The field of reinforcement learning is moving towards more efficient and reliable methods, particularly in offline and off-policy settings. Recent developments have focused on improving the stability and sampling efficiency of policy optimization algorithms, as well as addressing the challenges of fully offline reinforcement learning. Notable progress has been made in developing algorithms that can learn effective policies without requiring extensive online interactions, which has the potential to unlock the application of reinforcement learning in real-world settings. Additionally, there have been significant advances in off-policy evaluation, including the development of methods that can accurately estimate the performance of a policy in a new environment. Noteworthy papers include: SOReL and TOReL, which introduce two new algorithms for fully offline reinforcement learning that can accurately estimate regret and achieve competitive performance without requiring online interactions. Demystifying the Paradox of Importance Sampling, which provides a theoretical understanding of the benefits of using history-dependent behavior policy estimation in off-policy evaluation. Calibrated Value-Aware Model Learning, which analyzes the strengths and weaknesses of value-aware model learning losses and proposes corrections to ensure calibrated surrogate losses.