Strategy Logic
Strategy logic focuses on formalizing and analyzing strategic reasoning in multi-agent systems, aiming to model and predict the outcomes of interactions where agents pursue their own objectives. Current research emphasizes applying and extending these frameworks to diverse domains, including game playing (using LLMs and other models), robotics control (e.g., via deep reinforcement learning), and cybersecurity, often focusing on improving the efficiency and robustness of algorithms. This field is significant for its potential to improve the design of autonomous systems, enhance the understanding of complex interactions, and provide rigorous tools for evaluating the strategic capabilities of AI agents.
Papers
Automated Security Response through Online Learning with Adaptive Conjectures
Kim Hammar, Tao Li, Rolf Stadler, Quanyan Zhu
GTBench: Uncovering the Strategic Reasoning Limitations of LLMs via Game-Theoretic Evaluations
Jinhao Duan, Renming Zhang, James Diffenderfer, Bhavya Kailkhura, Lichao Sun, Elias Stengel-Eskin, Mohit Bansal, Tianlong Chen, Kaidi Xu