AI Paper Atlas

Research map / Large language models / Agents, code and math

Constructing Industrial-Scale Optimization Modeling Benchmark

Zhong Li, Hongliang Lu, Tao Wei, Yuxuan Chen, Wenyu Liu, Yuan Lan, Fan Zhang, Zaiwen Wen

ICML 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
evaluate on MIPLIB-NL instead of NL4OptSIRL-Qwen2.5-32B · NL4Opt vs MIPLIB-NLPass@1 accuracy (%): 87.90 -> 1.43Hurt
evaluate on MIPLIB-NL instead of NL4OptGPT-5.1 · NL4Opt vs MIPLIB-NLPass@1 accuracy (%): 81.78 -> 39.05Hurt
evaluate on MIPLIB-NL instead of NL4Optbare LLMs average · NL4Opt vs MIPLIB-NLPass@1 accuracy (%): 78.78 -> 17.85Hurt
evaluate on MIPLIB-NL instead of MAMO-EOptMATH-Qwen2.5-7B · MAMO-E vs MIPLIB-NLPass@1 accuracy (%): 97.06 -> 0.00Hurt
enable thinking modeClaude-Sonnet-4.5 · MIPLIB-NLPass@1 accuracy (%): 24.23 -> 30.00Improved
enable thinking modeClaude-Sonnet-4.5 · IndustryORPass@1 accuracy (%): 71.43 -> 76.19Improved
enable thinking modeDeepSeek-V3.2 · MIPLIB-NLPass@1 accuracy (%): 23.81 -> 22.38No clear effect
enable thinking modeQwen3-Max · MIPLIB-NLPass@1 accuracy (%): 21.43 -> 21.43No clear effect
enable thinking modeQwen3-Max · IndustryORPass@1 accuracy (%): 47.62 -> 59.52Improved
fine-tune on existing toy datasets (SIRL 32B vs 7B scaling)SIRL-Qwen2.5 · MIPLIB-NLPass@1 accuracy (%): 0.48 -> 1.43No clear effect
increase problem scale from small to very largebare LLMs · MIPLIB-NLaverage Pass@1Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy