TRACER: Token ReAssignment for Concept ERasure in Generative Recommendation

2026-06-05

AI Summary

This work addresses the challenge in generative recommender systems where concept unlearning and recommendation utility are difficult to balance due to semantic ID sharing. To tackle this, the authors propose an end-to-end concept unlearning framework that innovatively reassigns semantic IDs associated with target concepts to newly introduced tokens, thereby avoiding direct suppression of shared IDs. Additionally, a semantic consistency regularizer is incorporated to preserve the semantic coherence of retained items. Experimental results on real-world recommendation datasets demonstrate that the proposed framework effectively erases sensitive or harmful concepts while significantly outperforming existing methods, achieving a superior trade-off between unlearning efficacy and recommendation utility.

Abstract

Generative recommendation formulates next-item prediction as autoregressive generation over semantic ID (SID) sequences derived from users' historical interactions, making modern recommender systems structurally similar to large language models (LLMs). As privacy and safety concerns grow, these systems increasingly require concept unlearning to remove sensitive or harmful concepts associated with items. However, existing LLM unlearning methods cannot be directly applied to generative recommendation. Unlike word tokens with explicit semantics, SIDs are abstract identifiers that are often shared by both forget and retain items, leading to severe conflicts between concept removal and recommendation utility preservation. To address this challenge, we propose TRACER, an end-to-end concept unlearning framework based on token reassignment. Rather than directly suppressing shared SIDs, TRACER reassigns concept-related items to alternative tokens that better facilitate forgetting while minimizing side effects on retained items. We further introduce a coherence regularizer to preserve semantic consistency among retain items during unlearning. Experiments on real-world recommendation datasets demonstrate that TRACER effectively removes target concepts while substantially better preserving recommendation utility than existing unlearning baselines.

Problem

• concept unlearning
• generative recommendation
• semantic ID
• token reassignment
• recommendation utility

Innovation

• Token Reassignment
• Concept Unlearning
• Generative Recommendation
• Semantic ID
• Coherence Regularization