Sample efficiency in Reinforcement Learning (RL) has traditionally been driven by algorithmic enhancements. In this work, we demonstrate that scaling can also lead to substantial improvements. We conduct a thorough investigation into the interplay of scaling model capacity and domain-specific RL enhancements. These empirical findings inform the design choices underlying our proposed BRO (Bigger, Regularized, Optimistic) algorithm. The key innovation behind BRO is that strong regularization allows for effective scaling of the critic networks, which, paired with optimistic exploration, leads to superior performance. BRO achieves state-of-the-art results, significantly outperforming the leading model-based and model-free algorithms across 40 complex tasks from the DeepMind Control, MetaWorld, and MyoSuite benchmarks. BRO is the first model-free algorithm to achieve near-optimal policies in the notoriously challenging Dog and Humanoid tasks.

The level of interest in Galois Counter Mode (GCM) Authenticated Encryption rose significantly within the last few years. GCM is interesting because it is the only authenticated encryption standard that can be implemented in a fully pipelined or parallelized way and it is the most appropriate for encrypting packetized data. McGrew and Viega [CHECK END OF SENTENCE] described (but did not detail) how GHASH can be implemented with more than one multiplier operating in parallel. This paper details how that can be done and shows that, when N multipliers are used, and the multipliers use the approach of multiplying polynomials then applying a modular reduction, a single modular reduction can be used instead on N separate operations. This optimization can be used even when there is a single multiplier, which makes this implementation strategy have a broader appeal. Recently Intel has introduced new ISA instructions into the next generation CPU core, namely: AES family and PCLMULQDQ operating in XMM registers domain. In this paper, we discuss the example implementation of proposed GHASH modifications using above instructions.