Beyond Item Order: Temporal Gap Tokenization for Generative Recommendation with Semantic IDs

2026-07-04

AI Summary

This work addresses a critical limitation in existing semantic ID-based generative recommendation methods, which neglect temporal intervals between user interactions and thus struggle to capture interest continuity and preference drift. To overcome this, the authors propose ChronoSID, a novel framework that systematically integrates temporal interval information into semantic ID-based generative recommendation for the first time. ChronoSID employs a dual-view temporal enhancement strategy: it learns temporally regularized item representations via Time-Aware Field-Aware Masked Autoencoders (TA-FAMAE) and incorporates log-scale discretized time intervals as gap tokens interleaved with semantic IDs in the sequence-to-sequence generator. Evaluated on Amazon review datasets, this lightweight approach significantly outperforms baselines such as ReSID, with particularly notable gains in long-interval scenarios. Ablation studies confirm the individual effectiveness of both temporal components.

Abstract

Semantic-ID-based generative recommendation has recently emerged as a scalable paradigm for sequential recommendation, where each item is represented by a compact sequence of discrete codes and next-item prediction is formulated as code generation. Existing methods, however, typically construct user histories as sequences of static item identifiers, leaving the elapsed time between consecutive interactions outside the generative input. This temporal blindness is problematic because inter-interaction gaps provide useful cues about interest continuity and preference drift. In this paper, we propose ChronoSID, a lightweight temporal augmentation framework for semantic-ID-based generative recommendation. ChronoSID injects temporal signals into the standard three-stage semantic-ID pipeline from two complementary perspectives. First, we introduce Time-Aware Field-Aware Masked Auto-Encoding (TA-FAMAE), which regularizes item representation learning with an auxiliary time-gap prediction objective. Second, we discretize historical interaction intervals into fixed log-scale gap tokens and interleave them with semantic ID tuples as the encoder input of the sequence-to sequence generator. This design preserves the compact SID generation paradigm while enabling the model to capture time-aware transition patterns. Experiments on Amazon review benchmarks show that ChronoSID consistently improves over ReSID and other competitive generative recommendation baselines. Ablation studies further verify the contribution of both temporal components, and diagnostic analyses show clearer gains under long-gap scenarios where user interests are more likely to drift.

Problem

• generative recommendation
• semantic IDs
• temporal gap
• sequential recommendation
• time-aware modeling

Innovation

• Temporal Gap Tokenization
• Semantic ID
• Generative Recommendation
• Time-Aware Representation
• Sequence-to-Sequence Modeling