Enhancing LLM-Based Social Bot via an Adversarial Learning Framework

2025-08-25

AI Summary

Addressing the challenge of endowing LLM-based social bots with both individual heterogeneity and social adaptability, this paper proposes EvoBot: a framework that models user-specific traits via fine-grained user profiling and enables dynamic responsiveness to evolving social network structures through neighbor-aware contextual modeling. Methodologically, it introduces an adversarial co-evolution paradigm between a generator (EvoBot) and a detector, integrating supervised fine-tuning, direct preference optimization, and adversarial feedback in iterative training on real-world social network data. Its key innovation lies in unifying personalized modeling, social context awareness, and dual-model adversarial evolution within a single coherent framework. Experiments demonstrate that EvoBot generates highly persona-consistent content, substantially improves sociological authenticity and evasion capability against bot detectors, and accurately reproduces real-world opinion dynamics and information diffusion patterns.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Adversarial Learning & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Models for Web evolution

Abstract

Developing Large Language Model (LLM) agents that exhibit human-like behavior, encompassing not only individual heterogeneity rooted in unique user profiles but also adaptive response to socially connected neighbors, is a significant research challenge. Social media platforms, with their diverse user data and explicit social structures, provide an ideal testbed for such investigations. This paper introduces EvoBot, an extbf{Evo}lving LLM-based social extbf{Bot} that significantly enhances human-like generative capabilities through a novel adversarial learning framework. EvoBot is initialized by Supervised Fine-Tuning (SFT) on representative data from social media and then iteratively refines its generation of sophisticated, human-like content via Direct Preference Optimization (DPO). This refinement is guided by feedback from a co-adapting extbf{Detector} which concurrently improves its ability to distinguish EvoBot from humans, thereby creating an increasingly challenging learning environment for EvoBot. Experiments demonstrate that EvoBot generates content aligned with diverse user profiles, increasingly bypassing the co-adapting Detector through human-like expression. Moreover, it exhibits strong social responsiveness, more accurately modeling real-world opinion dynamics and information spread in multi-agent simulations. The framework also yields a more robust Detector, underscoring its broader utility for both advanced agent development and related detection tasks. The code is available at https://github.com/kfq20/EvoBot.

Problem

• Developing LLM agents with human-like behavior and individual heterogeneity
• Enhancing social responsiveness in multi-agent opinion dynamics modeling
• Creating adversarial framework to improve both bot generation and detection

Innovation

• Adversarial learning framework for LLM refinement
• Supervised Fine-Tuning with social media data
• Direct Preference Optimization for human-like generation