Materials Property Prediction on GNoME
1.22Mean Absolute Error (MAE)DeepSeek-V4-Pro
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DeepSeek-V4-Pro2026.07 | 1.22 | — | — | |
| Kimi-K2.62026.07 | 1.56 | — | — | |
| Opus-4.72026.07 | 1.94 | — | — | |
| GPT-5.52026.07 | 5.39 | — | — | |
| LLM-Prop2026.07 | 15.6 | — | — | |
| SciReasoner2026.07 | 21.91 | — | — | |
| SciReasoner2026.07 | 21.91 | — | — | |
| ALM CoreConfiguration=canonical understanding r=128+IT2026.06 | — | 0.034 | 0.051 | |
| CGCNNInput Format=CIF2026.06 | — | 0.014 | 0.045 | |
| LLM-PropInput Format=CIF2026.06 | — | 0.017 | 0.098 | |
| MatBERTInput Format=CIF2026.06 | — | 0.02 | 0.042 |