Research map / Large language models / Agents, code and math
SimulCost: A Cost-Aware Benchmark and Toolkit for Automating Physics Simulations with LLMs
Yadi Cao, Sicheng Lai, Jiahe Huang, Yang Zhang, Zach Lawrence, Rohan Bhakta, Izzy F. Thomas, Mingyun Cao, Chung-Hao Tsai, Zihao Zhou, Yidong Zhao, Hao Liu and 3 more
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 |
|---|---|---|---|
| use multi-round instead of single-round inference | LLMs · SimulCost (11 solvers) | success rate: 45-62% -> 66-81% | Improved |
| tighten accuracy requirement low to high | LLMs single-round · SimulCost | success rate (%): 64 -> 39 | Hurt |
| tighten accuracy requirement low to high | LLMs multi-round · SimulCost | success rate (%): 73 -> 71 | No clear effect |
| use multi-round instead of single-round at high accuracy | LLMs · SimulCost | success rate: +30.5 | Improved |
| add in-context examples (ICL) (ablation) | Claude-3.7-Sonnet single-round · SimulCost ablation subset | success rate: +28.7 | Improved |
| add in-context examples (ICL) (ablation) | Claude-3.7-Sonnet multi-round · SimulCost ablation subset | success rate | Hurt |
| include cost information in ICL examples (ablation) | Claude-3.7-Sonnet single-round · SimulCost ablation subset | efficiency | Improved |
| use LLM instead of Bayesian optimization (BO-GP) | multi-round tuning · SimulCost, low accuracy | efficiency: 1.05 -> 2.13 | Improved |
| increase GPT-5 reasoning effort (ablation) | GPT-5 · SimulCost | success rate | No clear effect |
| let LLM provide initial guess for search algorithm | search algorithms · SimulCost | success rate: +8.6 | Mixed |
| add multi-round inference for Misc parameters | LLMs · SimulCost task groups | success rate: +28 | Improved |
Closest papers by meaning
- Mosaic: A Benchmark Suite for Differentiable Physics Solvers (NeurIPS 2026 · similarity 0.85)
- PhysGym: Benchmarking LLMs in Interactive Physics Discovery with Controlled Priors (NeurIPS 2025 · similarity 0.85)
- ScaleSim: Serving Large-Scale Multi-Agent Simulation with Invocation Distance-Based Memory Management (ICML 2026 · similarity 0.84)
- MCPMark: A Benchmark for Stress-Testing Realistic and Comprehensive MCP Use (ICLR 2026 · similarity 0.83)
- The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements (NeurIPS 2025 · similarity 0.83)
- LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling (NeurIPS 2026 · similarity 0.83)
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.
