Reasoning over Semantic IDs Enhances Generative Recommendation

2026-03-24

AI Summary

This work addresses the limitations of existing semantic ID (SID)-based generative recommendation methods, which struggle to effectively harness the reasoning capabilities of large language models (LLMs) due to the non-native semantics of SIDs and the absence of high-quality reasoning supervision. To overcome these challenges, we propose SIDReasoner, a novel framework that enhances semantic alignment between SIDs and natural language through a two-stage training process and introduces a result-driven reinforcement optimization mechanism. This approach enables LLMs to learn effective recommendation reasoning paths without requiring explicit reasoning annotations. SIDReasoner represents the first method to achieve efficient LLM-based reasoning over SIDs, significantly improving recommendation accuracy, interpretability, and cross-domain generalization. Extensive experiments on three real-world datasets demonstrate its superior performance.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Recent advances in generative recommendation have leveraged pretrained LLMs by formulating sequential recommendation as autoregressive generation over a unified token space comprising language tokens and itemic identifiers, where each item is represented by a compact sequence of discrete tokens, namely Semantic IDs (SIDs). This SID-based formulation enables efficient decoding over large-scale item corpora and provides a natural interface for LLM-based recommenders to leverage rich world knowledge. Meanwhile, breakthroughs in LLM reasoning motivate reasoning-enhanced recommendation, yet effective reasoning over SIDs remains underexplored and challenging. Itemic tokens are not natively meaningful to LLMs; moreover, recommendation-oriented SID reasoning is hard to evaluate, making high-quality supervision scarce. To address these challenges, we propose SIDReasoner, a two-stage framework that elicits reasoning over SIDs by strengthening SID--language alignment to unlock transferable LLM reasoning, rather than relying on large amounts of recommendation-specific reasoning traces. Concretely, SIDReasoner first enhances SID-language alignment via multi-task training on an enriched SID-centered corpus synthesized by a stronger teacher model, grounding itemic tokens in diverse semantic and behavioral contexts. Building on this enhanced alignment, SIDReasoner further improves recommendation reasoning through outcome-driven reinforced optimization, which guides the model toward effective reasoning trajectories without requiring explicit reasoning annotations. Extensive experiments on three real-world datasets demonstrate the effectiveness of our reasoning-augmented SID-based generative recommendation. Beyond accuracy, the results highlight the broader potential of large reasoning models for generative recommendation, including improved interpretability and cross-domain generalization.

Problem

• Semantic IDs
• generative recommendation
• LLM reasoning
• recommendation reasoning
• SID-language alignment

Innovation

• Semantic IDs
• Generative Recommendation
• LLM Reasoning
• SID-Language Alignment
• Reinforced Optimization