OR-Space: A Full-Lifecycle Workspace Benchmark for Industrial Optimization Agents

Large language model (LLM) agents are increasingly used to assist with operations research (OR) modeling, yet existing OR-oriented benchmarks often reduce evaluation to one-shot translation from a self-contained problem statement into a mathematical formulation or solver program. Such settings abstract away two characteristics of real industrial OR workflows: persistent multi-artifact workspaces and multi-stage task lifecycles. We introduce OR-Space, a full-lifecycle workspace benchmark for evaluating industrial optimization agents across model construction, model revision, and grounded explanation. Each instance is an executable workspace containing business documents, structured data, optional code artifacts, solver outputs, and task-specific evaluators distributed across interdependent files. OR-Space defines three task modes: Build, where agents construct solver-ready optimization models from heterogeneous artifacts; Revise, where agents modify existing models under changing requirements or solver feedback while preserving valid prior logic; and Explain, where agents answer grounded questions about solutions, constraints, and business implications using evidence spread across workspace artifacts. By combining persistent workspaces with lifecycle-oriented tasks, OR-Space evaluates whether agents can perform reliable optimization work beyond end-to-end text generation. We describe the benchmark design, evaluation protocol, and quality-control pipeline, and position OR-Space as a benchmark for studying the reliability, failure modes, and practical readiness of LLM agents in industrial OR workflows.

OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention

Linear attention and state-space models offer constant-memory alternatives to softmax attention, but often struggle with in-context associative recall. The Delta Rule mitigates this by writing each token via one step of online gradient descent. However, its step size relies on a single scalar gate that ignores the feature-wise curvature of the inner objective. We propose Online Scaled DeltaNet (OSDN), which augments the scalar gate with a diagonal preconditioner updated online via hypergradient feedback. Crucially, this right-preconditioning is algebraically equivalent to a per-feature scaling of the write-side key. This equivalence allows OSDN to strictly preserve the hardware-friendly chunkwise parallel pipeline of DeltaNet without incurring high-dimensional state overhead. Theoretically, by exploiting the exact-quadratic structure of the inner regression loss, we establish super-geometric convergence against a right-Newton comparator and prove an algorithm-aligned token-local residual contraction bound. To handle non-stationary contexts, we further introduce Adaptive Preconditioner Forgetting (APF) to dynamically refresh stale calibration. Empirically, OSDN demonstrates strong performance across scales. At the 340M-parameter scale, OSDN improves JRT-style in-context recall by 32% over DeltaNet. Scaling to 1.3B parameters, it achieves a 39% reduction in the recall residual ratio while maintaining parity on general downstream tasks (e.g., perplexity and LongBench) -- demonstrating that our online-preconditioning mechanism effectively transfers and amplifies at the billion-parameter scale.

A Single-Sample Polylogarithmic Regret Bound for Nonstationary Online Linear Programming

We study nonstationary Online Linear Programming (OLP), where $n$ orders arrive sequentially with reward-resource consumption pairs that form a sequence of independent, but not necessarily identically distributed, random vectors. At the beginning of the planning horizon, the decision-maker is provided with a resource endowment that is sufficient to fulfill a significant portion of the requests. The decision-maker seeks to maximize the expected total reward by making immediate and irrevocable acceptance or rejection decisions for each order, subject to this resource endowment. We focus on the challenging single-sample setting, where only one sample from each of the $n$ distributions is available at the start of the planning horizon. We propose a novel re-solving algorithm that integrates a dynamic programming perspective with the dual-based frameworks traditionally employed in stationary environments. In the large-resource regime, where the resource endowment scales linearly with the number of orders, we prove that our algorithm achieves $O((\log n)^2)$ regret across a broad class of nonstationary distribution sequences. Our results demonstrate that polylogarithmic regret is attainable even under significant environmental shifts and minimal data availability, bridging the gap between stationary OLP and more volatile real-world resource allocation problems.

A Universal Load Balancing Principle and Its Application to Large Language Model Serving

Load balancing-the allocation of work across parallel resources to reduce delay, energy and cost-is a pervasive challenge in science and engineering, from large-scale simulation and data processing to cloud and manufacturing operations. Motivated by the emerging bottleneck in large language model (LLM) serving, we study a particularly stringent regime of load balancing that arises in barrier-synchronized, stateful systems: work cannot be freely migrated and progress is gated by the slowest participant at each step, so heterogeneity and temporal drift in workloads create persistent stragglers and substantial idle time. LLM serving under data-parallel decoding provides a prominent modern instance: in production traces, barrier-induced idle can exceed 40% of compute time per decode step. Here we develop a universal load-balancing principle, which admits a step-wise finite-horizon integer-optimization formulation and yields worst-case guarantees: across LLM decode models and a broader class of non-decreasing workload drift processes, it reduces long-run imbalance by a factor that grows with batch size and system scale. Extensive experiments corroborate the theory, showing substantial improvements in throughput and latency together with reductions in energy consumption. These results provide a general, theoretically grounded framework for load balancing, with immediate implications for sustainable LLM serving and broad relevance to other synchronization-gated resource-allocation problems.

Adaptively Robust LLM Inference Optimization under Prediction Uncertainty

We study the problem of optimizing Large Language Model (LLM) inference scheduling to minimize total latency. LLM inference is an online and multi-task service process and also heavily energy consuming by which a pre-trained LLM processes input requests and generates output tokens sequentially. Therefore, it is vital to improve its scheduling efficiency and reduce the power consumption while a great amount of prompt requests are arriving. A key challenge in LLM inference scheduling is that while the prompt length is known upon arrival, the output length, which critically impacts memory usage and processing time, is unknown. To address this uncertainty, we propose algorithms that leverage machine learning to predict output lengths, assuming the prediction provides an interval classification (min-max range) for each request. We first design a conservative algorithm, $\mathcal{A}_{\max}$, which schedules requests based on the upper bound of predicted output lengths to prevent memory overflow. However, this approach is overly conservative: as prediction accuracy decreases, performance degrades significantly due to potential overestimation. To overcome this limitation, we propose $\mathcal{A}_{\min}$, an adaptive algorithm that initially treats the predicted lower bound as the output length and dynamically refines this estimate during inferencing. We prove that $\mathcal{A}_{\min}$ achieves a log-scale competitive ratio. Through numerical simulations, we demonstrate that $\mathcal{A}_{\min}$ often performs nearly as well as the hindsight scheduler, highlighting both its efficiency and robustness in practical scenarios. Moreover, $\mathcal{A}_{\min}$ relies solely on the lower bound of the prediction interval--an advantageous design choice since upper bounds on output length are typically more challenging to predict accurately.

LLM Serving Optimization with Variable Prefill and Decode Lengths

We study the problem of serving LLM (Large Language Model) requests where each request has heterogeneous prefill and decode lengths. In LLM serving, the prefill length corresponds to the input prompt length, which determines the initial memory usage in the KV cache. The decode length refers to the number of output tokens generated sequentially, with each additional token increasing the KV cache memory usage by one unit. Given a set of n requests, our goal is to schedule and process them to minimize the total completion time. We show that this problem is NP-hard due to the interplay of batching, placement constraints, precedence relationships, and linearly increasing memory usage. We then analyze commonly used scheduling strategies in practice, such as First-Come-First-Serve (FCFS) and Shortest-First (SF), and prove that their competitive ratios scale up sublinearly with the memory limit-a significant drawback in real-world settings where memory demand is large. To address this, we propose a novel algorithm based on a new selection metric that efficiently forms batches over time. We prove that this algorithm achieves a constant competitive ratio. Finally, we develop and evaluate a few algorithm variants inspired by this approach, including dynamic programming variants, local search methods, and an LP-based scheduler, demonstrating through comprehensive simulations that they outperform standard baselines while maintaining computational efficiency.

Gradient Methods with Online Scaling Part I. Theoretical Foundations

This paper establishes the theoretical foundations of the online scaled gradient methods (OSGM), a framework that utilizes online learning to adapt stepsizes and provably accelerate first-order methods. OSGM quantifies the effectiveness of a stepsize by a feedback function motivated from a convergence measure and uses the feedback to adjust the stepsize through an online learning algorithm. Consequently, instantiations of OSGM achieve convergence rates that are asymptotically no worse than the optimal stepsize. OSGM yields desirable convergence guarantees on smooth convex problems, including 1) trajectory-dependent global convergence on smooth convex objectives; 2) an improved complexity result on smooth strongly convex problems, and 3) local superlinear convergence. Notably, OSGM constitutes a new family of first-order methods with non-asymptotic superlinear convergence, joining the celebrated quasi-Newton methods. Finally, OSGM explains the empirical success of the popular hypergradient-descent heuristic in optimization for machine learning.

Adaptive Resolving Methods for Reinforcement Learning with Function Approximations

Reinforcement learning (RL) problems are fundamental in online decision-making and have been instrumental in finding an optimal policy for Markov decision processes (MDPs). Function approximations are usually deployed to handle large or infinite state-action space. In our work, we consider the RL problems with function approximation and we develop a new algorithm to solve it efficiently. Our algorithm is based on the linear programming (LP) reformulation and it resolves the LP at each iteration improved with new data arrival. Such a resolving scheme enables our algorithm to achieve an instance-dependent sample complexity guarantee, more precisely, when we have $N$ data, the output of our algorithm enjoys an instance-dependent $\tilde{O}(1/N)$ suboptimality gap. In comparison to the $O(1/\sqrt{N})$ worst-case guarantee established in the previous literature, our instance-dependent guarantee is tighter when the underlying instance is favorable, and the numerical experiments also reveal the efficient empirical performances of our algorithms.

Solver-Informed RL: Grounding Large Language Models for Authentic Optimization Modeling

Optimization modeling is fundamental to decision-making across diverse domains.Despite progress in automating optimization formulation from natural language descriptions, Large Language Models (LLMs) often struggle to generate formally correct and usable models due to hallucinations, posing a challenge for reliable automation. Inspired by the success of Reinforcement Learning (RL) in enhancing Large Reasoning Models, we present Solver-Informed Reinforcement Learning (SIRL).This novel framework leverages external optimization solvers as verifiable reward mechanisms to significantly improve the authenticity of LLMs for optimization modeling.Acting as precise verifiers, these solvers automatically assess the executable code and the instance-level mathematical model represented by the associated LP file, yielding precise and comprehensive feedback signals -- including syntax, feasibility, and solution quality that directly inform the RL process. This automated verification process, powered by classic optimization solvers, also underpins our instance-enhanced self-consistency method to synthesize high-quality training data. Extensive experiments on diverse public benchmarks demonstrate that SIRL achieves state-of-the-art performance, substantially outperforming existing methods in generating accurate and executable optimization models.

Beyond $\mathcal{O}(\sqrt{T})$ Regret: Decoupling Learning and Decision-making in Online Linear Programming

Online linear programming plays an important role in both revenue management and resource allocation, and recent research has focused on developing efficient first-order online learning algorithms. Despite the empirical success of first-order methods, they typically achieve a regret no better than $\mathcal{O} ( \sqrt{T} )$, which is suboptimal compared to the $\mathcal{O} (\log T)$ bound guaranteed by the state-of-the-art linear programming (LP)-based online algorithms. This paper establishes a general framework that improves upon the $\mathcal{O} ( \sqrt{T} )$ result when the LP dual problem exhibits certain error bound conditions. For the first time, we show that first-order learning algorithms achieve $o( \sqrt{T} )$ regret in the continuous support setting and $\mathcal{O} (\log T)$ regret in the finite support setting beyond the non-degeneracy assumption. Our results significantly improve the state-of-the-art regret results and provide new insights for sequential decision-making.