Shape Your Feed: An LLM-based Agentic System for Conversational Recommendation

2026-08-06

AI Summary

Traditional recommender systems rely on passive behavioral signals, which struggle to capture users’ fine-grained and dynamic preferences expressed through natural language, often leading to recommendations misaligned with true intent. This work proposes an LLM-based agent-driven recommendation framework featuring a tripartite architecture—comprising perception, service, and self-evolution streams—that enables an interactive recommendation system integrating multimodal inputs, continuously updated semantic user profiles, and LLM-aligned preference modeling. The framework empowers users to actively shape their information feeds and incorporates agent-based reranking, pruning mechanisms, and a self-evolution optimization strategy leveraging DPO and LLM-as-a-Judge. Offline evaluation demonstrates a 98.85% alignment accuracy, substantially outperforming few-shot baselines, while online A/B experiments confirm significant improvements in recommendation relevance and user satisfaction.

Abstract

Industrial recommendation systems predominantly adopt a passive ranking paradigm that infers user preferences from implicit behavioral signals (e.g., clicks, dwell time) rather than explicit, natural language inputs. As a result, users experience a persistent discrepancy between their explicit interests and what passive behavioral algorithms deliver, limiting their ability to express nuanced preferences or steer their feed in real time. To address this growing gap between how recommendations are optimized and how users wish to articulate their interests, we present Shape Your Feed (SYF), an LLM-based agentic recommendation framework that enables real-time, multimodal co-curation of content. SYF employs a three-tier architecture: (i) a Perception Flow that captures fine-grained user intent from text prompts, voice commands, and UI interactions; (ii) a Serving Flow that performs real-time agentic re-ranking and pruning of candidate items, grounded in a persistent Semantic Profile encoding evolving user preferences; and (iii) a Self-Evolution Flow that aligns system behavior with human judgments via Direct Preference Optimization (DPO) and an LLM-as-a-Judge ensemble. Offline evaluations show that SYF's alignment scoring module achieves 98.85% accuracy, substantially improving over strong few-shot baselines. Large-scale online A/B experiments on production traffic further demonstrate that SYF improves feed relevance and user sentiment, indicating a practical and scalable path toward interactive, user-steerable recommendation in industrial settings.

Problem

• conversational recommendation
• user intent
• passive ranking
• recommendation systems
• natural language input

Innovation

• LLM-based agentic recommendation
• real-time co-curation
• Semantic Profile
• Direct Preference Optimization (DPO)
• multimodal user intent