PEARL: Solver-in-the-Loop Interactive Optimization Modeling from Natural Language
Summary
Introduces PEARL, a system for interactive optimization modeling that uses a solver-in-the-loop approach to iteratively improve formulations from natural language, outperforming much larger models like DeepSeek-V3.2-685B.
View Cached Full Text
Cached at: 07/22/26, 08:20 AM
# PEARL: Solver-in-the-Loop Interactive Optimization Modeling from Natural Language
Source: [https://arxiv.org/abs/2607.18256](https://arxiv.org/abs/2607.18256)
[View PDF](https://arxiv.org/pdf/2607.18256)
> Abstract: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\.
## Submission history
From: Hongliang Lu \[[view email](https://arxiv.org/show-email/2067fae2/2607.18256)\] **\[v1\]**Thu, 14 May 2026 06:47:20 UTC \(1,486 KB\)Similar Articles
MILP-Evo: Closed-Loop Fully Automatic Design of MILP Solvers
The paper introduces MILP-Evo, a closed-loop framework that uses LLM-guided program evolution to automatically design white-box MILP solver components (cut selectors and branching rules) by iteratively generating and evaluating candidate programs via end-to-end solver performance on MILP instances.
SocraticPO: Policy Optimization via Interactive Guidance
SocraticPO augments RL rollouts with Socratic-style natural language guidance and reward decay to improve scientific reasoning in LLMs, outperforming strong baselines on SciKnowEval benchmarks.
SPEAR: Code-Augmented Agentic Prompt Optimization
SPEAR is a code-augmented agentic prompt optimizer that uses a Python sandbox for structural error analysis, achieving state-of-the-art performance on multiple LLM evaluation suites including industrial judge tasks, BBH, and GSM8K.
OPERA: Aligning Open-Ended Reasoning via Objective Perplexity-based Reinforcement Learning
OPERA proposes a reinforcement learning method for open-ended tasks using intrinsic rewards based on perplexity dynamics, replacing unreliable LLM-as-a-judge reward models. It achieves state-of-the-art results on Qwen3-8B, matching proprietary models in creative writing and other open-ended tasks.
Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts
Introduces ReElicit, a Bayesian optimization framework that uses LLMs to elicit and adapt feature spaces for optimizing system prompts under aggregate scalar feedback, achieving strong performance across ten benchmark tasks.