator and a view discriminator to learn generating fair augmentation strategies in an adversarial style. Then, we propose two dynamic, learnable models to generate contrastive views within contrastive learning framework, which automatically fine-tune the augmentation strategies. Meanwhile, we theoretically show that FairDgcl can simultaneously generate enhanced representations that possess both fairness and accuracy. Lastly, comprehensive experiments conducted on four real-world datasets demonstrate the effectiveness of the proposed FairDgcl. The code can be found athttps://github.com/cwei01/FairDgcl.

With the thriving of the pre-trained language model (PLM) widely verified in various NLP tasks, pioneer efforts attempt to explore the possible cooperation of the general textual information in PLM with the personalized behavioral information in user historical behavior sequences to enhance sequential recommendation (SR). However, despite the commonalities of input format and task goal, there are huge gaps between the behavioral and textual information, which obstruct thoroughly modeling SR as language modeling via PLM. To bridge the gap, we propose a novel unified pre-trained language model enhanced sequential recommendation (UPSR) that thoroughly transfers the next item prediction task to a text generation task, aiming to build a unified pre-trained recommendation model for multi-domain recommendation tasks. We formally design five key indicators, namely naturalness, domain consistency, informativeness, noise and ambiguity, and text length, to guide the text \(\rightarrow\) item adaptation (selecting appropriate text to form the item textual representation) and behavior sequence \(\rightarrow\) text sequence adaptation (transferring the sequence of item textual representations into a text sequence) differently for pre-training and fine-tuning stages, which are essential but under-explored by previous works. In experiments, we conduct extensive evaluations on seven datasets with both supervised and zero-shot settings and achieve the overall best performance. Comprehensive model analyses also provide valuable insights for behavior modeling via PLM, shedding light on large pre-trained recommendation models. The source codes will be released in the future.

Classical sequential recommendation models generally adopt ID embeddings to store knowledge learned from user historical behaviors and represent items. However, these unique IDs are challenging to be transferred to new domains. With the thriving of pre-trained language model (PLM), some pioneer works adopt PLM for pre-trained recommendation, where modality information is considered universal across domains via PLM. Unfortunately, the behavioral information in ID embeddings is verified to currently dominate in recommendation compared to modality information and thus limits these models’ performance. In this work, we propose a novel ID-centric recommendation pre-training paradigm (IDP), which directly transfers informative ID embeddings learned in pre-training domains to item representations in new domains. Specifically, in pre-training stage, besides the ID-based sequential recommendation model, we also build a Cross-domain ID-matcher (CDIM) learned by both behavioral and modality information. In the tuning stage, modality information of new domain items is regarded as a cross-domain bridge built by CDIM. They first adopted to retrieve behaviorally and semantically similar items from pre-training domains using CDIM. Next, these retrieved items’ pre-trained ID embeddings are directly adopted to generate downstream new items’ embeddings. Through extensive experiments on real-world datasets, we demonstrate that our proposed model significantly outperforms all baselines.

In the past few decades, recommender systems have attracted much attention in both research and industry communities. Existing recommendation models mainly learn the underlying user preference from historical behavior data (typically in the forms of item IDs), and then estimate the user–item matching relationships for recommendations. Inspired by the recent progress on large language models (LLMs), we develop a different recommendation paradigm, considering recommendation as instruction following by LLMs. The key idea is that the needs of a user can be expressed in natural language descriptions (called instructions ), so that LLMs can understand and further execute the instruction for fulfilling the recommendation. For this purpose, we instruction tune the 3B Flan-T5-XL, to better adapt LLMs to recommender systems. We first design a general instruction format for describing the preference, intention, and task form of a user in natural language. Then we manually design 39 instruction templates and automatically generate large amounts of user-personalized instruction data with varying types of preferences and intentions. To demonstrate the effectiveness of our approach, we instantiate the instructions into several widely studied recommendation (or search) tasks, and conduct extensive experiments with real-world datasets. Experiment results show that our approach can outperform several competitive baselines, including the powerful GPT-3.5, on these evaluation tasks. Our approach sheds light on developing user-friendly recommender systems, in which users can freely communicate with the system and obtain accurate recommendations via natural language instructions.

Pioneering efforts have verified the effectiveness of the diffusion models in exploring the informative uncertainty for recommendation. Considering the difference between recommendation and image synthesis tasks, existing methods have undertaken tailored refinements to the diffusion and reverse process. However, these approaches typically use the highest-score item in corpus for user interest prediction, leading to the ignorance of the user's generalized preference contained within other items, thereby remaining constrained by the data sparsity issue. To address this issue, this paper presents a novel Plug-in Diffusion Model for Recommendation (PDRec) framework, which employs the diffusion model as a flexible plugin to jointly take full advantage of the diffusion-generating user preferences on all items. Specifically, PDRec first infers the users' dynamic preferences on all items via a time-interval diffusion model and proposes a Historical Behavior Reweighting (HBR) mechanism to identify the high-quality behaviors and suppress noisy behaviors. In addition to the observed items, PDRec proposes a Diffusion-based Positive Augmentation (DPA) strategy to leverage the top-ranked unobserved items as the potential positive samples, bringing in informative and diverse soft signals to alleviate data sparsity. To alleviate the false negative sampling issue, PDRec employs Noise-free Negative Sampling (NNS) to select stable negative samples for ensuring effective model optimization. Extensive experiments and analyses on four datasets have verified the superiority of the proposed PDRec over the state-of-the-art baselines and showcased the universality of PDRec as a flexible plugin for commonly-used sequential encoders in different recommendation scenarios. The code is available in https://github.com/hulkima/PDRec.

Xiaozhi Wang, Hao Peng, Yong Guan, Kaisheng Zeng, Jianhui Chen, Lei Hou, Xu Han, Yankai Lin, Zhiyuan Liu, Ruobing Xie, Jie Zhou, Juanzi Li. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Generative models have emerged as a promising utility to enhance recommender systems. It is essential to model both item content and user-item collaborative interactions in a unified generative framework for better recommendation. Although some existing large language model (LLM)-based methods contribute to fusing content information and collaborative signals, they fundamentally rely on textual language generation, which is not fully aligned with the recommendation task. How to integrate content knowledge and collaborative interaction signals in a generative framework tailored for item recommendation is still an open research challenge.

This paper studies vision-language (V&L) pre-training for deep cross-modal representations. Recently, pre-trained V&L models have shown great success in V&L tasks. However, most existing models apply multi-modal encoders to encode the image and text, at the cost of high training complexity because of the input sequence length. In addition, they suffer from noisy training corpora caused by V&L mismatching. In this work, we propose a lightweight vision-language pre-training (LightVLP) for efficient and effective V&L pre-training. First, we design a new V&L framework with two autoencoders. Each autoencoder involves an encoder, which only takes in unmasked tokens (removes masked ones), as well as a lightweight decoder that reconstructs the masked tokens. Besides, we mask and remove large portions of input tokens to accelerate the training. Moreover, we propose a gated interaction mechanism to cope with noise in aligned image-text pairs. As for a matched image-text pair, the model tends to apply cross-modal representations for reconstructions. By contrast, for an unmatched pair, the model conducts reconstructions mainly using uni-modal representations. Benefiting from the above-mentioned designs, our base model shows competitive results compared to ALBEF while saving 44% FLOPs. Further, we compare our large model with ALBEF under the setting of similar FLOPs on six datasets and show the superiority of LightVLP. In particular, our model achieves 2.2% R@1 gains on COCO Text Retrieval and 1.1% on refCOCO+.

Yiju Guo, Ganqu Cui, Lifan Yuan, Ning Ding, Zexu Sun, Bowen Sun, Huimin Chen, Ruobing Xie, Jie Zhou, Yankai Lin, Zhiyuan Liu, Maosong Sun. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

This paper focuses on detecting clickbait posts on the Web.These posts often use eye-catching disinformation in mixed modalities to mislead users to click for profit.That affects the user experience and thus would be blocked by content provider.To escape detection, malicious creators use tricks to add some irrelevant nonbait content into bait posts, dressing them up as legal to fool the detector.This content often has biased relations with non-bait labels, yet traditional detectors tend to make predictions based on simple co-occurrence rather than grasping inherent factors that lead to malicious behavior.This spurious bias would easily cause misjudgments.To address this problem, we propose a new debiased method based on causal inference.We first employ a set of features in multiple modalities to characterize the posts.Considering these features are often mixed up with unknown biases, we then disentangle three kinds of latent factors from them, including the invariant factor that indicates intrinsic bait intention; the causal factor which reflects deceptive patterns in a certain scenario, and non-causal noise.By eliminating the noise that causes bias, we can use invariant and causal factors to build a robust model with good generalization ability.Experiments on three popular datasets show the effectiveness of our approach.

Autonomous agents empowered by Large Language Models (LLMs) have undergone significant improvements, enabling them to generalize across a broad spectrum of tasks. However, in real-world scenarios, cooperation among individuals is often required to enhance the efficiency and effectiveness of task accomplishment. Hence, inspired by human group dynamics, we propose a multi-agent framework AgentVerse that can effectively orchestrate a collaborative group of expert agents as a greater-than-the-sum-of-its-parts system. Our experiments demonstrate that AgentVerse can proficiently deploy multi-agent groups that outperform a single agent. Extensive experiments on text understanding, reasoning, coding, tool utilization, and embodied AI confirm the effectiveness of AgentVerse. Moreover, our analysis of agent interactions within AgentVerse reveals the emergence of specific collaborative behaviors, contributing to heightened group efficiency. We will release our codebase, AgentVerse, to further facilitate multi-agent research.

Despite the advancements of open-source large language models (LLMs), e.g., LLaMA, they remain significantly limited in tool-use capabilities, i.e., using external tools (APIs) to fulfill human instructions. The reason is that current instruction tuning largely focuses on basic language tasks but ignores the tool-use domain. This is in contrast to the excellent tool-use capabilities of state-of-the-art (SOTA) closed-source LLMs, e.g., ChatGPT. To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation. We first present ToolBench, an instruction-tuning dataset for tool use, which is constructed automatically using ChatGPT. Specifically, the construction can be divided into three stages: (i) API collection: we collect 16,464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub; (ii) instruction generation: we prompt ChatGPT to generate diverse instructions involving these APIs, covering both single-tool and multi-tool scenarios; (iii) solution path annotation: we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction. To enhance the reasoning capabilities of LLMs, we develop a novel depth-first search-based decision tree algorithm. It enables LLMs to evaluate multiple reasoning traces and expand the search space. Moreover, to evaluate the tool-use capabilities of LLMs, we develop an automatic evaluator: ToolEval. Based on ToolBench, we fine-tune LLaMA to obtain an LLM ToolLLaMA, and equip it with a neural API retriever to recommend appropriate APIs for each instruction. Experiments show that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT. Our ToolLLaMA also demonstrates strong zero-shot generalization ability in an out-of-distribution tool-use dataset: APIBench.

Learning from human feedback has become a pivot technique in aligning large language models (LLMs) with human preferences. However, acquiring vast and premium human feedback is bottlenecked by time, labor, and human capability, resulting in small sizes or limited topics of current datasets. This further hinders feedback learning as well as alignment research within the open-source community. To address this issue, we explore how to go beyond human feedback and collect high-quality AI feedback automatically for a scalable alternative. Specifically, we identify scale and diversity as the key factors for feedback data to take effect. Accordingly, we first broaden instructions and responses in both amount and breadth to encompass a wider range of user-assistant interactions. Then, we meticulously apply a series of techniques to mitigate annotation biases for more reliable AI feedback. We finally present UltraFeedback, a large-scale, high-quality, and diversified AI feedback dataset, which contains over 1 million GPT-4 feedback for 250k user-assistant conversations from various aspects. Built upon UltraFeedback, we align a LLaMA-based model by best-of-$n$ sampling and reinforcement learning, demonstrating its exceptional performance on chat benchmarks. Our work validates the effectiveness of scaled AI feedback data in constructing strong open-source chat language models, serving as a solid foundation for future feedback learning research.

Pre-trained Transformers inherently possess the characteristic of sparse activation, where only a small fraction of the neurons are activated for each token. While sparse activation has been explored through post-training methods, its potential in pre-training remains untapped. In this work, we first study how activation properties change during pre-training. Our examination reveals that Transformers exhibit sparse activation throughout the majority of the pre-training process while the activation correlation keeps evolving as training progresses. Leveraging this observation, we propose Switchable Sparse-Dense Learning (SSD). SSD adaptively switches between the Mixtures-of-Experts (MoE) based sparse training and the conventional dense training during the pre-training process, leveraging the efficiency of sparse training and avoiding the static activation correlation of sparse training. Compared to dense training, SSD achieves comparable performance with identical model size and reduces pre-training costs. Moreover, the models trained with SSD can be directly used as MoE models for sparse inference and achieve the same performance as dense models with up to $2\times$ faster inference speed. Codes are available at https://github.com/thunlp/moefication.

The graph-based recommendation has achieved great success in recent years. However, most existing graph-based recommendations focus on capturing user preference based on positive edges/feedback, while ignoring negative edges/feedback (e.g., dislike, low rating) that widely exist in real-world recommender systems. How to utilize negative feedback in graph-based recommendations still remains underexplored. In this study, we first conducted a comprehensive experimental analysis and found that (1) existing graph neural networks are not well-suited for modeling negative feedback, which acts as a high-frequency signal in a user-item graph. (2) The graph-based recommendation suffers from the representation degeneration problem. Based on the two observations, we propose a novel model that models positive and negative feedback from a frequency filter perspective called Dual-frequency Graph Neural Network for Sign-aware Recommendation (DFGNN). Specifically, in DFGNN, the designed dual-frequency graph filter (DGF) captures both low-frequency and high-frequency signals that contain positive and negative feedback. Furthermore, the proposed signed graph regularization is applied to maintain the user/item embedding uniform in the embedding space to alleviate the representation degeneration problem. Additionally, we conduct extensive experiments on real-world datasets and demonstrate the effectiveness of the proposed model. Codes of our model will be released upon acceptance.

Multi-modal recommender systems (MRSs) are pivotal in diverse online web platforms and have garnered considerable attention in recent years. However, previous studies overlook the challenges of (1)noisy multi-modal content, (2) noisy user feedback, and (3) aligning multi-modal content and user feedback. To tackle these challenges, we propose Denoising and Aligning Multi-modal Recommender System (DA-MRS). To mitigate noise in multi-modal content, DA-MRS first constructs item-item graphs determined by consistent content similarity across modalities. To denoise user feedback, DA-MRS associates the probability of observed feedback with multi-modal content and devises a denoised BPR loss. Furthermore, DA-MRS implements Alignment guided by User preference to enhance task-specific item representation and Alignment guided by graded Item relations to provide finer-grained alignment. Extensive experiments verify that DA-MRS is a plug-and-play framework and achieves significant and consistent improvements across various datasets, backbone models, and noisy scenarios.

Large Vision-Language Models (LVLMs) have demonstrated their powerful multimodal capabilities. However, they also face serious safety problems, as adversaries can induce robustness issues in LVLMs through the use of well-designed adversarial examples. Therefore, LVLMs are in urgent need of detection tools for adversarial examples to prevent incorrect responses. In this work, we first discover that LVLMs exhibit regular attention patterns for clean images when presented with probe questions. We propose an unconventional method named PIP, which utilizes the attention patterns of one randomly selected irrelevant probe question (e.g., "Is there a clock''') to distinguish adversarial examples from clean examples. Regardless of the image to be tested and its corresponding question, PIP only needs to perform one additional inference of the image to be tested and the probe question, and then achieves successful detection of adversarial examples. Even under black-box attacks and open dataset scenarios, our PIP, coupled with a simple SVM, still achieves more than 98% recall and a precision of over 90%. Our PIP is the first attempt to detect adversarial attacks on LVLMs via simple irrelevant probe questions, shedding light on deeper understanding and introspection within LVLMs. The code is available at https://github.com/btzyd/pip.

Sequential recommendation (SR) has seen significant advancements with the help of Pre-trained Language Models (PLMs). Some PLM-based SR models directly use PLM to encode user historical behavior’s text sequences to learn user representations, while there is seldom an in-depth exploration of the capability and suitability of PLM in behavior sequence modeling. In this work, we first conduct extensive model analyses between PLMs and PLM-based SR models, discovering great underutilization and parameter redundancy of PLMs in behavior sequence modeling. Inspired by this, we explore different lightweight usages of PLMs in SR, aiming to maximally stimulate the ability of PLMs for SR while satisfying the efficiency and usability demands of practical systems. We discover that adopting behavior-tuned PLMs for item initializations of conventional ID-based SR models is the most economical framework of PLM-based SR, which would not bring in any additional inference cost but could achieve a dramatic performance boost compared with the original version. Extensive experiments on five datasets show that our simple and universal framework leads to significant improvement compared to classical SR and SOTA PLM-based SR models without additional inference costs. Our code can be found in https://github.com/777pomingzi/Rethinking-PLM-in-RS.

Conversational recommender systems (CRSs) are able to elicit user preferences through multi-turn dialogues. They typically incorporate external knowledge and pre-trained language models to capture the dialogue context. Most CRS approaches, trained on benchmark datasets, assume that the standard items and responses in these benchmarks are optimal. However, they overlook that users may express negative emotions with the standard items and may not feel emotionally engaged by the standard responses. This issue leads to a tendency to replicate the logic of recommenders in the dataset instead of aligning with user needs. To remedy this misalignment, we introduce empathy within a CRS. With empathy we refer to a system’s ability to capture and express emotions. We propose an empathetic conversational recommender (ECR) framework.

In recent years, we have witnessed significant improvements in pre-trained language models (PLM) brought about by the scaling of parameter sizes and data amounts. However, this also brings high computational and storage costs. In this paper, we present a new direction to improve PLMs without scaling parameters and data: adopting a geometric feature space that is more suitable for encoding the intrinsic structured features of text. Although text is generally considered unstructured data, it possesses rich intrinsic structured features that signify syntactic and semantic relationships. Leveraging these structured features is vital for text understanding. Given that structured features are better encoded in hyperbolic spaces than in the Euclidean spaces used by conventional PLMs, we propose that PLMs should operate entirely within hyperbolic spaces. Our experiments demonstrate the superiority of hyperbolic PLMs over Euclidean PLMs across a wide variety of tasks, using the same parameter and data settings. This suggests that altering the geometry of model representation is a promising direction for model enhancement. The code is released athttps://github.com/thunlp/hyperbolic_llm

Recommendation systems, widely adopted in social networks, personalize user experiences through advanced technologies such as Reinforcement Learning (RL), known for producing high-performance, list- wise recommendations. However, RL-based recommendation methods exhibit biases, specifically: 1) Online bias, which stems from a complex real-worldonline policycomposed of various rules and models rather than a single policy; 2) Training bias, a distributional shift resulting from differences between thetarget policyand thebehavior policy. To address these issues, we introduce a novel framework named Contrastive Imitation Proximal Policy Optimization (CIPPO) for recommendation based on RL. This approach leverages extensively labeled feedback data and incorporates a Masked Imitation Network (MIN) that closely emulates the online policy, thus reducing discrepancies between online and offline environments. Additionally, the clipping function in Proximal Policy Optimization, combined with a specially designed contrastive module, effectively reduces the distributional shift between the behavior and target policies. We conduct offline and online experiments to show the improvements of CIPPO, providing details including ablation tests and parameter analysis to validate the effectiveness and robustness. CIPPO gains 12.79% on ACN and in WeChat Top Stories, a large media platform with over 50 million users.

Pre-training models have shown their power in sequential recommendation. Recently, prompt has been widely explored and verified for tuning after pre-training in NLP, which helps to more effectively and parameter-efficiently extract useful knowledge from pre-training models for downstream tasks, especially in cold-start scenarios. However, it is challenging to bring prompt-tuning from NLP to recommendation, since the tokens of recommendation (i.e., items) are million-level and do not have concrete explainable semantics, and the sequence modeling in recommendation should be personalized. In this work, we first introduce prompt to recommendation models and propose a novel Personalized prompt-based recommendation (PPR) framework for cold-start recommendation. Specifically, we build personalized soft prompt via a prompt generator based on user profiles, and enable a sufficient training on prompts via a new prompt-oriented contrastive learning. PPR is effective, parameter-efficient, and universal in various tasks. In both few-shot and zero-shot recommendation tasks, PPR models achieve significant improvements over baselines in three large-scale datasets. We also verify PPR's universality in adopting different recommendation models as the backbone. Finally, we explore and confirm the capability of PPR on other tasks such as cross-domain recommendation and user profile prediction, shedding lights on the promising future directions of better using large-scale pre-trained recommendation models.

The emergence of Graph Neural Networks (GNNs) has greatly advanced the development of recommendation systems. Recently, many researchers have leveraged GNN-based models to learn fair representations for users and items. However, current GNN-based models suffer from biased user–item interaction data, which negatively impacts recommendation fairness. Although there have been several studies employing adversarial learning to mitigate this issue in recommendation systems, they mostly focus on modifying the model training approach with fairness regularization and neglect direct intervention of biased interaction. In contrast to these models, this article introduces a novel perspective by directly intervening in observed interactions to generate a counterfactual graph (called FairGap) that is not influenced by sensitive node attributes, enabling us to learn fair representations for users and items easily. We design FairGap to answer the key counterfactual question: “Would interactions with an item remain unchanged if a user’s sensitive attributes were concealed?”. We also provide theoretical proofs to show that our learning strategy via the counterfactual graph is unbiased in expectation. Moreover, we propose a fairness-enhancing mechanism to continuously improve user fairness in the graph-based recommendation. Extensive experimental results against state-of-the-art competitors and base models on three real-world datasets validate the effectiveness of our proposed model.

Cross-domain recommendation (CDR) aims at leveraging the correlation of users’ behaviors in both the source and target domains to improve the user preference modeling in the target domain. Conventional CDR methods typically explore the dual-relations between the source and target domains’ behaviors. However, this may ignore the informative mixed behaviors that naturally reflect the user’s global preference. To address this issue, we present a novel framework, termed triple sequence learning for cross-domain recommendation (Tri-CDR), which jointly models the source, target, and mixed behavior sequences to highlight the global and target preference and precisely model the triple correlation in CDR. Specifically, Tri-CDR independently models the hidden representations for the triple behavior sequences and proposes a triple cross-domain attention (TCA) method to emphasize the informative knowledge related to both user’s global and target-domain preference. To comprehensively explore the cross-domain correlations, we design a triple contrastive learning (TCL) strategy that simultaneously considers the coarse-grained similarities and fine-grained distinctions among the triple sequences, ensuring the alignment while preserving information diversity in multi-domain. We conduct extensive experiments and analyses on six cross-domain settings. The significant improvements of Tri-CDR with different sequential encoders verify its effectiveness and universality. The source code is available at https://github.com/hulkima/Tri-CDR .

Large-scale commonsense knowledge bases empower a broad range of AI applications, where the automatic extraction of commonsense knowledge (CKE) is a fundamental and challenging problem. CKE from text is known for suffering from the inherent sparsity and reporting bias of commonsense in text. Visual perception, on the other hand, contains rich commonsense knowledge about real-world entities, e.g., (person, can_hold, bottle), which can serve as promising sources for acquiring grounded commonsense knowledge. In this work, we present CLEVER, which formulates CKE as a distantly supervised multi-instance learning problem, where models learn to summarize commonsense relations from a bag of images about an entity pair without any human annotation on image instances. To address the problem, CLEVER leverages vision-language pre-training models for deep understanding of each image in the bag, and selects informative instances from the bag to summarize commonsense entity relations via a novel contrastive attention mechanism. Comprehensive experimental results in held-out and human evaluation show that CLEVER can extract commonsense knowledge in promising quality, outperforming pre-trained language model-based methods by 3.9 AUC and 6.4 mAUC points. The predicted commonsense scores show strong correlation with human judgment with a 0.78 Spearman coefficient. Moreover, the extracted commonsense can also be grounded into images with reasonable interpretability. The data and codes can be obtained at https://github.com/thunlp/CLEVER.

To overcome the overparameterized problem in Pre-trained Language Models (PLMs), pruning is widely used as a simple and straightforward compression method by directly removing unimportant weights.Previous first-order methods successfully compress PLMs to extremely high sparsity with little performance drop.These methods, such as movement pruning, use first-order information to prune PLMs while fine-tuning the remaining weights.In this work, we argue fine-tuning is redundant for first-order pruning, since first-order pruning is sufficient to converge PLMs to downstream tasks without fine-tuning.Under this motivation, we propose Static Model Pruning (SMP), which only uses first-order pruning to adapt PLMs to downstream tasks while achieving the target sparsity level.In addition, we also design a new masking function and training objective to further improve SMP.Extensive experiments at various sparsity levels show SMP has significant improvements over firstorder and zero-order methods.Unlike previous first-order methods, SMP is also applicable to low sparsity and outperforms zero-order methods.Meanwhile, SMP is more parameter efficient than other methods due to it does not require fine-tuning.Our code is available at https://github.com/kongds/SMP.

Yujia Qin, Zihan Cai, Dian Jin, Lan Yan, Shihao Liang, Kunlun Zhu, Yankai Lin, Xu Han, Ning Ding, Huadong Wang, Ruobing Xie, Fanchao Qi, Zhiyuan Liu, Maosong Sun, Jie Zhou. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Recommendation usually focuses on immediate accuracy metrics like CTR as training objectives. User retention rate, which reflects the percentage of today’s users that will return to the recommender system in the next few days, should be paid more attention to in real-world systems. User retention is the most intuitive and accurate reflection of user long-term satisfaction. However, most existing recommender systems are not focused on user retention-related objectives, since their complexity and uncertainty make it extremely hard to discover why a user will or will not return to a system and which behaviors affect user retention. In this work, we conduct a series of preliminary explorations on discovering and making full use of the reasons for user retention in recommendation. Specifically, we make a first attempt to design a rationale contrastive multi-instance learning framework to explore the rationale and improve the interpretability of user retention. Extensive offline and online evaluations with detailed analyses of a real-world recommender system verify the effectiveness of our user retention modeling. We further reveal the real-world interpretable factors of user retention from both user surveys and explicit negative feedback quantitative analyses to facilitate future model designs. The source codes are released at https://github.com/dinry/IURO.

Negative Sampling in recommendation aims to capture informative negative instances for the sparse user-item interactions to improve the performance. Conventional negative sampling methods tend to select informative hard negative samples (HNS) besides the default random samples. However, these hard negative sampling methods usually struggle with false hard negative samples (FHNS), which happens when a user-item interaction has not been observed yet and is picked as a negative sample, while the user will actually interact with this item once exposed to it. Such FHNS issues may seriously confuse the model training, while most conventional hard negative sampling methods do not systematically explore and distinguish FHNS from HNS. To address this issue, we propose a novel model-agnostic Real Hard Negative Sampling (RealHNS) framework specially for cross-domain recommendation (CDR), which aims to discover the false and refine the real from all HNS via both general and cross-domain real hard negative sample selectors. For the general part, we conduct the coarse- and fine-grained real HNS selectors sequentially, armed with a dynamic item-based FHNS filter to find high-quality HNS. For the cross-domain part, we further design a new cross-domain HNS for alleviating negative transfer in CDR and discover its corresponding FHNS via a dynamic user-based FHNS filter to keep its power. We conduct experiments on four datasets based on three representative hard negative sampling methods, along with extensive model analyses, ablation studies, and universality analyses. The consistent improvements indicate the effectiveness, robustness, and universality of RealHNS, which is also easy-to-deploy in real-world systems as a plug-and-play strategy. The source code is avaliable in https://github.com/hulkima/RealHNS.

Next-basket Recommendation (NBR) refers to the task of predicting a set of items that a user will purchase in the next basket. However, most of existing works merely focus on the correlations between user preferences and predicted items, ignoring the essential correlations among items in the next basket, which often results in over-homogenization of predicted items. In this work, we presents a Generative next-basket Recommendation model (GenRec), a novel NBR paradigm that generates the recommended items one by one to form the next basket via an autoregressive decoder. This generative NBR paradigm contributes to capturing and considering item correlations inside each baskets in both training and serving. Moreover, we jointly consider user’s both item- and basket-level contextual information to better capture user’s multi-granularity preferences. Extensive experiments on three real-world datasets demonstrate the effectiveness of our model.

Recently, a series of pioneer studies have shown the potency of pre-trained models in sequential recommendation, illuminating the path of building an omniscient unified pre-trained recommendation model for different downstream recommendation tasks. Despite these advancements, the vulnerabilities of classical recommender systems also exist in pre-trained recommendation in a new form, while the security of pre-trained recommendation model is still unexplored, which may threaten its widely practical applications. In this study, we propose a novel framework for backdoor attacking in pre-trained recommendation. We demonstrate the provider of the pre-trained model can easily insert a backdoor in pre-training, thereby increasing the exposure rates of target items to target user groups. Specifically, we design two novel and effective backdoor attacks: basic replacement and prompt-enhanced, under various recommendation pre-training usage scenarios. Experimental results on real-world datasets show that our proposed attack strategies significantly improve the exposure rates of target items to target users by hundreds of times in comparison to the clean model. The source codes are released in https://github.com/wyqing20/APRec.

Cross-domain recommendation (CDR) aims to leverage the source domain information to provide better recommendation for the target domain, which is widely adopted in recommender systems to alleviate the data sparsity and cold-start problems. However, existing CDR methods mostly focus on designing effective model architectures to transfer the source domain knowledge, ignoring the behavior-level effect during the loss optimization process, where behaviors regarding different aspects in the source domain may have different importance for the CDR model optimization. The ignorance of the behavior-level effect will cause the carefully designed model architectures ending up with sub-optimal parameters, which limits the recommendation performance. To tackle the problem, we propose a generic behavioral importance-aware optimization framework for cross-domain recommendation (BIAO). Specifically, we propose a behavioral perceptron which predicts the importance of each source behavior according to the corresponding item’s global impact and local user-specific impact. The joint optimization process of the CDR model and the behavioral perceptron is formulated as a bi-level optimization problem. In the lower optimization, only the CDR model is updated with weighted source behavior loss and the target domain loss, while in the upper optimization, the behavioral perceptron is updated with implicit gradient from a developing dataset obtained through the proposed reorder-and-reuse strategy. Extensive experiments show that our proposed optimization framework consistently improves the performance of different cross-domain recommendation models in 7 cross-domain scenarios, demonstrating that our method can serve as a generic and powerful tool for cross-domain recommendation1.

Knowledge graph entity typing (KGET) aims to infer missing entity typing instances in KGs, which is a significant subtask of KG completion. Despite of its progress, however, it still faces two non-trivial challenges: (i) most existing KGET methods extract features by encoding the existing entity typing tuples, while ignoring rich relational knowledge. (ii) they typically treat each entity typing tuple in KGs independently, and thus inevitably fail to take account of the inherent and valuable neighborhood information surrounding a tuple. To address these challenges, we build a novel Heterogeneous Relational Graph (HRG), and propose a Multiplex Relational Graph Attention Networks (MRGAT) to learn on HRG, and then utilize a Connecting Embeddings model (ConnectE) to make entity type inference. Specifically, the framework contains three components. Firstly, to effectively integrate the entity typing tuples and entity relation triples in KGs, we construct a HRG that consists of three semantic subgraphs. Secondly, we employ MRGAT to learn embeddings on HRG. In MRGAT, each subgraph of HRG is fed to its corresponding model that is capable of capturing neighborhood information. Finally, given the learned embeddings, we make entity type prediction by ConnectE. Experimental results validate the superiority of our model against various state-of-the-art baselines.

Semantically connecting users and items is a fundamental problem for the matching stage of an industrial recommender system. Recent advances in this topic are based on multi-channel retrieval to efficiently measure users' interest on items from the massive candidate pool. However, existing studies are primarily built upon pre-defined retrieval channels, including User-CF (U2U), Item-CF (I2I), and Embedding-based Retrieval (U2I), thus access to the limited correlation between users and items which solely entail from partial information of latent interactions. In this paper, we propose a model-agnostic integrated cross-channel (MIC) approach for the large-scale recommendation, which maximally leverages the inherent multi-channel mutual information to enhance the matching performance. Specifically, MIC robustly models correlation within user-item, user-user, and item-item from latent interactions in a universal schema. For each channel, MIC naturally aligns pairs with semantic similarity and distinguishes them otherwise with more uniform anisotropic representation space. While state-of-the-art methods require specific architectural design, MIC intuitively considers them as a whole by enabling the complete information flow among users and items. Thus MIC can be easily plugged into other retrieval recommender systems. Extensive experiments show that our MIC helps several state-of-the-art models boost their performance on four real-world benchmarks. The satisfactory deployment of the proposed MIC on industrial online services empirically proves its scalability and flexibility.

Personalized recommendation aims to provide appropriate items according to user preferences mainly from their behaviors. Excessive homogeneous user behaviors on similar items will lead to fatigue, which may decrease user activeness and degrade user experience. However, existing models seldom consider user fatigue in recommender systems. In this work, we propose a novel multi-granularity fatigue, modeling user fatigue from coarse to fine. Specifically, we focus on the recommendation feed scenario, where the underexplored global session fatigue and coarse-grained taxonomy fatigue have large impacts. We conduct extensive analyses to demonstrate the characteristics and influence of different types of fatigues in real-world recommender systems. In experiments, we verify the effectiveness of multi-granularity fatigue in both offline and online evaluations. Currently, the fatigue-enhanced model has also been deployed on a widely-used recommendation system of WeChat.

Cross-domain recommendation (CDR) aims to provide better recommendation results in the target domain with the help of the source domain, which is widely used and explored in real-world systems. However, CDR in the matching (i.e., candidate generation) module struggles with the data sparsity and popularity bias issues in both representation learning and knowledge transfer. In this work, we propose a novel Contrastive Cross-Domain Recommendation (CCDR) framework for CDR in matching. Specifically, we build a huge diversified preference network to capture multiple information reflecting user diverse interests, and design an intra-domain contrastive learning (intra-CL) and three inter-domain contrastive learning (inter-CL) tasks for better representation learning and knowledge transfer. The intra-CL enables more effective and balanced training inside the target domain via a graph augmentation, while the inter-CL builds different types of cross-domain interactions from user, taxonomy, and neighbor aspects. In experiments, CCDR achieves significant improvements on both offline and online evaluations in a real-world system. Currently, we have deployed our CCDR on WeChat Top Stories, affecting plenty of users. The source code is in https://github.com/lqfarmer/CCDR.

Conversational recommender systems (CRS) aim to provide highquality recommendations in conversations. However, most conventional CRS models mainly focus on the dialogue understanding of the current session, ignoring other rich multi-aspect information of the central subjects (i.e., users) in recommendation. In this work, we highlight that the user's historical dialogue sessions and look-alike users are essential sources of user preferences besides the current dialogue session in CRS. To systematically model the multi-aspect information, we propose a User-Centric Conversational Recommendation (UCCR) model, which returns to the essence of user preference learning in CRS tasks. Specifically, we propose a historical session learner to capture users' multi-view preferences from knowledge, semantic, and consuming views as supplements to the current preference signals. A multi-view preference mapper is conducted to learn the intrinsic correlations among different views in current and historical sessions via self-supervised objectives. We also design a temporal look-alike user selector to understand users via their similar users. The learned multi-aspect multi-view user preferences are then used for the recommendation and dialogue generation. In experiments, we conduct comprehensive evaluations on both Chinese and English CRS datasets. The significant improvements over competitive models in both recommendation and dialogue generation verify the superiority of UCCR.

Recommendation fairness has attracted great attention recently. In real-world systems, users usually have multiple sensitive attributes (e.g. age, gender, and occupation), and users may not want their recommendation results influenced by those attributes. Moreover, which of and when these user attributes should be considered in fairness-aware modeling should depend on users' specific demands. In this work, we define the selective fairness task, where users can flexibly choose which sensitive attributes should the recommendation model be bias-free. We propose a novel parameter-efficient prompt-based fairness-aware recommendation (PFRec) framework, which relies on attribute-specific prompt-based bias eliminators with adversarial training, enabling selective fairness with different attribute combinations on sequential recommendation. Both task-specific and user-specific prompts are considered. We conduct extensive evaluations to verify PFRec's superiority in selective fairness. The source codes are released in \urlhttps://github.com/wyqing20/PFRec.

An effective online recommendation system should jointly capture users' long-term and short-term preferences in both users' internal behaviors (from the target recommendation task) and external behaviors (from other tasks). However, it is extremely challenging to conduct fast adaptations to real-time new trends while making full use of all historical behaviors in large-scale systems, due to the real-world limitations in real-time training efficiency and external behavior acquisition. To address these practical challenges, we propose a novel Long Short-Term Temporal Meta-learning framework (LSTTM) for online recommendation. It arranges user multi-source behaviors in a global long-term graph and an internal short-term graph, and conducts different GAT-based aggregators and training strategies to learn user short-term and long-term preferences separately. To timely capture users' real-time interests, we propose a temporal meta-learning method based on MAML under an asynchronous optimization strategy for fast adaptation, which regards recommendations at different time periods as different tasks. In experiments, LSTTM achieves significant improvements on both offline and online evaluations. It has been deployed on a widely-used online recommendation system named WeChat Top Stories, affecting millions of users.

Personalized recommendation often relies on user historical behaviors to provide items for users. It is intuitive that future information also contains essential messages as supplements to user historical behaviors. However, we cannot directly encode future information into models, since we are unable to get future information in online serving. In this work, we propose a novel adversarial future encoding (AFE) framework to make full use of informative future features in different types of recommendation models. Specifically, AFE contains a future-aware discriminator and a generator. The future-aware discriminator takes both common features and future features as inputs, working as a recommendation prophet to judge user-item pairs. In contrast, the generator is considered as a challenger, which generates items with only common features, aiming to confuse the future-aware prophet. The future-aware discriminator can inspire the generator (to be deployed online) to produce better results. We further conduct a multi-factor optimization to enable a fast and stable model convergence via the direct learning and knowledge distillation losses. Moreover, we have adopted AFE on both a list-wise RL-based ranking model and a point-wise ranking model to verify its universality. In experiments, we conduct sufficient evaluations on two large-scale datasets, achieving significant improvements on both offline and online evaluations. Currently, we have deployed AFE on a real-world system, affecting millions of users. The source code is in https://github.com/modriczhang/AFE.

Cold-start problem is still a very challenging problem in recommender systems. Fortunately, the interactions of the cold-start users in the auxiliary source domain can help cold-start recommendations in the target domain. How to transfer user's preferences from the source domain to the target domain, is the key issue in Cross-domain Recommendation (CDR) which is a promising solution to deal with the cold-start problem. Most existing methods model a common preference bridge to transfer preferences for all users. Intuitively, since preferences vary from user to user, the preference bridges of different users should be different. Along this line, we propose a novel framework named Personalized Transfer of User Preferences for Cross-domain Recommendation (PTUPCDR). Specifically, a meta network fed with users' characteristic embeddings is learned to generate personalized bridge functions to achieve personalized transfer of preferences for each user. To learn the meta network stably, we employ a task-oriented optimization procedure. With the meta-generated personalized bridge function, the user's preference embedding in the source domain can be transformed into the target domain, and the transformed user preference embedding can be utilized as the initial embedding for the cold-start user in the target domain. Using large real-world datasets, we conduct extensive experiments to evaluate the effectiveness of PTUPCDR on both cold-start and warm-start stages. The code has been available at https://github.com/easezyc/WSDM2022-PTUPCDR.

Real-world recommendation systems need to deal with millions of item candidates. Therefore, most practical large-scale recommendation systems usually contain two modules. The matching module aims to efficiently retrieve hundreds of high-quality items from large corpora, while the ranking module aims to generate specific ranks for these items. Recommendation diversity is an essential factor that strongly impacts user experience. There are lots of efforts that have explored recommendation diversity in ranking, while the matching module should take more responsibility for diversity. In this article, we propose a novel Heterogeneous graph neural network framework for diversified recommendation (GraphDR) in matching to improve both recommendation accuracy and diversity. Specifically, GraphDR builds a huge heterogeneous preference network to record different types of user preferences, and conducts a field-level heterogeneous graph attention network for node aggregation. We conduct a neighbor-similarity based loss with a multi-channel matching to improve both accuracy and diversity. In experiments, we conduct extensive online and offline evaluations on a real-world recommendation system with various accuracy and diversity metrics and achieve significant improvements. GraphDR has been deployed on a well-known recommendation system named WeChat Top Stories, which affects millions of users. The source code will be released inhttps://github.com/lqfarmer/GraphDR.

WeChat is the largest social instant messaging platform in China, with 1.1 billion monthly active users. “Top Stories” is a novel friend-enhanced recommendation engine in WeChat, in which users can read articles based on preferences of both their own and their friends. Specifically, when a user reads an article by opening it, the “click” behavior is private. Moreover, if the user clicks the “wow” button, (only) her/his direct connections will be aware of this action/preference. Based on the unique WeChat data, we aim to understand user preferences and “wow” diffusion in Top Stories at different levels. We have made some interesting discoveries. For instance, the “wow” probability of one user is negatively correlated with the number of connected components that are formed by her/his active friends, but the click probability is the opposite. We further study to what extent users’ “wow” and click behavior can be predicted from their social connections. To address this problem, we present a hierarchical graph representation learning based model DiffuseGNN, which is capable of capturing the structure-based social observations discovered above. Our experiments show that the proposed method can significantly improve the prediction performance compared with alternative methods.

Integrated recommendation aims to jointly recommend heterogeneous items in the main feed from different sources via multiple channels, which needs to capture user preferences on both item and channel levels. It has been widely used in practical systems by billions of users, while few works concentrate on the integrated recommendation systematically. In this work, we propose a novel Hierarchical reinforcement learning framework for integrated recommendation (HRL-Rec), which divides the integrated recommendation into two tasks to recommend channels and items sequentially. The low-level agent is a channel selector, which generates a personalized channel list. The high-level agent is an item recommender, which recommends specific items from heterogeneous channels under the channel constraints. We design various rewards for both recommendation accuracy and diversity, and propose four losses for fast and stable model convergence. We also conduct an online exploration for sufficient training. In experiments, we conduct extensive offline and online experiments on a billion-level real-world dataset to show the effectiveness of HRL-Rec. HRL-Rec has also been deployed on WeChat Top Stories, affecting millions of users. The source codes are released in https://github.com/modriczhang/HRL-Rec.

Reinforcement learning (RL) has been verified in real-world list-wise recommendation. However, RL-based recommendation suffers from huge memory and computation costs due to its large-scale models. Knowledge distillation (KD) is an effective approach for model compression widely used in practice. However, RL-based models strongly rely on sufficient explorations on the enormous user-item space due to the data sparsity issue, which multiplies the challenges of KD with RL models. What the teacher should teach and how much the student should learn from each lesson need to be carefully designed. In this work, we propose a novel Distilled reinforcement learning framework for recommendation (DRL-Rec), which aims to improve both effectiveness and efficiency in list-wise recommendation. Specifically, we propose an Exploring and filtering module before the distillation, which decides what lessons the teacher should teach from both teachers' and students' aspects. We also conduct a Confidence-guided distillation at both output and intermediate levels with a list-wise KL divergence loss and a Hint loss, which aims to understand how much the student should learn for each lesson. We achieve significant improvements on both offline and online evaluations in a well-known recommendation system. DRL-Rec has been deployed on WeChat Top Stories for more than six months, affecting millions of users. The source codes are released in https://github.com/modriczhang/DRL-Rec.

Search and recommendation are the two most common approaches used by people to obtain information. They share the same goal -- satisfying the user's information need at the right time. There are already a lot of Internet platforms and Apps providing both search and recommendation services, showing us the demand and opportunity to simultaneously handle both tasks. However, most platforms consider these two tasks independently -- they tend to train separate search model and recommendation model, without exploiting the relatedness and dependency between them. In this paper, we argue that jointly modeling these two tasks will benefit both of them and finally improve overall user satisfaction. We investigate the interactions between these two tasks in the specific information content service domain. We propose first integrating the user's behaviors in search and recommendation into a heterogeneous behavior sequence, then utilizing a joint model for handling both tasks based on the unified sequence. More specifically, we design the Unified Information SEarch and Recommendation model (USER), which mines user interests from the integrated sequence and accomplish the two tasks in a unified way. Experiments on a dataset from a real-world information content service platform verify that our model outperforms separate search and recommendation baselines.

Real-world super platforms such as Google and WeChat usually have different recommendation scenarios to provide heterogeneous items for users' diverse demands. Multi-domain recommendation (MDR) is proposed to improve all recommendation domains simultaneously, where the key point is to capture informative domain-specific features from all domains. To address this problem, we propose a novel Adversarial feature translation (AFT) model for MDR, which learns the feature translations between different domains under a generative adversarial network framework. Precisely, in the multi-domain generator, we propose a domain-specific masked encoder to highlight inter-domain feature interactions, and then aggregate these features via a transformer and a domain-specific attention. In the multi-domain discriminator, we explicitly model the relationships between item, domain and users' general/domain-specific representations with a two-step feature translation inspired by the knowledge representation learning. In experiments, we evaluate AFT on a public and an industrial MDR datasets and achieve significant improvements. We also conduct an online evaluation on a real-world MDR system. We further give detailed ablation tests and model analyses to verify the effectiveness of different components. Currently, we have deployed AFT on WeChat Top Stories. The source code is in https://github.com/xiaobocser/AFT.

In recommender systems and advertising platforms, marketers always want to deliver products, contents, or advertisements to potential audiences over media channels such as display, video, or social. Given a set of audiences or customers (seed users), the audience expansion technique (look-alike modeling) is a promising solution to identify more potential audiences, who are similar to the seed users and likely to finish the business goal of the target campaign. However, look-alike modeling faces two challenges: (1) In practice, a company could run hundreds of marketing campaigns to promote various contents within completely different categories every day, e.g., sports, politics, society. Thus, it is difficult to utilize a common method to expand audiences for all campaigns. (2) The seed set of a certain campaign could only cover limited users. Therefore, a customized approach based on such a seed set is likely to be overfitting.

Kai Zhang, Yuan Yao, Ruobing Xie, Xu Han, Zhiyuan Liu, Fen Lin, Leyu Lin, Maosong Sun. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Learning disentangled representations for user intentions from multi-feedback (i.e., positive and negative feedback) can enhance the accuracy and explainability of recommendation algorithms. However, learning such disentangled representations from multi-feedback data is challenging because i) multi-feedback is complex: there exist complex relations among different types of feedback (e.g., click, unclick, and dislike, etc) as well as various user intentions, and ii) multi-feedback is noisy: there exists noisy (useless) information both in features and labels, which may deteriorate the recommendation performance. Existing works on disentangled representation learning only focus on positive feedback, failing to handle the complex relations and noise hidden in multi-feedback data. To solve this problem, in this work we propose a Curriculum Disentangled Recommendation (CDR) model that is capable of efficiently learning disentangled representations from complex and noisy multi-feedback for better recommendation. Concretely, we design a co-filtering dynamic routing mechanism that simultaneously captures the complex relations among different behavioral feedback and user intentions as well as denoise the representations in the feature level. We then present an adjustable self-evaluating curriculum that is able to evaluate sample difficulties for better model training and conduct denoising in the label level via disregarding useless information. Our extensive experiments on several real-world datasets demonstrate that the proposed CDR model can significantly outperform several state-of-the-art methods in terms of recommendation accuracy.

With the booming of online social networks in the mobile internet, an emerging recommendation scenario has played a vital role in information acquisition for user, where users are no longer recommended with a single item or item list, but a combination of heterogeneous and diverse objects (called a package, e.g., a package including news, publisher, and friends viewing the news). Different from the conventional recommendation where users are recommended with the item itself, in package recommendation, users would show great interests on the explicitly displayed objects that could have a significant influence on the user behaviors. However, to the best of our knowledge, few effort has been made for package recommendation and existing approaches can hardly model the complex interactions of diverse objects in a package. Thus, in this paper, we make a first study on package recommendation and propose an Intra- and inter-package attention network for Package Recommendation (IPRec). Specifically, for package modeling, an intra-package attention network is put forward to capture the object-level intention of user interacting with the package, while an inter-package attention network acts as a package-level information encoder that captures collaborative features of neighboring packages. In addition, to capture users preference representation, we present a user preference learner equipped with a fine-grained feature aggregation network and coarse-grained package aggregation network. Extensive experiments on three real-world datasets demonstrate that IPRec significantly outperforms the state of the arts. Moreover, the model analysis demonstrates the interpretability of our IPRec and the characteristics of user behaviors. Codes and datasets can be obtained at https://github.com/LeeChenChen/IPRec.

The key to personalized search is to build the user profile based on historical behaviour. To deal with the users who lack historical data, group based personalized models were proposed to incorporate the profiles of similar users when re-ranking the results. However, similar users are mostly found based on simple lexical or topical similarity in search behaviours. In this paper, we propose a neural network enhanced method to highlight similar users in semantic space. Furthermore, we argue that the behaviour-based similar users are still insufficient to understand a new query when user's historical activities are limited. To tackle this issue, we introduce the friend network into personalized search to determine the closeness between users in another way. Since the friendship is often formed based on similar background or interest, there are plenty of personalized signals hidden in the friend network naturally. Specifically, we propose a friend network enhanced personalized search model, which groups the user into multiple friend circles based on search behaviours and friend relations respectively. These two types of friend circles are complementary to construct a more comprehensive group profile for refining the personalization. Experimental results show the significant improvement of our model over existing personalized search models.

Cold-start problems are enormous challenges in practical recommender systems. One promising solution for this problem is cross-domain recommendation (CDR) which leverages rich information from an auxiliary (source) domain to improve the performance of recommender system in the target domain. In these CDR approaches, the family of Embedding and Mapping methods for CDR (EMCDR) is very effective, which explicitly learn a mapping function from source embeddings to target embeddings with overlapping users. However, these approaches suffer from one serious problem: the mapping function is only learned on limited overlapping users, and the function would be biased to the limited overlapping users, which leads to unsatisfying generalization ability and degrades the performance on cold-start users in the target domain. With the advantage of meta learning which has good generalization ability to novel tasks, we propose a transfer-meta framework for CDR (TMCDR) which has a transfer stage and a meta stage. In the transfer (pre-training) stage, a source model and a target model are trained on source and target domains, respectively. In the meta stage, a task-oriented meta network is learned to implicitly transform the user embedding in the source domain to the target feature space. In addition, the TMCDR is a general framework that can be applied upon various base models, e.g., MF, BPR, CML. By utilizing data from Amazon and Douban, we conduct extensive experiments on 6 cross-domain tasks to demonstrate the superior performance and compatibility of TMCDR.

Recently, embedding techniques have achieved impressive success in recommender systems. However, the embedding techniques are data demanding and suffer from the cold-start problem. Especially, for the cold-start item which only has limited interactions, it is hard to train a reasonable item ID embedding, called cold ID embedding, which is a major challenge for the embedding techniques. The cold item ID embedding has two main problems: (1) A gap is existing between the cold ID embedding and the deep model. (2) Cold ID embedding would be seriously affected by noisy interaction. However, most existing methods do not consider both two issues in the cold-start problem, simultaneously. To address these problems, we adopt two key ideas: (1) Speed up the model fitting for the cold item ID embedding (fast adaptation). (2) Alleviate the influence of noise. Along this line, we propose Meta Scaling and Shifting Networks to generate scaling and shifting functions for each item, respectively. The scaling function can directly transform cold item ID embeddings into warm feature space which can fit the model better, and the shifting function is able to produce stable embeddings from the noisy embeddings. With the two meta networks, we propose Meta Warm Up Framework (MWUF) which learns to warm up cold ID embeddings. Moreover, MWUF is a general framework that can be applied upon various existing deep recommendation models. The proposed model is evaluated on three popular benchmarks, including both recommendation and advertising datasets. The evaluation results demonstrate its superior performance and compatibility.

Users of recommendation systems usually focus on one topic at a time. When finishing reading an item, users may want to access more relevant items related to the last read one as extended reading. However, conventional recommendation systems are hard to provide the continuous extended reading function of these relevant items, since the main recommendation results should be diversified. In this paper, we propose a new task named recommendation suggestion, which aims to (1) predict whether users want extended reading, and (2) provide appropriate relevant items as suggestions. These recommended relevant items are arranged in a relevant box and instantly inserted below the clicked item in the main feed. The challenge of recommendation suggestion on relevant items is that it should further consider semantic relevance and information gain besides CTR-related factors. Moreover, the real-time relevant box insertion may also harm the overall performance when users do not want extended reading. To address these issues, we propose a novel Real-time relevant recommendation suggestion (R3S) framework, which consists of an Item recommender and a Box trigger. We extract features from multiple aspects including feature interaction, semantic similarity and information gain as different experts, and propose a new Multi-critic multi-gate mixture-of-experts (M3oE) strategy to jointly consider different experts with multi-head critics. In experiments, we conduct both offline and online evaluations on a real-world recommendation system with detailed ablation tests. The significant improvements in item/box related metrics verify the effectiveness of R3S. Moreover, we have deployed R3S on WeChat Top Stories, which affects millions of users. The source codes are in https://github.com/modriczhang/R3S.

Real-world recommendation systems usually have different learning objectives and evaluation criteria on accuracy, diversity or novelty. Therefore, multi-objective recommendation (MOR) has been widely explored to jointly model different objectives. Pareto efficiency, where no objective can be further improved without hurting others, is viewed as an optimal situation in multi-objective optimization. Recently, Pareto efficiency model has been introduced to MOR, while all existing scalarization methods only have shared objective weights for all instances. To capture users’ objective-level preferences and enhance personalization in Pareto-efficient recommendation, we propose a novel Personalized Approximate Pareto-Efficient Recommendation (PAPERec) framework for multi-objective recommendation. Specifically, we design an approximate Pareto-efficient learning based on scalarization with KKT conditions that closely mimics Pareto efficiency, where users have personalized weights on different objectives. We propose a Pareto-oriented reinforcement learning module to find appropriate personalized objective weights for each user, with the weighted sum of multiple objectives’ gradients considered in reward. In experiments, we conduct extensive offline and online evaluations on a real-world recommendation system. The significant improvements verify the effectiveness of PAPERec in practice. We have deployed PAPERec on WeChat Top Stories, affecting millions of users. The source codes are released in https://github.com/onepunch-cyber/PAPERec.

Knowledge graphs typically undergo open-ended growth of new relations. This cannot be well handled by relation extraction that focuses on pre-defined relations with sufficient training data. To address new relations with few-shot instances, we propose a novel bootstrapping approach, Neural Snowball, to learn new relations by transferring semantic knowledge about existing relations. More specifically, we use Relational Siamese Networks (RSN) to learn the metric of relational similarities between instances based on existing relations and their labeled data. Afterwards, given a new relation and its few-shot instances, we use RSN to accumulate reliable instances from unlabeled corpora; these instances are used to train a relation classifier, which can further identify new facts of the new relation. The process is conducted iteratively like a snowball. Experiments show that our model can gather high-quality instances for better few-shot relation learning and achieves significant improvement compared to baselines. Codes and datasets are released on https://github.com/thunlp/Neural-Snowball.

Knowledge graph (KG) entity typing aims at inferring possible missing entity type instances in KG, which is a very significant but still under-explored subtask of knowledge graph completion.In this paper, we propose a novel approach for KG entity typing which is trained by jointly utilizing local typing knowledge from existing entity type assertions and global triple knowledge from KGs.Specifically, we present two distinct knowledge-driven effective mechanisms of entity type inference.Accordingly, we build two novel embedding models to realize the mechanisms.Afterward, a joint model with them is used to infer missing entity type instances, which favors inferences that agree with both entity type instances and triple knowledge in KGs.Experimental results on two real-world datasets (Freebase and YAGO) demonstrate the effectiveness of our proposed mechanisms and models for improving KG entity typing.

In tag-enhanced video recommendation systems, videos are attached with some tags that highlight the contents of videos from different aspects. Tag ranking in such recommendation systems provides personalized tag lists for videos from their tag candidates. A better tag ranking model could attract users to click more tags, enter their corresponding tag channels, and watch more tag-specific videos, which improves both tag click rate and video watching time. However, most conventional tag ranking models merely concentrate on tag-video relevance or tag-related behaviors, ignoring the rich information in video-related behaviors. We should consider user preferences on both tags and videos. In this paper, we propose a novel Graph neural network based tag ranking (GraphTR) framework on a huge heterogeneous network with video, tag, user and media. We design a novel graph neural network that combines multi-field transformer, GraphSAGE and neural FM layers in node aggregation. We also propose a neighbor-similarity based loss to encode various user preferences into heterogeneous node representations. In experiments, we conduct both offline and online evaluations on a real-world video recommendation system in WeChat Top Stories. The significant improvements in both video and tag related metrics confirm the effectiveness and robustness in real-world tag-enhanced video recommendation. Currently, GraphTR has been deployed on WeChat Top Stories for more than six months. The source codes are in https://github.com/lqfarmer/GraphTR.

Few-shot classification requires classifiers to adapt to new classes with only a few training instances.State-of-the-art meta-learning approaches such as MAML learn how to initialize and fast adapt parameters from limited instances, which have shown promising results in few-shot classification.However, existing meta-learning models solely rely on implicit instance-based statistics, and thus suffer from instance unreliability and weak interpretability.To solve this problem, we propose a novel meta-information guided meta-learning (MIML) framework, where semantic concepts of classes provide strong guidance for meta-learning in both initialization and adaptation.In effect, our model can establish connections between instance-based information and semantic-based information, which enables more effective initialization and faster adaptation.Comprehensive experimental results on few-shot relation classification demonstrate the effectiveness of the proposed framework.Notably, MIML achieves comparable or superior performance to humans with only one shot on FewRel evaluation.The source code and experiment details of this paper can be obtained from https://github.com/thunlp/MIML.

Distant supervision (DS) has been widely used to generate auto-labeled data for sentencelevel relation extraction (RE), which improves RE performance.However, the existing success of DS cannot be directly transferred to the more challenging document-level relation extraction (DocRE), since the inherent noise in DS may be even multiplied in document level and significantly harm the performance of RE.To address this challenge, we propose a novel pre-trained model for DocRE, which denoises the document-level DS data via multiple pre-training tasks.Experimental results on the large-scale DocRE benchmark show that our model can capture useful information from noisy DS data and achieve promising results.The source code of this paper can be found in https://github.com/thunlp/DSDocRE.

Both explicit and implicit feedbacks can reflect user opinions on items, which are essential for learning user preferences in recommendation. However, most current recommendation algorithms merely focus on implicit positive feedbacks (e.g., click), ignoring other informative user behaviors. In this paper, we aim to jointly consider explicit/implicit and positive/negative feedbacks to learn user unbiased preferences for recommendation. Specifically, we propose a novel Deep feedback network (DFN) modeling click, unclick and dislike behaviors. DFN has an internal feedback interaction component that captures fine-grained interactions between individual behaviors, and an external feedback interaction component that uses precise but relatively rare feedbacks (click/dislike) to extract useful information from rich but noisy feedbacks (unclick). In experiments, we conduct both offline and online evaluations on a real-world recommendation system WeChat Top Stories used by millions of users. The significant improvements verify the effectiveness and robustness of DFN. The source code is in https://github.com/qqxiaochongqq/DFN.

Real-world integrated personalized recommendation systems usually deal with millions of heterogeneous items. It is extremely challenging to conduct full corpus retrieval with complicated models due to the tremendous computation costs. Hence, most large-scale recommendation systems consist of two modules: a multi-channel matching module to efficiently retrieve a small subset of candidates, and a ranking module for precise personalized recommendation. However, multi-channel matching usually suffers from cold-start problems when adding new channels or new data sources. To solve this issue, we propose a novel Internal and contextual attention network (ICAN), which highlights channel-specific contextual information and feature field interactions between multiple channels. In experiments, we conduct both offline and online evaluations with case studies on a real-world integrated recommendation system. The significant improvements confirm the effectiveness and robustness of ICAN, especially for cold-start channels. Currently, ICAN has been deployed on WeChat Top Stories used by millions of users. The source code can be obtained from https://github.com/zhijieqiu/ICAN.

The social media prediction task is aiming at predicting content popularity which includes social multimedia data such as photos, videos, and news. The task can not only help make better decisions for recommendation, but also reveals the public attention from evolutionary social systems. In this paper, we propose a novel approach named curriculum learning for wide multimedia-based transformer with graph target detection(CL-WMTG). The curriculum learning is designed for the transformer to improve the efficiency of model convergence. The mechanism of wide multimedia-based transformer is to make the model capable of learning cross information from text, pictures and other features(e.g. categories, location). Moreover, the graph target detection part can extract different features in the picture by pretrained model and reconstruct the features with a homogeneous graph network. We achieved third place in the SMP Challenge 2020.

Ruidong Wu, Yuan Yao, Xu Han, Ruobing Xie, Zhiyuan Liu, Fen Lin, Leyu Lin, Maosong Sun. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Knowledge graphs (KGs), which could provide essential relational information between entities, have been widely utilized in various knowledge-driven applications. Since the overall human knowledge is innumerable that still grows explosively and changes frequently, knowledge construction and update inevitably involve automatic mechanisms with less human supervision, which usually bring in plenty of noises and conflicts to KGs. However, most conventional knowledge representation learning methods assume that all triple facts in existing KGs share the same significance without any noises. To address this problem, we propose a novel confidence-aware knowledge representation learning framework (CKRL), which detects possible noises in KGs while learning knowledge representations with confidence simultaneously. Specifically, we introduce the triple confidence to conventional translation-based methods for knowledge representation learning. To make triple confidence more flexible and universal, we only utilize the internal structural information in KGs, and propose three kinds of triple confidences considering both local and global structural information. In experiments, We evaluate our models on knowledge graph noise detection, knowledge graph completion and triple classification. Experimental results demonstrate that our confidence-aware models achieve significant and consistent improvements on all tasks, which confirms the capability of CKRL modeling confidence with structural information in both KG noise detection and knowledge representation learning.

Sememes are minimum semantic units of concepts in human languages, such that each word sense is composed of one or multiple sememes.Words are usually manually annotated with their sememes by linguists, and form linguistic commonsense knowledge bases widely used in various NLP tasks.Recently, the lexical sememe prediction task has been introduced.It consists of automatically recommending sememes for words, which is expected to improve annotation efficiency and consistency.However, existing methods of lexical sememe prediction typically rely on the external context of words to represent the meaning, which usually fails to deal with low-frequency and out-ofvocabulary words.To address this issue for Chinese, we propose a novel framework to take advantage of both internal character information and external context information of words.We experiment on HowNet, a Chinese sememe knowledge base, and demonstrate that our framework outperforms state-of-the-art baselines by a large margin, and maintains a robust performance even for low-frequency words.i

Most language modeling methods rely on large-scale data to statistically learn the sequential patterns of words.In this paper, we argue that words are atomic language units but not necessarily atomic semantic units.Inspired by HowNet, we use sememes, the minimum semantic units in human languages, to represent the implicit semantics behind words for language modeling, named Sememe-Driven Language Model (SDLM).More specifically, to predict the next word, SDLM first estimates the sememe distribution given textual context.Afterwards, it regards each sememe as a distinct semantic expert, and these experts jointly identify the most probable senses and the corresponding word.In this way, SDLM enables language models to work beyond word-level manipulation to fine-grained sememe-level semantics, and offers us more powerful tools to fine-tune language models and improve the interpretability as well as the robustness of language models.Experiments on language modeling and the downstream application of headline generation demonstrate the significant effectiveness of SDLM.

Sememes are defined as the minimum semantic units of human languages.As important knowledge sources, sememe-based linguistic knowledge bases have been widely used in many NLP tasks.However, most languages still do not have sememe-based linguistic knowledge bases.Thus we present a task of cross-lingual lexical sememe prediction, aiming to automatically predict sememes for words in other languages.We propose a novel framework to model correlations between sememes and multi-lingual words in low-dimensional semantic space for sememe prediction.Experimental results on real-world datasets show that our proposed model achieves consistent and significant improvements as compared to baseline methods in cross-lingual sememe prediction.The codes and data of this paper are available at https: //github.com/thunlp/CL-SP.

Sememes are minimum semantic units of word meanings, and the meaning of each word sense is typically composed by several sememes.Since sememes are not explicit for each word, people manually annotate word sememes and form linguistic common-sense knowledge bases.In this paper, we present that, word sememe information can improve word representation learning (WRL), which maps words into a low-dimensional semantic space and serves as a fundamental step for many NLP tasks.The key idea is to utilize word sememes to capture exact meanings of a word within specific contexts accurately.More specifically, we follow the framework of Skip-gram and present three sememe-encoded models to learn representations of sememes, senses and words, where we apply the attention scheme to detect word senses in various contexts.We conduct experiments on two tasks including word similarity and word analogy, and our models significantly outperform baselines.The results indicate that WRL can benefit from sememes via the attention scheme, and also confirm our models being capable of correctly modeling sememe information.

Entity images could provide significant visual information for knowledge representation learning. Most conventional methods learn knowledge representations merely from structured triples, ignoring rich visual information extracted from entity images. In this paper, we propose a novel Image-embodied Knowledge Representation Learning model (IKRL), where knowledge representations are learned with both triple facts and images. More specifically, we first construct representations for all images of an entity with a neural image encoder. These image representations are then integrated into an aggregated image-based representation via an attention-based method. We evaluate our IKRL models on knowledge graph completion and triple classification. Experimental results demonstrate that our models outperform all baselines on both tasks, which indicates the significance of visual information for knowledge representations and the capability of our models in learning knowledge representations with images.

Sememes are defined as the minimum semantic units of human languages. People have manually annotated lexical sememes for words and form linguistic knowledge bases. However, manual construction is time-consuming and labor-intensive, with significant annotation inconsistency and noise. In this paper, we for the first time explore to automatically predict lexical sememes based on semantic meanings of words encoded by word embeddings. Moreover, we apply matrix factorization to learn semantic relations between sememes and words. In experiments, we take a real-world sememe knowledge base HowNet for training and evaluation, and the results reveal the effectiveness of our method for lexical sememe prediction. Our method will be of great use for annotation verification of existing noisy sememe knowledge bases and annotation suggestion of new words and phrases.

Entity alignment aims to link entities and their counterparts among multiple knowledge graphs (KGs). Most existing methods typically rely on external information of entities such as Wikipedia links and require costly manual feature construction to complete alignment. In this paper, we present a novel approach for entity alignment via joint knowledge embeddings. Our method jointly encodes both entities and relations of various KGs into a unified low-dimensional semantic space according to a small seed set of aligned entities. During this process, we can align entities according to their semantic distance in this joint semantic space. More specifically, we present an iterative and parameter sharing method to improve alignment performance. Experiment results on real-world datasets show that, as compared to baselines, our method achieves significant improvements on entity alignment, and can further improve knowledge graph completion performance on various KGs with the favor of joint knowledge embeddings.

Representation learning (RL) of knowledge graphs aims to project both entities and relations into a continuous low-dimensional space. Most methods concentrate on learning representations with knowledge triples indicating relations between entities. In fact, in most knowledge graphs there are usually concise descriptions for entities, which cannot be well utilized by existing methods. In this paper, we propose a novel RL method for knowledge graphs taking advantages of entity descriptions. More specifically, we explore two encoders, including continuous bag-of-words and deep convolutional neural models to encode semantics of entity descriptions. We further learn knowledge representations with both triples and descriptions. We evaluate our method on two tasks, including knowledge graph completion and entity classification. Experimental results on real-world datasets show that, our method outperforms other baselines on the two tasks, especially under the zero-shot setting, which indicates that our method is capable of building representations for novel entities according to their descriptions. The source code of this paper can be obtained from https://github.com/xrb92/DKRL.