How Tencent Leverages LLMs: RAG, GraphRAG, and Agents in Real‑World Apps

Overview of Tencent LLM Applications

In this article we explore how Tencent applies its large language models (LLMs) to a wide range of business scenarios, aiming to enhance model intelligence and user experience. The discussion covers major use cases—including content generation, intelligent customer service, and role‑play—and delves into the technical foundations of Retrieval‑Augmented Generation (RAG), GraphRAG, and Agent technologies.

Core Business Scenarios

Content Generation: automatic creation of copy, advertising text, comment assistance, etc.

Content Understanding: text moderation, fraud detection, and other analysis tasks.

Intelligent Customer Service: knowledge‑base Q&A, user guidance, and related services.

Development Copilot: automated code review, test‑case generation, and other programming aids.

Role‑Play: intelligent NPC interactions in gaming environments.

Key Enabling Technologies

Supervised Fine‑Tuning (SFT)

SFT fine‑tunes a base LLM with domain‑specific data, embedding business knowledge directly into the model to achieve targeted task performance.

Retrieval‑Augmented Generation (RAG)

RAG combines external knowledge bases and retrieval mechanisms with generation, allowing the model to incorporate up‑to‑date factual information, improve explainability, and reduce hallucinations. Typical applications include intelligent customer service and document assistants.

Agent (Intelligent Agent)

Agents integrate external tools and multi‑step reasoning capabilities, enabling the model to plan, execute, and complete complex tasks that require sequential decision making, such as advanced workflow automation.