[arXiv]score: 0.54
AutoSG Translates Natural Language Prompts into Custom Solvers for Expensive Optimization
May 26, 2026
AutoSG uses retrieval-augmented generation grounded in verified literature, a structure-preserving one-step self-refinement operator, and instance-adaptive evaluation to generate specialized solvers for expensive optimization tasks from plain text prompts.
cs.CLcs.AI
HOW THIS AFFECTS YOU
●
builderFully automated prompt-to-solver pipeline could reduce engineering effort for domain-specific optimization problems in logistics, science, or engineering.
●
researcherRAG-grounded code generation with structural preservation during refinement addresses hallucination and local-optima destruction in LLM-driven solver synthesis.