In large-scale multi-agent systems with shared resource constraints, an upstream planner must iteratively evaluate candidate resource plans – assessing feasibility, aggregate response, and marginal cost – before committing to one. Lagrangian relaxation separates local decisions through a broadcast cost signal, but the planner still needs the cost-to-utilization response map to explore plan space, and this map depends on population composition that changes across planning cycles. We propose population-aware coordination interfaces: learned primal and dual maps, conditioned on compact population summaries, that the planner queries inside its iterative loop. The primal map predicts aggregate utilization under a proposed cost trajectory; the dual map predicts the cost trajectory for a target plan. By encoding response-relevant population structure, these maps remain reliable across evolving populations without per-cycle retraining, and support coordination of large populations from compact subsamples. We additionally cast Sim2Real transfer as a backtestable procedure, enabling evaluation before deployment. In a supply-chain capacity-control case study, population-aware interfaces reduce forecast error by 16–19% and capacity violations by 20–51% relative to population-unaware baselines under composition shift; 20K-agent cohorts support accurate coordination of 500K-agent populations; and simulator-trained primal maps achieve 11.1% MAPE on real observations versus 13–24% for baselines.

This paper investigates the application of Deep Reinforcement Learning (DRL) to classical inventory management problems, with a focus on practical implementation considerations. We apply a DRL algorithm based on DirectBackprop to several fundamental inventory management scenarios including multi-period systems with lost sales (with and without lead times), perishable inventory management, dual sourcing, and joint inventory procurement and removal. The DRL approach learns policies across products using only historical information that would be available in practice, avoiding unrealistic assumptions about demand distributions or access to distribution parameters. We demonstrate that our generic DRL implementation performs competitively against or outperforms established benchmarks and heuristics across these diverse settings, while requiring minimal parameter tuning. Through examination of the learned policies, we show that the DRL approach naturally captures many known structural properties of optimal policies derived from traditional operations research methods. To further improve policy performance and interpretability, we propose a Structure-Informed Policy Network technique that explicitly incorporates analytically-derived characteristics of optimal policies into the learning process. This approach can help interpretability and add robustness to the policy in out-of-sample performance, as we demonstrate in an example with realistic demand data. Finally, we provide an illustrative application of DRL in a non-stationary setting. Our work bridges the gap between data-driven learning and analytical insights in inventory management while maintaining practical applicability.

Large language models (LLMs) are good at generating code, but remain brittle for formal verification in systems like Lean4. A core scalability challenge is that verified synthesis requires consistent outputs across multiple artifacts: executable code, precise specifications, theorem statements, and ultimately proofs. Existing approaches rarely treat these as a unified pipeline. We present BRIDGE, a structured prompting framework that decomposes verification into three interconnected domains: Code (implementations), Specifications (formal intent), and Theorem Statements (constructive correctness claims), and elicits domain-specific intermediate reasoning to connect them. In Lean4, BRIDGE often adopts a code-first workflow, using the generated implementation as a semantic anchor for downstream specification and theorem statement generation. Across 178 algorithmic problems and five LLMs, BRIDGE improves Lean executable correctness by nearly 1.5x (pass at 5) over direct baselines and can be 2x more sample-efficient at inference time, requiring fewer samples per verified solution at comparable generation lengths. We further find that specification-driven prompting improves Python pass rates by up to 17.5 percent. Beyond inference-time prompting, supervised fine-tuning on BRIDGE-style reasoning traces yields nearly 1.5x higher Lean pass success than code-only SFT, indicating that these intermediate representations are learnable. BRIDGE provides a practical foundation for scaling verified synthesis and motivates future work on expert iteration and full proof generation.

Self-improvement is a mechanism in Large Language Model (LLM) pre-training, post-training and test-time inference. We explore a framework where the model verifies its own outputs, filters or reweights data based on this verification, and distills the filtered data. Despite several empirical successes, a fundamental understanding is still lacking. In this work, we initiate a comprehensive, modular and controlled study on LLM self-improvement. We provide a mathematical formulation for self-improvement, which is largely governed by a quantity which we formalize as the **generation-verification gap**. Through experiments with various model families and tasks, we discover a scaling phenomenon of self-improvement -- a variant of the generation-verification gap scales monotonically with the model pre-training flops. We also examine when self-improvement is possible, an iterative self-improvement procedure, and ways to improve its performance. Our findings not only advance understanding of LLM self-improvement with practical implications, but also open numerous avenues for future research into its capabilities and boundaries.

In this work, we study how to efficiently apply reinforcement learning (RL) for solving large-scale stochastic optimization problems by leveraging intervention models. The key of the proposed methodology is to better explore the solution space by simulating and composing the stochastic processes using pre-trained deep learning (DL) models. We demonstrate our approach on a challenging real-world application, the multi-sourcing multi-period inventory management problem in supply chain optimization. In particular, we employ deep RL models for learning and forecasting the stochastic supply chain processes under a range of assumptions. Moreover, we also introduce a constraint coordination mechanism, designed to forecast dual costs given the cross-products constraints in the inventory network. We highlight that instead of directly modeling the complex physical constraints into the RL optimization problem and solving the stochastic problem as a whole, our approach breaks down those supply chain processes into scalable and composable DL modules, leading to improved performance on large real-world datasets. We also outline open problems for future research to further investigate the efficacy of such models.

We study the problem of forecasting the number of units fulfilled (or “drained”) from each inventory warehouse to meet customer demand, along with the associated outbound shipping costs. The actual drain and shipping costs are determined by complex production systems that manage the planning and execution of customers' orders fulfillment, i.e. from where and how to ship a unit to be delivered to a customer. Accurately modeling these processes is critical for regional inventory planning, especially when using Reinforcement Learning (RL) to develop control policies. For the RL usecase, a drain model is incorporated into a simulator to produce long rollouts, which we desire to be differentiable. While simulating the calls to the internal software systems can be used to recover this transition, they are non-differentiable and too slow and costly to run within an RL training environment. Accordingly, we frame this as a probabilistic forecasting problem, modeling the joint distribution of outbound drain and shipping costs across all warehouses at each time period, conditioned on inventory positions and exogenous customer demand. To ensure robustness in an RL environment, the model must handle out-of-distribution scenarios that arise from off-policy trajectories. We propose a validation scheme that leverages production systems to evaluate the drain model on counterfactual inventory states induced by RL policies. Preliminary results demonstrate the model's accuracy within the in-distribution setting.

This paper addresses the capacitated periodic review inventory control problem, focusing on a retailer managing multiple products with limited shared resources, such as storage or inbound labor at a facility. Specifically, this paper is motivated by the questions of (1) what does it mean to backtest a capacity control mechanism, (2) can we devise and backtest a capacity control mechanism that is compatible with recent advances in deep reinforcement learning for inventory management? First, because we only have a single historic sample path of Amazon's capacity limits, we propose a method that samples from a distribution of possible constraint paths covering a space of real-world scenarios. This novel approach allows for more robust and realistic testing of inventory management strategies. Second, we extend the exo-IDP (Exogenous Decision Process) formulation of Madeka et al. 2022 to capacitated periodic review inventory control problems and show that certain capacitated control problems are no harder than supervised learning. Third, we introduce a `neural coordinator', designed to produce forecasts of capacity prices, guiding the system to adhere to target constraints in place of a traditional model predictive controller. Finally, we apply a modified DirectBackprop algorithm for learning a deep RL buying policy and a training the neural coordinator. Our methodology is evaluated through large-scale backtests, demonstrating RL buying policies with a neural coordinator outperforms classic baselines both in terms of cumulative discounted reward and capacity adherence (we see improvements of up to 50

We apply deep reinforcement learning (RL) to solve the periodic review inventory control problem with general arrival dynamics. In this work, we incorporate a learned model of transition dynamics (inventory arrivals) into the inventory control problem formulation, increasing the fidelity of the resulting simulator. Leveraging recent results (Madeka et al., 2022), we demonstrate a reduction of the complexity of the inventory control problem we consider to that of supervised learning, proving that under mild assumptions our backtest of inventory control policies is accurate. We also propose several metrics by which to evaluate the inventory arrivals model, and demonstrate the impact of an improved arrivals model on policy performance via a comparison of policies learned on our simulator with one learned on a simulator with less accurate arrivals dynamics. Finally, we use data from a real world A/B test of an RL agent trained using our simulator with learned dynamics to evaluate the performance of the arrivals model, showing that empirically it generalizes well to the off-policy state distribution induced by the RL agent in an actual supply chain.

In this paper we address the problem of learning and backtesting inventory control policies in the presence of general arrival dynamics -- which we term as a quantity-over-time arrivals model (QOT). We also allow for order quantities to be modified as a post-processing step to meet vendor constraints such as order minimum and batch size constraints -- a common practice in real supply chains. To the best of our knowledge this is the first work to handle either arbitrary arrival dynamics or an arbitrary downstream post-processing of order quantities. Building upon recent work (Madeka et al., 2022) we similarly formulate the periodic review inventory control problem as an exogenous decision process, where most of the state is outside the control of the agent. Madeka et al., 2022 show how to construct a simulator that replays historic data to solve this class of problem. In our case, we incorporate a deep generative model for the arrivals process as part of the history replay. By formulating the problem as an exogenous decision process, we can apply results from Madeka et al., 2022 to obtain a reduction to supervised learning. Via simulation studies we show that this approach yields statistically significant improvements in profitability over production baselines. Using data from a real-world A/B test, we show that Gen-QOT generalizes well to off-policy data and that the resulting buying policy outperforms traditional inventory management systems in real world settings.

Recent advances in neural forecasting have produced major improvements in accuracy for probabilistic demand prediction. In this work, we propose novel improvements to the current state of the art by incorporating changes inspired by recent advances in Transformer architectures for Natural Language Processing. We develop a novel decoder-encoder attention for context-alignment, improving forecasting accuracy by allowing the network to study its own history based on the context for which it is producing a forecast. We also present a novel positional encoding that allows the neural network to learn context-dependent seasonality functions as well as arbitrary holiday distances. Finally we show that the current state of the art MQ-Forecaster (Wen et al., 2017) models display excess variability by failing to leverage previous errors in the forecast to improve accuracy. We propose a novel decoder-self attention scheme for forecasting that produces significant improvements in the excess variation of the forecast.

Multi-horizon probabilistic time series forecasting has wide applicability to real-world tasks such as demand forecasting. Recent work in neural time-series forecasting mainly focus on the use of Seq2Seq architectures. For example, MQTransformer - an improvement of MQCNN - has shown the state-of-the-art performance in probabilistic demand forecasting. In this paper, we consider incorporating cross-entity information to enhance model performance by adding a cross-entity attention mechanism along with a retrieval mechanism to select which entities to attend over. We demonstrate how our new neural architecture, MQRetNN, leverages the encoded contexts from a pretrained baseline model on the entire population to improve forecasting accuracy. Using MQCNN as the baseline model (due to computational constraints, we do not use MQTransformer), we first show on a small demand forecasting dataset that it is possible to achieve ~3% improvement in test loss by adding a cross-entity attention mechanism where each entity attends to all others in the population. We then evaluate the model with our proposed retrieval methods - as a means of approximating an attention over a large population - on a large-scale demand forecasting application with over 2 million products and observe ~1% performance gain over the MQCNN baseline.

This work provides a Deep Reinforcement Learning approach to solving a periodic review inventory control system with stochastic vendor lead times, lost sales, correlated demand, and price matching. While this dynamic program has historically been considered intractable, our results show that several policy learning approaches are competitive with or outperform classical methods. In order to train these algorithms, we develop novel techniques to convert historical data into a simulator. On the theoretical side, we present learnability results on a subclass of inventory control problems, where we provide a provable reduction of the reinforcement learning problem to that of supervised learning. On the algorithmic side, we present a model-based reinforcement learning procedure (Direct Backprop) to solve the periodic review inventory control problem by constructing a differentiable simulator. Under a variety of metrics Direct Backprop outperforms model-free RL and newsvendor baselines, in both simulations and real-world deployments.

Motivated by modern applications in which one constructs graphical models based on a very large number of features, this paper introduces a new class of cluster-based graphical models, in which variable clustering is applied as an initial step for reducing the dimension of the feature space. We employ model assisted clustering, in which the clusters contain features that are similar to the same unobserved latent variable. Two different cluster-based Gaussian graphical models are considered: the latent variable graph, corresponding to the graphical model associated with the unobserved latent variables, and the cluster-average graph, corresponding to the vector of features averaged over clusters. Our study reveals that likelihood based inference for the latent graph, not analyzed previously, is analytically intractable. Our main contribution is the development and analysis of alternative estimation and inference strategies, for the precision matrix of an unobservable latent vector $Z$. We replace the likelihood of the data by an appropriate class of empirical risk functions, that can be specialized to the latent graphical model and to the simpler, but under-analyzed, cluster-average graphical model. The estimators thus derived can be used for inference on the graph structure, for instance on edge strength or pattern recovery. Inference is based on the asymptotic limits of the entry-wise estimates of the precision matrices associated with the conditional independence graphs under consideration. While taking the uncertainty induced by the clustering step into account, we establish Berry-Esseen central limit theorems for the proposed estimators. It is noteworthy that, although the clusters are estimated adaptively from the data, the central limit theorems regarding the entries of the estimated graphs are proved under the same conditions one would use if the clusters were known....

Motivated by the task of clustering either d variables or d points into K groups, we investigate efficient algorithms to solve the Peng–Wei (P–W) K-means semi-definite programming (SDP) relaxation. The P–W SDP has been shown in the literature to have good statistical properties in a variety of settings, but remains intractable to solve in practice. To this end we propose FORCE, a new algorithm to solve this SDP relaxation. Compared to off-the-shelf interior point solvers, our method reduces the computational complexity of solving the SDP from $${\widetilde{{\mathcal {O}}}}(d^7\log \epsilon ^{-1})$$ to $${\widetilde{{\mathcal {O}}}}(d^{6}K^{-2}\epsilon ^{-1})$$ arithmetic operations for an $$\epsilon $$ -optimal solution. Our method combines a primal first-order method with a dual optimality certificate search, which when successful, allows for early termination of the primal method. We show for certain variable clustering problems that, with high probability, FORCE is guaranteed to find the optimal solution to the SDP relaxation and provide a certificate of exact optimality. As verified by our numerical experiments, this allows FORCE to solve the P–W SDP with dimensions in the hundreds in only tens of seconds. For a variation of the P–W SDP where K is not known a priori a slight modification of FORCE reduces the computational complexity of solving this problem as well: from $${\widetilde{{\mathcal {O}}}}(d^7\log \epsilon ^{-1})$$ using a standard SDP solver to $${\widetilde{{\mathcal {O}}}}(d^{4}\epsilon ^{-1})$$ .

Many complex domains, such as robotics control and real-time strategy (RTS) games, require an agent to learn a continuous control. In the former, an agent learns a policy over $\mathbb{R}^d$ and in the latter, over a discrete set of actions each of which is parametrized by a continuous parameter. Such problems are naturally solved using policy based reinforcement learning (RL) methods, but unfortunately these often suffer from high variance leading to instability and slow convergence. Unnecessary variance is introduced whenever policies over bounded action spaces are modeled using distributions with unbounded support by applying a transformation $T$ to the sampled action before execution in the environment. Recently, the variance reduced clipped action policy gradient (CAPG) was introduced for actions in bounded intervals, but to date no variance reduced methods exist when the action is a direction, something often seen in RTS games. To this end we introduce the angular policy gradient (APG), a stochastic policy gradient method for directional control. With the marginal policy gradients family of estimators we present a unified analysis of the variance reduction properties of APG and CAPG; our results provide a stronger guarantee than existing analyses for CAPG. Experimental results on a popular RTS game and a navigation task show that the APG estimator offers a substantial improvement over the standard policy gradient.

Many complex domains, such as robotics control and real-time strategy (RTS) games, require an agent to learn a continuous control. In the former, an agent learns a policy over $\mathbb{R}^d$ and in the latter, over a discrete set of actions each of which is parametrized by a continuous parameter. Such problems are naturally solved using policy based reinforcement learning (RL) methods, but unfortunately these often suffer from high variance leading to instability and slow convergence. Unnecessary variance is introduced whenever policies over bounded action spaces are modeled using distributions with unbounded support by applying a transformation $T$ to the sampled action before execution in the environment. Recently, the variance reduced clipped action policy gradient (CAPG) was introduced for actions in bounded intervals, but to date no variance reduced methods exist when the action is a direction, something often seen in RTS games. To this end we introduce the angular policy gradient (APG), a stochastic policy gradient method for directional control. With the marginal policy gradients family of estimators we present a unified analysis of the variance reduction properties of APG and CAPG; our results provide a stronger guarantee than existing analyses for CAPG. Experimental results on a popular RTS game and a navigation task show that the APG estimator offers a substantial improvement over the standard policy gradient.

Most existing deep reinforcement learning (DRL) frameworks consider action spaces that are either discrete or continuous space. Motivated by the project of design Game AI for King of Glory (KOG), one the world’s most popular mobile game, we consider the scenario with the discrete-continuous hybrid action space. To directly apply existing DLR frameworks, existing approaches either approximate the hybrid space by a discrete set or relaxing it into a continuous set, which is usually less efficient and robust. In this paper, we propose a parametrized deep Q-network (P-DQN) for the hybrid action space without approximation or relaxation. Our algorithm combines DQN and DDPG and can be viewed as an extension of the DQN to hybrid actions. The empirical study on the game KOG validates the efficiency and effectiveness of our method.