Mitigating Collaborative Semantic ID Staleness in Generative Retrieval

2026-04-14

AI Summary

In generative retrieval, interaction-aware semantic IDs suffer from semantic obsolescence due to temporal drift in user behavior, leading to degraded performance. This work presents the first systematic analysis of this temporal obsolescence issue and introduces a lightweight, model-agnostic alignment method for updating semantic IDs. The proposed approach enables efficient warm-start fine-tuning to align the retriever with the latest log-derived semantic IDs, without requiring vocabulary reconstruction or full retraining. Evaluated on three public benchmarks, the method significantly improves Recall@K and nDCG@K at high cutoff ranks while reducing computational overhead by 8–9× compared to full retraining.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Generative retrieval with Semantic IDs (SIDs) assigns each item a discrete identifier and treats retrieval as a sequence generation problem rather than a nearest-neighbor search. While content-only SIDs are stable, they do not take into account user-item interaction patterns, so recent systems construct interaction-informed SIDs. However, as interaction patterns drift over time, these identifiers become stale, i.e., their collaborative semantics no longer match recent logs. Prior work typically assumes a fixed SID vocabulary during fine-tuning, or treats SID refresh as a full rebuild that requires retraining. However, SID staleness under temporal drift is rarely analyzed explicitly. To bridge this gap, we study SID staleness under strict chronological evaluation and propose a lightweight, model-agnostic SID alignment update. Given refreshed SIDs derived from recent logs, we align them to the existing SID vocabulary so the retriever checkpoint remains compatible, enabling standard warm-start fine-tuning without a full rebuild-and-retrain pipeline. Across three public benchmarks, our update consistently improves Recall@K and nDCG@K at high cutoffs over naive fine-tuning with stale SIDs and reduces retriever-training compute by approximately 8-9 times compared to full retraining.

Problem

• Semantic ID Staleness
• Generative Retrieval
• Temporal Drift
• Collaborative Semantics
• User-Item Interaction

Innovation

• Semantic ID
• generative retrieval
• temporal drift
• SID alignment
• warm-start fine-tuning