The field of bandit learning and game theory is witnessing significant developments, with a focus on improving the efficiency and fairness of algorithms in complex decision-making scenarios. Researchers are exploring new approaches to address the challenges of learning in structured bandits, including the development of novel probing frameworks and the analysis of computational hardness. The incorporation of concepts from game theory, such as Nash Equilibrium, is also leading to innovative solutions for multi-agent systems. Furthermore, the study of learnability in bandit settings is revealing new insights into the boundaries of what can be learned, and the importance of considering computational observations and restricted input sources. Noteworthy papers in this area include:

• A study on two-player zero-sum games with bandit feedback, which proposes and analyzes two algorithms with instance-dependent regret upper bounds.
• An investigation into the hardness of bandit learning, which shows that no combinatorial dimension can characterize bandit learnability and constructs a reward function class with inherent computational hardness.
• A proposal for a multi-agent multi-armed bandit framework with a probing strategy to ensure fair outcomes, which achieves sublinear regret and outperforms baseline methods.