Code generation on MooseBench standard non-registry, 220-case
81.6IFSPDE-Refine
Evaluation Results
| Method | Links | |
|---|---|---|
| PDE-RefineLLM=Claude Sonnet 4.62026.05 | 81.6 | |
| PDE-RefineLLM=GPT-5.42026.05 | 80.2 | |
| SpecGenLLM=Claude Sonnet 4.62026.05 | 80 | |
| PDE-RefineLLM=DeepSeek V4 Flash2026.05 | 78.2 | |
| SpecGenLLM=GPT-5.42026.05 | 76.9 | |
| DirectLLM=GPT-5.42026.05 | 75.6 | |
| DirectLLM=Claude Sonnet 4.62026.05 | 74.4 | |
| SpecGenLLM=DeepSeek V4 Flash2026.05 | 66.8 | |
| DirectLLM=DeepSeek V4 Flash2026.05 | 59.9 |