ABXI: Invariant Interest Adaptation for Task-Guided Cross-Domain Sequential Recommendation

TL;DR

CDSR suffers from data sparsity and a prediction-mismatch problem when cross-domain knowledge is not carefully disentangled. ABXI addresses this by introducing a task-guided alignment that unifies cross-domain and domain-specific ground truths, together with a single shared encoder augmented by domain LoRAs and an invariant projector powered by iLoRAs. The model optimizes with an InfoNCE objective $f(h) = - \log \frac{\exp(h \cdot e^+ / \tau)}{\sum_{e \in \boldsymbol{E}} \exp(h \cdot e / \tau)}$, enabling effective alignment of representations across domains while preserving domain-specific signals, with a complexity of $O(dL^2 + Ld^2 + dj)$. Empirically, ABXI outperforms state-of-the-art CDSR baselines on three Amazon-derived datasets, and ablations confirm the critical roles of task-guided alignment, invariant interest extraction, and the LoRA modules in delivering robust cross-domain transfer.

Abstract

Cross-Domain Sequential Recommendation (CDSR) has recently gained attention for countering data sparsity by transferring knowledge across domains. A common approach merges domain-specific sequences into cross-domain sequences, serving as bridges to connect domains. One key challenge is to correctly extract the shared knowledge among these sequences and appropriately transfer it. Most existing works directly transfer unfiltered cross-domain knowledge rather than extracting domain-invariant components and adaptively integrating them into domain-specific modelings. Another challenge lies in aligning the domain-specific and cross-domain sequences. Existing methods align these sequences based on timestamps, but this approach can cause prediction mismatches when the current tokens and their targets belong to different domains. In such cases, the domain-specific knowledge carried by the current tokens may degrade performance. To address these challenges, we propose the A-B-Cross-to-Invariant Learning Recommender (ABXI). Specifically, leveraging LoRA's effectiveness for efficient adaptation, ABXI incorporates two types of LoRAs to facilitate knowledge adaptation. First, all sequences are processed through a shared encoder that employs a domain LoRA for each sequence, thereby preserving unique domain characteristics. Next, we introduce an invariant projector that extracts domain-invariant interests from cross-domain representations, utilizing an invariant LoRA to adapt these interests into modeling each specific domain. Besides, to avoid prediction mismatches, all domain-specific sequences are aligned to match the domains of the cross-domain ground truths. Experimental results on three datasets demonstrate that our approach outperforms other CDSR counterparts by a large margin. The codes are available in https://github.com/DiMarzioBian/ABXI.

Figures (4)

• Figure 1: Our proposal on generating recommendations by integrating domain-specific interests with domain-invariant interests extracted from the cross-domain sequence.
• Figure 2: Illustration of the sequence splits under different alignments, where gt denotes the ground truth. (a) illustrates the input raw sequence. (b) and (c) demonstrate the split outcomes of timestamp-guided and task-guided alignment, respectively.
• Figure 3: Proposed ABXI model.
• Figure 4: MRR performance for varying rank values $r_d$ for dLoRA and $r_i$ for iLoRA (RQ4).