Game-Theoretic Multi-Agent Control for Robust Contextual Reasoning in LLMs

Abstract

Large Language Models (LLMs) in multi-turn interactions maintain evolving context rather than generating isolated responses, making them vulnerable to prompt-injection and context-poisoning attacks in which locally plausible adversarial fragments gradually distort reasoning trajectories. Existing defenses mainly filter individual outputs and often ignore context evolution across turns, leaving long-horizon reasoning exposed. Although the Model Context Protocol (MCP) standardizes context exchange

Categories

Framework mappings

Suggested from the entry's categories.

@misc{jamshidi2026gametheoretic,
  title = {{Game-Theoretic Multi-Agent Control for Robust Contextual Reasoning in LLMs}},
  author = {Saeid Jamshidi and Amin Nikanjam and Arghavan Moradi Dakhel and Kawser Wazed Nafi and Foutse Khomh},
  year = {2026},
  month = jun,
  eprint = {2606.10322},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.10322}
}