About RobustOPT
From uncertainty to a model you can solve.
RobustOPT helps you formulate, reformulate, implement and review optimization models under uncertainty.
Start with a problem description, or bring an existing mathematical model directly to Step 6. Review its reformulation, generate implementation code, solve a numerical instance, and inspect the results and report.
Method 01
Memory-augmented LLM
Relevant reformulation experience is retrieved from memory and supplied to a large language model to help derive a tractable counterpart. You review the resulting mathematics before continuing to code and numerical solving.
This implementation adapts the AutoREM approach to RobustOPT’s workflow. It does not reproduce the paper’s experimental pipeline verbatim.
AutoREM paperMethod 02
RSOME
RSOME reformulates supported robust and distributionally robust models into a finite, deterministic counterpart. Begin with RSOME model code you provide, or review code translated from your mathematical model.
Inspect the reformulated model and its conversion steps before numerical solving. A separate solver computes the solution; reformulation and solving are distinct steps.
RSOME in Python paperMethod 03
LLM only
Your currently selected language model works directly from the reviewed mathematical model, without retrieved Memory experience in its task prompt. Review the counterpart and implementation before numerical solving.
This is a direct LLM workflow. The papers below describe the other two methods.
Research data use
Research data use notice
Notice version: research-use-v1-20260919
New accounts must read and agree to this notice. Existing accounts and earlier activity are not automatically covered; research use must follow the consent scope and time recorded for each account.
With your agreement, we may use de-identified records of website use, problems and data you submit, generated models, code and solving results, and feedback or questionnaire answers you voluntarily submit for academic research, system evaluation and improvement.
Passwords and API keys are excluded from research use. Research publications will not disclose your identity or raw submitted content without separate authorization.
Registration agreement does not enroll you in the optional study or activate its additional task activity or content collection. Those separate permissions remain independent, and questionnaires remain voluntary.
To ask questions or request withdrawal from research use, contact the team through Feedback. The team needs to verify and process your request; submitting feedback does not automatically change permissions or delete data. Aggregate findings that have already been published may not be retractable.
Further reading