Newest papers

ASTOR: Benchmarking and Improving Autoformulation for Soft-Constraint Optimization with a Formal Grammar

Amer Goel, Yeganeh Alimohammadi, Andrew Daw · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26819 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Los problemas de decisión de código en el mundo real combinan dos tipos de restricciones cualitativamente diferentes: requisitos que deben ser satisfechos, llamados restricciones duras, y preferencias que pueden ser violadas, llamadas restricciones blandas. Debido a que estos problemas a menudo se plantean a los tomadores de decisiones en lengua natural, los modelos de idioma grande ofrecen una herramienta natural para ayudar con su solución. Esto ha dado lugar a la autoformulación, que se centra en el uso de LLMs para traducir descripciones en lengua natural de los problemas de decisión en modelos de optimización matemática ejecutables. Sin embargo, esta investigación a menudo descuida restricciones blandas como las propuestas por los consumidores, gerentes y planificadores, a pesar del hecho de que muchos problemas prácticos requieren que los modelos explícitamente. Introducimos ASTOR,

Machine translation (Meta M2M100); the English original follows.

Abstract

Real-world decision problems combine two qualitatively different kinds of constraints: requirements that must be satisfied, called hard constraints, and preferences that can be violated, called soft constraints. Because these problems are often posed to decision-makers in natural language, large language models offer a natural tool for assisting with their solution. This has given rise to autoformulation, which focuses on using LLMs to translate natural-language descriptions of decision problems into executable mathematical optimization models. However, this research often neglects soft constraints like those proposed by consumers, managers, and planners, despite the fact that many practical problems require modeling them explicitly. We introduce ASTOR, an inference-time framework that maps a natural-language decision problem to a grammar-guided abstract syntax tree (AST) before deterministically compiling the resulting representation into executable optimization code. We also introduce SoftOR, a unified benchmark of natural language problems which focuses on soft constraint formulations. On SoftOR, ASTOR achieves the best aggregate accuracy among comparable single-request and multi-agent LLM methods while remaining competitive on standard autoformulation benchmarks that include only hard constraints. Code and data are available at https://anonymous.4open.science/r/astor-softor-iclr-E34C/README.md.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Executable, Benchmark (surveying), Programming language, Syntax, Natural language

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange