CFChilin Fu

Chilin Fu

*This is an algorithmically generated author record based on publicly available information.

Chilin Fu has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Corrigendum: One Model for All: Large Language Models Are Domain-Agnostic Recommendation Systems

    by Zuoli Tang, Zhaoxin Huan, Zihao Li, Xiaolu Zhang, Jun Hu, Chilin Fu, Jun Zhou, Lixin Zou, Chenliang Li

    This is a corrigendum for the article “One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems” published in ACM Trans. Inf. Syst . 43, 5, Article 118 (July 2025), 27 pages.

  • Top Cited Papers
    Open Access

    One Model for All: Large Language Models Are Domain-Agnostic Recommendation Systems

    by Zuoli Tang, Zhaoxin Huan, Zihao Li, Xiaolu Zhang, Jun Hu, Chilin Fu, Jun Zhou, Lixin Zou, Chenliang Li

    Sequential recommendation systems aim to predict users’ next likely interaction based on their history. However, these systems face data sparsity and cold-start problems. Utilizing data from other domains, known as multi-domain methods, is useful for alleviating these problems. However, traditional multi-domain methods rely on meaningless ID-based item representation, which makes it difficult to align items with similar meanings from different domains, yielding sup-optimal knowledge transfer. This article introduces LLM-Rec , a framework that utilizes pre-trained Large Language Models (LLMs) for domain-agnostic recommendation. Specifically, we mix user’s behaviors from multiple domains and concatenate item titles into a sentence, then use LLMs for generating user and item representations. By mixing behaviors across different domains, we can exploit the knowledge encoded in LLMs to bridge the semantic across over multi-domain behaviors, thus obtaining semantically rich representations and improving performance in all domains. Furthermore, we explore the underlying reasons why LLMs are effective and investigate whether LLMs can understand the semantic correlations as the recommendation model, and if advanced techniques like scaling laws in NLP also work in recommendations. We conduct extensive experiments with LLMs ranging from 40 M to 6.7 B to answer the above questions and to verify the effectiveness of LLM-Rec in multi-domain recommendation. The source code is available at https://github.com/WHUIR/LLMRec .


  • Open Access

    Enhancing Sequential Recommendation via LLM-based Semantic Embedding Learning

    by Jun Hu, Wenwen Xia, Xiaolu Zhang, Chilin Fu, Weichang Wu, Zhaoxin Huan, Ang Li, Zuoli Tang, Jun Zhou

    Sequential recommendation systems (SRS) are crucial in various applications as they enable users to discover relevant items based on their past interactions. Recent advancements involving large language models (LLMs) have shown significant promise in addressing intricate recommendation challenges. However, these efforts exhibit certain limitations. Specifically, directly extracting representations from an LLM based on items' textual features and feeding them into a sequential model hold no guarantee that the semantic information of texts could be preserved in these representations. Additionally, concatenating textual descriptions of all items in an item sequence into a long text and feeding it into an LLM for recommendation results in lengthy token sequences, which largely diminishes the practical efficiency. In this paper, we introduce SAID, a framework that utilizes LLMs to explicitly learn Semantically Aligned item ID embeddings based on texts. For each item, SAID employs a projector module to transform an item ID into an embedding vector, which will be fed into an LLM to elicit the exact descriptive text tokens accompanied by the item. The item embeddings are forced to preserve fine-grained semantic information of textual descriptions. Further, the learned embeddings can be integrated with lightweight downstream sequential models for practical recommendations. In this way, SAID circumvents lengthy token sequences in previous works, reducing resources required in industrial scenarios and also achieving superior recommendation performance. Experiments on six public datasets demonstrate that SAID outperforms baselines by about 5% to 15% in terms of NDCG@10. Moreover, SAID has been deployed in Alipay's online advertising platform, achieving a 3.07% relative improvement of cost per mille (CPM) over baselines, with an online response time of under 20 milliseconds.


  • Open Access

    Robust User Behavioral Sequence Representation via Multi-scale Stochastic Distribution Prediction

    by Chilin Fu, Weichang Wu, Xiaolu Zhang, Jun Hu, Jing Wang, Jun Zhou

    User behavior representation learned by self-supervised pre-training tasks is widely used in various domains and applications. Conventional methods usually follow the methodology in Natural Language Processing (NLP) to set the pre-training tasks. They either randomly mask some of the behaviors in the sequence and predict the masked ones or predict the next k behaviors. These methods fit for text sequence, in which the tokens are sequentially arranged subject to linguistic criterion. However, the user behavior sequences can be stochastic with noise and randomness. The same paradigm is intractable for learning a robust user behavioral representation. Though the next user behavior can be stochastic, the behavior distribution over a period of time is much more stable and less noisy. Based on this, we propose a Multi-scale Stochastic Distribution Prediction (MSDP) algorithm for learning robust user behavioral sequence representation. Instead of using predictions on concrete behavior as pre-training tasks, we take the prediction on user's behaviors distribution over a period of time as the self-supervision signal. Moreover, inspired by the recent success of the multi-task prompt training method on Large Language Models (LLM), we propose using the window size of the predicted time period as a prompt, enabling the model to learn user behavior representations that can be applied to prediction tasks across various future time periods. We generate different window size prompts through stochastic sampling. It effectively improves the generalization capability of the learned sequence representation. Extensive experiments demonstrate that our approach can learn robust user behavior representation successfully, which significantly outperforms state-of-the-art (SOTA) baselines.


  • Multi-Task Decouple Learning With Hierarchical Attentive Point Process

    by Weichang Wu, Xiaolu Zhang, Shiwan Zhao, Chilin Fu, Jun Zhou

    Sequential data mining is ubiquitous in various scenarios. Modeling event sequence and predicting event occurrence is of vital importance in sequential data mining, and Temporal Point Processes (TPP) are widely used in this area. Conventional TPP use objective functions as sum of classification loss for event type and regression loss for occurrence time, leading to practical limitations that conventional TPP is unable to predict the occurrence of each type of event and distinguish the dependency within and between different event types. To tackle these defects, we propose a Multi-task Decouple Learning (MTDL) framework to model TPP from a novel perspective of Multi-task Learning (MTL), i.e. predicting the next-step occurrence time for all event types using a weighted multi-task regression loss. We experiment with three state-of-the-arts, showing that the proposed MTDL framework can improve the performance of original TPP models. Moreover, we develop a Hierarchical Attentive Point Process (HAPP) to further exploit the potential of the proposed MTDL framework, using a hierarchical attention mechanism to capture the inner-sequence time dependency within the same type of events and the inter-sequence dependency between different types of events. Experiments on real-world business dataset and public datasets show the efficacy of the proposed method.


  • Open Access

    Multi-Source Domain Adaptation via Latent Domain Reconstruction

    by Jun Zhou, Chilin Fu, Xiaolu Zhang

    Multi-Source Domain Adaptation (MSDA) is widely used in various machine learning scenarios for domain shifts between labeled source domains and unlabeled target domains. Conventional MSDA methods are built on a strong hypothesis that data samples from the same source belong to the same domain with the same latent distribution. However, in practice sources and their latent domains are not necessarily one-to-one correspondence. To tackle this problem, a novel Multi-source Reconstructed Domain Adaptation (MRDA) framework for MSDA is proposed. We use an Expectation-Maximization (EM) mechanism that iteratively reconstructs the source domains to recover the latent domains and performs domain adaptation on the reconstructed domains. Specifically, in the E-step, we cluster the samples from multiple sources into different latent domains, and a soft assignment strategy is proposed to avoid cluster imbalance. In the M-step, we freeze the latent domains clustered in the E-step and optimize the objective function for domain adaptation, and a global-specific feature extractor is used to capture both domain-invariant and domain-specific features. Extensive experiments demonstrate that our approach can reconstruct source domains and perform domain adaptation on the reconstructed domains effectively, thus significantly outperforming state-of-the-art (SOTA) baselines (e.g., 1% to 3.1% absolute improvement in AUC).