PEARL: Solver-in-the-Loop Interactive Optimization Modeling from Natural Language
Quick Answer
PEARL is an interactive optimization modeling system that integrates Python execution and solver diagnostics, significantly improving verified solve rates.
Quick Take
The PEARL-Qwen3-4B model outperforms the larger DeepSeek-V3.2-685B in accuracy across various optimization benchmarks, demonstrating the effectiveness of iterative model refinement.
Key Points
- PEARL uses iterative solve-debug-revise cycles for optimization modeling.
- It learns when to test partial models and revise based on solver feedback.
- The system improves both formulations and solver code before finalization.
- PEARL-Qwen3-4B achieves higher accuracy than DeepSeek-V3.2-685B.
- The approach enhances verified solve rates across diverse optimization benchmarks.
DeepSignal Analysis
What happened
PEARL is an interactive optimization modeling system that enhances the process of translating natural language decision problems into executable solver code. It utilizes Python execution and solver diagnostics to improve the accuracy of optimization tasks, outperforming larger models like DeepSeek-V3.2-685B in benchmark tests.
Key evidence
- PEARL integrates Python execution and solver diagnostics, which allows for iterative refinement of optimization models.
- The PEARL-Qwen3-4B model demonstrates superior accuracy compared to the larger DeepSeek-V3.2-685B across various optimization benchmarks.
- PEARL operates through a multi-turn process, utilizing feedback from solver diagnostics to improve both model formulations and solver code.
Why it matters
The development of PEARL addresses the limitations of existing one-shot optimization modeling approaches by introducing an interactive framework. This could lead to more effective solutions in real-world applications where iterative refinement is crucial. The ability to learn from solver diagnostics may significantly enhance the reliability and accuracy of optimization models.
What to watch
Paper Resources
📖 Reader Mode
~2 min readAbstract:Optimization modeling is the process of translating real-world decision problems, often described in natural language, into formal mathematical formulations and executable solver code. While recent advances in large language models have shown promise in automating this process, most existing approaches remain one-shot: a model produces a formulation once, without executing it, conditioning on solver feedback, or iteratively revising errors. This stands in sharp contrast to real-world optimization modeling, which is inherently interactive and proceeds through repeated solve-debug-revise cycles. We introduce PEARL, a system for interactive optimization modeling that uses Python execution and mathematical programming solvers inside this loop. Rather than relying on a fixed repair workflow, PEARL learns when to test partial models, how to revise from solver diagnostics, and when to stop. It operates in a multi-turn tool-integrated setting where intermediate execution results, feasibility signals, and solution checks are used to improve both formulations and solver code before finalization. Across diverse optimization benchmarks, PEARL substantially improves verified solve rates over strong one-shot and tool-augmented baselines; notably, our PEARL-Qwen3-\textbf{4B} model outperforms the much larger DeepSeek-V3.2-\textbf{685B} in both macro- and micro-averaged accuracy on optimization modeling tasks.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.18256 [cs.AI] |
| (or arXiv:2607.18256v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.18256 arXiv-issued DOI via DataCite |
Submission history
From: Hongliang Lu [view email]
[v1]
Thu, 14 May 2026 06:47:20 UTC (1,486 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.