This section consists of three parts, with each subsequent part building upon the previous one. Appendix A.1 covers the fundamentals of RL, where the actor-critic method is introduced. Appendix A.2 describes the RL algorithm for a single fulfillment agent, which is the proximal policy Appendix A.3 presents the MARL algorithm for the Currently, policy-based methods [Deisenroth et al., 2013] are prevalent because they are compatible with stochastic To sum up, the complete procedure is given in Algorithm 1.Algorithm 1 Heterogeneous Multi-Agent Reinforcement Learning for Order Fulfillment. With regard to the advantage estimator, we set the GAE parameters [Schulman et al., 2016] To highlight how our proposed benchmark differs from existing approaches focused on sub-tasks of order fulfillment, we compare the objectives, observations, and actions in Table 1. It should be noted that multiple formulations exist for each sub-task.

We investigate batch policy optimization for cost-sensitive classification and contextual bandits---two related tasks that obviate exploration but require generalizing from observed rewards to action selections in unseen contexts. When rewards are fully observed, we show that the expected reward objective exhibits suboptimal plateaus and exponentially many local optima in the worst case. To overcome the poor landscape, we develop a convex surrogate that is calibrated with respect to entropy regularized expected reward. We then consider the partially observed case, where rewards are recorded for only a subset of actions. Here we generalize the surrogate to partially observed data, and uncover novel objectives for batch contextual bandit training. We find that surrogate objectives remain provably sound in this setting and empirically demonstrate state-of-the-art performance.

This study presents a systematic comparison of three Reinforcement Learning (RL) algorithms (PPO, GRPO, and DAPO) for improving complex reasoning in large language models (LLMs). Our main contribution is a controlled transfer-learning evaluation: models are first fine-tuned on the specialized Countdown Game and then assessed on a suite of general-purpose reasoning benchmarks. Across all tasks, RL-trained models outperform their corresponding base models, although the degree of improvement differs by benchmark. Our parametric analysis offers practical guidance for RL-based LLM training. Increasing the group size in GRPO and DAPO leads to more stable training dynamics and higher accuracy, while the impact of the KL-penalty coefficient is non-monotonic. Additionally, we find that the Dynamic Sampling (DS) component in DAPO does not improve performance; in fact, the best overall results are achieved with DAPO when DS is disabled.

Deep neural networks often fail when deployed in real-world contexts due to distribution shift, a critical barrier to building safe and reliable systems. An emerging approach to address this problem relies on \emph{disagreement discrepancy} -- a measure of how the disagreement between two models changes under a shifting distribution. The process of maximizing this measure has seen applications in bounding error under shifts, testing for harmful shifts, and training more robust models. However, this optimization involves the non-differentiable zero-one loss, necessitating the use of practical surrogate losses. We prove that existing surrogates for disagreement discrepancy are not Bayes consistent, revealing a fundamental flaw: maximizing these surrogates can fail to maximize the true disagreement discrepancy. To address this, we introduce new theoretical results providing both upper and lower bounds on the optimality gap for such surrogates. Guided by this theory, we propose a novel disagreement loss that, when paired with cross-entropy, yields a provably consistent surrogate for disagreement discrepancy. Empirical evaluations across diverse benchmarks demonstrate that our method provides more accurate and robust estimates of disagreement discrepancy than existing approaches, particularly under challenging adversarial conditions.

Partial client participation has been widely adopted in Federated Learning (FL) to reduce the communication burden efficiently.