Research map / Large language models / Agents, code and math
Improving LLM-based Global Optimization with Search Space Partitioning
Andrej Schwanke, Lyubomir Ivanov, David Salinas, Fabio Ferreira, Aaron Klein, Frank Hutter, Arber Zela
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 |
|---|---|---|---|
| partition search space with KD-tree and sample LLM within regions | global LLM optimizer · synthetic functions (Hartmann, Rosenbrock, Rastrigin, Levy) | best function value | Improved |
| partition search space with KD-tree and sample LLM within regions | global LLM optimizer · Rastrigin 10D (Gemini-1.5-Flash) | best function value: -32.08 -> -0.17 | Improved |
| partition search space with KD-tree and sample LLM within regions | global LLM optimizer · Ackley 20D (Gemini-1.5-Flash) | best function value: -6.10 -> -0.00 | Improved |
| partition search space with KD-tree and sample LLM within regions | global LLM optimizer · Rosenbrock 8D (Gemini-1.5-Flash) | best function value: -4.66 -> -7.03 | Hurt |
| partition search space with KD-tree | global LLM optimizer · FCNet hyperparameter optimization | validation MSE | No clear effect |
| use LLM to sample inside regions instead of random sampling (ablation) | RS + KD-Tree · FCNet hyperparameter optimization | validation MSE | Improved |
| replace UCB1 with UCB-V exploration term (ablation) | HOLLM · Vehicle Safety (Gemini-2.0-Flash) | best function value | Improved |
| combine exploitation and exploration terms in score (ablation) | HOLLM · Vehicle Safety (Gemini-2.0-Flash) | best function value | Improved |
| score-based region selection instead of uniform selection (ablation) | HOLLM · Vehicle Safety (Gemini-2.0-Flash) | best function value | Improved |
| use partitioning with open LLMs (Llama-3.1-8B, Llama-3.3-70B, Qwen3-30B-A3B) | global LLM optimizer · Vehicle Safety (Gemini-2.0-Flash) | best function value | Improved |
| use HOLLM under 1% observation noise | global LLM optimizer · Vehicle Safety (Gemini-2.0-Flash) noisy | best function value | Improved |
| coarser leaf size m0=d instead of d/4 (finer) (ablation) | HOLLM · Levy 10D | best function value | Hurt |
Closest papers by meaning
- Large Language Models to Enhance Bayesian Optimization (ICLR 2024 · similarity 0.90)
- EVOLvE: Evaluating and Optimizing LLMs For In-Context Exploration (ICML 2025 · similarity 0.89)
- LensLLM: Unveiling Fine-Tuning Dynamics for LLM Selection (ICML 2025 · similarity 0.89)
- LLMOPT: Learning to Define and Solve General Optimization Problems from Scratch (ICLR 2025 · similarity 0.89)
- Benchmarking Optimizers for Large Language Model Pretraining (NeurIPS 2026 · similarity 0.88)
- SolverLLM: Leveraging Test-Time Scaling for Optimization Problem via LLM-Guided Search (NeurIPS 2025 · similarity 0.88)
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.
