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

2026-05-27

AI Summary

Existing benchmarks for operations research (OR) optimization overlook the realistic characteristics of industrial settings—specifically, the multi-stage task lifecycle and persistent, multi-artifact workspaces—rendering them inadequate for reliably evaluating large language model (LLM) agents in practical optimization workflows. To address this gap, this work proposes OR-Space, a novel benchmark that introduces, for the first time, a full-lifecycle evaluation paradigm with persistent, multi-artifact workspaces. By integrating business documents, structured data, code, and solver outputs into an executable workspace, OR-Space defines three core tasks—Build, Revise, and Explain—to comprehensively assess agent capabilities across modeling, refinement, and explanation phases. Leveraging multi-source heterogeneous artifact integration, task-specific evaluators, and cross-file evidence tracing, the benchmark enables end-to-end evaluation, establishing a systematic foundation for assessing the reliability, failure modes, and production readiness of LLMs in industrial OR applications.

Abstract

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.

Problem

• operations research
• LLM agents
• benchmark
• industrial optimization
• workspace

Innovation

• full-lifecycle benchmark
• optimization agents
• persistent workspace
• model revision
• grounded explanation