MEIDNet Benchmarks¶
Reproducible results for multimodal materials representation and inverse design, one leaderboard per dataset. Different datasets are different tasks, so there is no ranking across them: compare rows only within a page. The idea follows Matbench Discovery: fixed data, inspectable submissions, plots and an evidence level per row.
3datasets
5results
2reproduced here
1DFT or experiment
Datasets¶
| dataset | role | results | reproduced | highest evidence |
|---|---|---|---|---|
| Carbon-24 | published: structure-representation generalization (not a multimodal benchmark) | 0 | 0 | – |
| MP-20 | published: structure-representation generalization (not a multimodal benchmark) | 0 | 0 | – |
| Perov-5 | the published multimodal benchmark: alignment, property reconstruction and inverse design | 5 | 2 | DFT validated |
The evidence ladder¶
| level | means |
|---|---|
| ○○○○○ generated | the model produced it |
| ●○○○○ ML filtered | it passed the family's rules and the model's own checks |
| ●●○○○ MLIP validated | relaxed and found stable with a machine-learned potential (meidnet screen, MACE) |
| ●●●○○ DFT validated | confirmed by density-functional calculations |
| ●●●●○ experimentally validated | synthesised and measured |
Status of a row¶
Published = quoted from the paper · Community submitted = a pull request, validated by CI · MEIDNet verified = reproduced here from the submitted configuration (python scripts/benchmarks.py reproduce <id>). How to contribute · Databases by application
Latest results¶
| date | result | dataset | category | evidence | status |
|---|---|---|---|---|---|
| 2026-10-03 | MLIP screening of the shipped paper candidates (MACE-MP-0) | Perov-5 | Validation level | ●●●○○ MLIP validated | Community submitted |
| 2026-10-03 | Shipped checkpoint re-evaluated on the training split (this code) | Perov-5 | Representation quality | ●○○○○ generated | MEIDNet verified |
| 2026-10-03 | Shipped checkpoint re-evaluated on the held-out validation split (this code) | Perov-5 | Representation quality | ●○○○○ generated | MEIDNet verified |
| 2026-05-29 | MEIDNet (paper): inverse design of perovskites from property targets | Perov-5 | Conditional design | ●●●●○ DFT validated | Published (from the paper) |
| 2026-05-29 | MEIDNet (paper): alignment of structure and property latents on Perov-5 | Perov-5 | Representation quality | ●○○○○ generated | Published (from the paper) |