TagLLM: A Fine-Grained Tag Generation Approach for Note Recommendation

Abstract

Large Language Models (LLMs) have shown promising potential in E-commerce community recommendation. While LLMs and Multimodal LLMs (MLLMs) are widely used to encode notes into implicit embeddings, leveraging their generative capabilities to represent notes with interpretable tags remains unexplored. In the field of tag generation, traditional close-ended methods heavily rely on the design of tag pools, while existing open-ended methods applied directly to note recommendations face two limitations: (1) MLLMs lack guidance during generation, resulting in redundant tags that fail to capture user interests; (2) The generated tags are often coarse and lack fine-grained representation of notes, interfering with downstream recommendations. To address these limitations, we propose TagLLM, a fine-grained tag generation method for note recommendation. TagLLM captures user interests across note categories through a User Interest Handbook and constructs fine-grained tag data using multimodal CoT Extraction. A Tag Knowledge Distillation method is developed to equip small models with competitive generation capabilities, enhancing inference efficiency. In online A/B test, TagLLM increases average view duration per user by 0.31%, average interactions per user by 0.96%, and page view click-through rate in cold-start scenario by 32.37%, demonstrating its effectiveness.

Figures (18)

• Figure 1: Our Motivation for TagLLM. Traditional tag generation methods are close-ended, making them difficult to scale and heavily reliant on the tag pool design. Recent LLM-based methods lack guidance, resulting in coarse and unfocused tags. Therefore, we propose TagLLM, a fine-grained generation method that is better suited for note recommendation.
• Figure 2: The Overall Pipeline of TagLLM. First, a User Interest Handbook is generated based on different note categories to reflect user interest. Based on the Handbook, we design a CoT Extraction method to construct tag training data. Next, a Tag Knowledge Distillation method is employed to distill the tag generation capabilities into smaller models. Finally, the tags generated by small models are converted into specific features for recommendations.
• Figure 3: Representative Examples of User Interest Handbook (left) and Tag Data Construction (right). Through LLM expansion and expert refinement, the guidance in the handbook is fine-grained, covering basic category attributes and user interests. With the three-step CoT Extraction process, LLMs are able to generate high-quality tags based on the guidance.
• Figure 4: Ablation Study of Input Modalities. The evaluated model is the distilled Qwen3-VL-4B-Instruct.
• Figure 5: Failure Case Study of TagLLM. Here are two common types: Reasoning Hallucination and User Misdirection.