AgenticTagger: Structured Item Representation for Recommendation with LLM Agents

2026-02-05

AI Summary

This work addresses the challenge of generating high-quality, structured product descriptions with large language models (LLMs), where the open-ended nature of generation often results in a high-cardinality, noisy vocabulary that poorly aligns with downstream recommendation tasks. To this end, we propose AgenticTagger, a novel framework that constructs a low-cardinality, hierarchical descriptor vocabulary and introduces a multi-agent reflective mechanism. In this mechanism, a scaffolding agent and multiple parallel annotation agents collaboratively and iteratively refine the vocabulary, guiding the LLM to accurately assign structured descriptors. Experimental results demonstrate that our approach significantly improves performance across multiple tasks—including generative retrieval, keyword-based retrieval, ranking, and feedback-driven controllable recommendation—on both public and proprietary datasets.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

High-quality representations are a core requirement for effective recommendation. In this work, we study the problem of LLM-based descriptor generation, i.e., keyphrase-like natural language item representation generation frameworks with minimal constraints on downstream applications. We propose AgenticTagger, a framework that queries LLMs for representing items with sequences of text descriptors. However, open-ended generation provides little control over the generation space, leading to high cardinality, low-performance descriptors that renders downstream modeling challenging. To this end, AgenticTagger features two core stages: (1) a vocabulary building stage where a set of hierarchical, low-cardinality, and high-quality descriptors is identified, and (2) a vocabulary assignment stage where LLMs assign in-vocabulary descriptors to items. To effectively and efficiently ground vocabulary in the item corpus of interest, we design a multi-agent reflection mechanism where an architect LLM iteratively refines the vocabulary guided by parallelized feedback from annotator LLMs that validates the vocabulary against item data. Experiments on public and private data show AgenticTagger brings consistent improvements across diverse recommendation scenarios, including generative and term-based retrieval, ranking, and controllability-oriented, critique-based recommendation.

Problem

• item representation
• LLM-based descriptor generation
• structured representation
• recommendation systems
• controlled text generation

Innovation

• AgenticTagger
• LLM agents
• structured item representation
• vocabulary grounding
• multi-agent reflection