ReaSeq: Unleashing World Knowledge via Reasoning for Sequential Modeling

2025-12-24

AI Summary

Industrial recommendation systems face two key challenges: (1) impoverished ID-based representations, leading to fragile interest modeling in sparse scenarios, and (2) log boundary limitations, preventing models from capturing off-platform user behavior. To address these, we propose a world-knowledge-enhanced sequential recommendation paradigm featuring a novel explicit-implicit dual-path reasoning mechanism: explicitly distilling structured item semantics via multi-agent Chain-of-Thought reasoning, and implicitly inferring cross-domain latent behaviors using a diffusion-based large language model (Diffusion LLM). Our method integrates semantically aligned embeddings with industrial real-time ranking systems. Deployed on Taobao’s primary search engine, it achieves >6.0% improvements in IPV and CTR, and 2.9% and 2.5% gains in order volume and GMV, respectively. This work marks the first systematic integration and large-scale industrial deployment of world knowledge in sequential recommendation.

Technology Category

• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Industrial recommender systems face two fundamental limitations under the log-driven paradigm: (1) knowledge poverty in ID-based item representations that causes brittle interest modeling under data sparsity, and (2) systemic blindness to beyond-log user interests that constrains model performance within platform boundaries. These limitations stem from an over-reliance on shallow interaction statistics and close-looped feedback while neglecting the rich world knowledge about product semantics and cross-domain behavioral patterns that Large Language Models have learned from vast corpora. To address these challenges, we introduce ReaSeq, a reasoning-enhanced framework that leverages world knowledge in Large Language Models to address both limitations through explicit and implicit reasoning. Specifically, ReaSeq employs explicit Chain-of-Thought reasoning via multi-agent collaboration to distill structured product knowledge into semantically enriched item representations, and latent reasoning via Diffusion Large Language Models to infer plausible beyond-log behaviors. Deployed on Taobao's ranking system serving hundreds of millions of users, ReaSeq achieves substantial gains: >6.0% in IPV and CTR, >2.9% in Orders, and >2.5% in GMV, validating the effectiveness of world-knowledge-enhanced reasoning over purely log-driven approaches.

Problem

• Enhances item representations with world knowledge to combat data sparsity
• Infers beyond-log user interests using reasoning to break platform boundaries
• Leverages LLM reasoning to overcome limitations of shallow interaction statistics

Innovation

• Leverages world knowledge from Large Language Models for reasoning
• Uses explicit Chain-of-Thought reasoning to enrich item representations
• Employs latent reasoning with Diffusion LLMs to infer beyond-log behaviors