The field of trajectory forecasting and multi-agent systems is moving towards the development of more efficient and interpretable models. Researchers are leveraging techniques such as Koopman operator theory to enable linear representation of nonlinear dynamics, allowing for more accurate predictions and better understanding of complex systems. Decentralized cooperative multi-agent reinforcement learning is also an area of focus, with novel methods being proposed to address issues such as non-stationarity and relative overgeneralization. Additionally, distributed state estimation and control methods are being developed for multi-agent systems, enabling agents to reconstruct the state of the system and make cooperative decisions. Noteworthy papers include: KoopCast, which presents a lightweight yet efficient model for trajectory forecasting, and Fully Decentralized Cooperative Multi-Agent Reinforcement Learning is A Context Modeling Problem, which proposes a novel method for addressing non-stationarity and relative overgeneralization in decentralized cooperative multi-agent reinforcement learning. Distributed Koopman Operator Learning from Sequential Observations is also notable, as it presents a distributed framework for learning Koopman operators from sequential observations.