RecGPT-V2 Technical Report

2025-12-16

AI Summary

To address four critical limitations of existing LLM-based recommender systems—computational redundancy, monolithic explanation generation, poor generalization, and misaligned evaluation—this paper proposes the first LLM-augmented intent reasoning framework for recommendation. Methodologically, it introduces: (1) hierarchical multi-agent collaborative reasoning to decouple user interest modeling from item understanding; (2) a meta-prompting mechanism for dynamic, diverse explanation generation; (3) constraint-aware reinforcement learning for multi-objective decoupled optimization; and (4) an “agent-as-judge” evaluation paradigm aligned with human judgment criteria. Experiments demonstrate a 60% reduction in GPU resource consumption, a 1.6-percentage-point gain in domain-specific recall, a 24.1% improvement in label prediction accuracy, and 7.3% and 13.0% gains in explanation diversity and acceptance rate, respectively. Online A/B tests on Taobao show significant improvements: +2.98% CTR, +3.71% IPV, +2.19% TV, and +11.46% NER—advancing recommendation from behavioral matching to explicit intent understanding.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Large language models (LLMs) have demonstrated remarkable potential in transforming recommender systems from implicit behavioral pattern matching to explicit intent reasoning. While RecGPT-V1 successfully pioneered this paradigm by integrating LLM-based reasoning into user interest mining and item tag prediction, it suffers from four fundamental limitations: (1) computational inefficiency and cognitive redundancy across multiple reasoning routes; (2) insufficient explanation diversity in fixed-template generation; (3) limited generalization under supervised learning paradigms; and (4) simplistic outcome-focused evaluation that fails to match human standards. To address these challenges, we present RecGPT-V2 with four key innovations. First, a Hierarchical Multi-Agent System restructures intent reasoning through coordinated collaboration, eliminating cognitive duplication while enabling diverse intent coverage. Combined with Hybrid Representation Inference that compresses user-behavior contexts, our framework reduces GPU consumption by 60% and improves exclusive recall from 9.39% to 10.99%. Second, a Meta-Prompting framework dynamically generates contextually adaptive prompts, improving explanation diversity by +7.3%. Third, constrained reinforcement learning mitigates multi-reward conflicts, achieving +24.1% improvement in tag prediction and +13.0% in explanation acceptance. Fourth, an Agent-as-a-Judge framework decomposes assessment into multi-step reasoning, improving human preference alignment. Online A/B tests on Taobao demonstrate significant improvements: +2.98% CTR, +3.71% IPV, +2.19% TV, and +11.46% NER. RecGPT-V2 establishes both the technical feasibility and commercial viability of deploying LLM-powered intent reasoning at scale, bridging the gap between cognitive exploration and industrial utility.

Problem

• Improves computational efficiency and reduces redundancy in intent reasoning
• Enhances explanation diversity through adaptive prompt generation
• Addresses generalization and evaluation gaps in LLM-based recommendation systems

Innovation

• Hierarchical Multi-Agent System for efficient intent reasoning
• Meta-Prompting framework for adaptive and diverse explanations
• Constrained reinforcement learning to optimize multi-reward outcomes