PKU-JX-LAB/Molecular-Dynamics-Benchmark

MD-EvalBench: Molecular Dynamics LLM Benchmark

The first comprehensive benchmark for evaluating Large Language Models in the Molecular Dynamics domain, from the paper "MDAgent2: Large Language Model for Code Generation and Knowledge Q&A in Molecular Dynamics".

Benchmark Components

• —MD-KnowledgeEval (336 questions): Theoretical knowledge in molecular dynamics
• —LAMMPS-SyntaxEval (368 questions): LAMMPS command and syntax understanding
• —LAMMPS-CodeGenEval (566 tasks): Automatic LAMMPS code generation

Leaderboard Tabs

1. 1.QA Benchmark: Knowledge and syntax QA performance comparison
2. 2.Code Generation: LAMMPS code generation performance across methods (Direct Prompting, MDAgent, MDAgent2-RUNTIME)

Configuration

• —Main leaderboard data is in app.py
• —Task definitions in src/about.py
• —Column definitions in src/display/utils.py