arXiv subjects

Longtao Xiao

Publications and source records attributed to Longtao Xiao.

4 recordsLinked to original sources

arXiv2026

RoTE: Coarse-to-Fine Multi-Level Rotary Time Embedding for Sequential Recommendation

Sequential recommendation models have been widely adopted for modeling user behavior. Existing approaches typically construct user interaction sequences by sorting items according to timestamps and then model user preferences from historical behaviors. While effective, such a process only considers the order of temporal information but overlooks the actual time spans between interactions, resulting in a coarse representation of users' temporal dynamics and limiting the model's ability to capture long-term and short-term interest evolution. To address this limitation, we propose RoTE, a novel multi-level temporal embedding module that explicitly models time span information in sequential recommendation. RoTE decomposes each interaction timestamp into multiple temporal granularities, ranging from coarse to fine, and incorporates the resulting temporal representations into item embeddings. This design enables models to capture heterogeneous temporal patterns and better perceive temporal distances among user interactions during sequence modeling. RoTE is a lightweight, plug-and-play module that can be seamlessly integrated into existing Transformer-based sequential recommendation models without modifying their backbone architectures. We apply RoTE to several representative models and conduct extensive experiments on three public benchmarks. Experimental results demonstrate that RoTE consistently enhances the corresponding backbone models, achieving up to a 20.11% improvement in NDCG@5, which confirms the effectiveness and generality of the proposed approach. Our code is available at https://github.com/XiaoLongtaoo/RoTE.

cs.IR↗

arXiv2025

FAIR: Focused Attention Is All You Need for Generative Recommendation

Recently, transformer-based generative recommendation has garnered significant attention for user behavior modeling. However, it often requires discretizing items into multi-code representations (e.g., typically four code tokens or more), which sharply increases the length of the original item sequence. This expansion poses challenges to transformer-based models for modeling user behavior sequences with inherent noises, since they tend to overallocate attention to irrelevant or noisy context. To mitigate this issue, we propose FAIR, the first generative recommendation framework with focused attention, which enhances attention scores to relevant context while suppressing those to irrelevant ones. Specifically, we propose (1) a focused attention mechanism integrated into the standard Transformer, which learns two separate sets of Q and K attention weights and computes their difference as the final attention scores to eliminate attention noise while focusing on relevant contexts; (2) a noise-robustness objective, which encourages the model to maintain stable attention patterns under stochastic perturbations, preventing undesirable shifts toward irrelevant context due to noise; and (3) a mutual information maximization objective, which guides the model to identify contexts that are most informative for next-item prediction. We validate the effectiveness of FAIR on four public benchmarks, demonstrating its superior performance compared to existing methods.

cs.IR↗

arXiv2025

UNGER: Generative Recommendation with A Unified Code via Semantic and Collaborative Integration

With the rise of generative paradigms, generative recommendation has garnered increasing attention. The core component is the item code, generally derived by quantizing collaborative or semantic representations to serve as candidate items identifiers in the context. However, existing methods typically construct separate codes for each modality, leading to higher computational and storage costs and hindering the integration of their complementary strengths. Considering this limitation, we seek to integrate two different modalities into a unified code, fully unleashing the potential of complementary nature among modalities. Nevertheless, the integration remains challenging: the integrated embedding obtained by the common concatenation method would lead to underutilization of collaborative knowledge, thereby resulting in limited effectiveness. To address this, we propose a novel method, named UNGER, which integrates semantic and collaborative knowledge into a unified code for generative recommendation. Specifically, we propose to adaptively learn an integrated embedding through the joint optimization of cross-modality knowledge alignment and next item prediction tasks. Subsequently, to mitigate the information loss caused by the quantization process, we introduce an intra-modality knowledge distillation task, using the integrated embeddings as supervised signals to compensate. Extensive experiments on three widely used benchmarks demonstrate the superiority of our approach compared to existing methods.

cs.IR↗

arXiv2025

Act-With-Think: Chunk Auto-Regressive Modeling for Generative Recommendation

Generative recommendation (GR) typically encodes behavioral or semantic aspects of item information into discrete tokens, leveraging the standard autoregressive (AR) generation paradigm to make predictions. However, existing methods tend to overlook their intrinsic relationship, that is, the semantic usually provides some reasonable explainability "$\textbf{why}$" for the behavior "$\textbf{what}$", which may constrain the full potential of GR. To this end, we present Chunk AutoRegressive Modeling (CAR), a new generation paradigm following the decision pattern that users usually think semantic aspects of items (e.g. brand) and then take actions on target items (e.g. purchase). Our CAR, for the $\textit{first time}$, incorporates semantics (SIDs) and behavior (UID) into a single autoregressive transformer from an ``act-with-think'' dual perspective via chunk-level autoregression. Specifically, CAR packs SIDs and UID into a conceptual chunk for item unified representation, allowing each decoding step to make a holistic prediction. Experiments show that our CAR significantly outperforms existing methods based on traditional AR, improving Recall@5 by 7.93% to 22.30%. Furthermore, we verify the scaling effect between model performance and SIDs bit number, demonstrating that CAR preliminary emulates a kind of slow-thinking style mechanism akin to the reasoning processes observed in large language models (LLMs).

cs.IR↗