Research map / Large language models / Agents, code and math
OPT-Engine: Benchmarking the Limits of LLMs in Optimization Modeling via Complexity Scaling
Yitian Chen, Cheng Cheng, Yinan Sun, Zi Ling, Dongdong Ge
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add RLVR post-training to Qwen3-4B-Instruct | Qwen3-4B-Instruct SIR · OPT-Engine Knapsack | execution rate: 23.1% -> 99.6% | Improved |
| add RLVR post-training to Qwen3-4B-Instruct | Qwen3-4B-Instruct SIR · OPT-Engine TSP | execution rate: 23.1% -> 38.1% | Improved |
| add RLVR post-training to Qwen3-4B-Instruct | Qwen3-4B-Instruct SIR · OPT-Engine Inventory | execution rate: 35.6% -> 38.1% | Improved |
| add Python computational tools to PTR | DeepSeek-V3.2 PTR · OPT-Engine TSP (Hard) | accuracy: 2.0% -> 44.0% | Improved |
| add Python computational tools to PTR | DeepSeek-V3.2 PTR · OPT-Engine Netflow (Hard) | accuracy: 0.0% -> 75.0% | Improved |
| add Python computational tools to PTR | DeepSeek-V3.2 PTR · OPT-Engine Knapsack (Easy) | accuracy: 89.2% -> 87.7% | Hurt |
| add external solver (SIR) instead of CompPTR | DeepSeek-V3.2 CompPTR · OPT-Engine TSP (Hard) | accuracy: 44.0% -> 80.0% | Improved |
| add external solver (SIR) instead of pure-text reasoning | DeepSeek-V3.2 PTR · OPT-Engine Netflow (Hard) | accuracy: 0.0% -> 100.0% | Improved |
| increase linguistic complexity (PPL) of problem text (ablation) | DeepSeek-V3.2 · OPT-Engine | accuracy | No clear effect |
| add constant K to the objective function (ablation) | DeepSeek-V3.2 · OPT-Engine | accuracy | No clear effect |
| add simple auxiliary constraints (ablation) | DeepSeek-V3.2 PTR and SIR · OPT-Engine | accuracy | Hurt |
Closest papers by meaning
- SolverLLM: Leveraging Test-Time Scaling for Optimization Problem via LLM-Guided Search (NeurIPS 2025 · similarity 0.91)
- OptiBench Meets ReSocratic: Measure and Improve LLMs for Optimization Modeling (ICLR 2025 · similarity 0.90)
- REASONING COMPILER: LLM-Guided Optimizations for Efficient Model Serving (NeurIPS 2025 · similarity 0.90)
- MM-OptBench: A Solver-Grounded Benchmark for Multimodal Optimization Modeling (NeurIPS 2026 · similarity 0.90)
- Autoformulation of Mathematical Optimization Models Using LLMs (ICML 2025 · similarity 0.90)
- Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces (NeurIPS 2026 · similarity 0.90)
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.
