MEIDNet (paper): alignment of structure and property latents on Perov-5¶
Published (from the paper) · ●○○○○ generated
| Dataset | Perov-5 |
| Modalities | structure, property:heat_all, property:dir_gap |
| Model | MEIDNet early fusion + curriculum (production model) (MEIDNet 1.0 (paper), perovskite_abx3) |
| Category | Representation quality |
| Submitted by | Anand Babu (UCLouvain) on 2026-05-29 |
| Configuration | examples/perov5/meidnet.yaml |
| Checkpoint | dual_autoencoder_clip_earlyfusion_propertyaware_2k.pth |
| Notes on the model | 128-dimensional shared latent space; 2000 epochs, contrastive warm-up over the first 1200 |
Results¶
| metric | value | split | meaning | note |
|---|---|---|---|---|
cosine_matched |
0.97 | validation | mean cosine similarity between the structure and property latents of the same material (1 = aligned) | mean cosine similarity between the structure and property latents of the same material, as reported in the paper |
l2_matched |
0.24 | validation | mean L2 distance between those two latents (0 = identical) | mean L2 distance between the two latents of the same material, as reported in the paper |
Validation¶
Evidence level: ●○○○○ generated.
Evidence:
Reproduce¶
Configuration examples/perov5/meidnet.yaml, split val:
python scripts/benchmarks.py reproduce paper-representation
These numbers are quoted from the paper.
Re-run here did not agree¶
On 2026-10-03 this repository's code re-ran the recipe above (AMD64 CPU, MEIDNet 2.0.0) and obtained different numbers. The row keeps its status; the comparison is shown so that readers can judge it. Possible reasons: a different split or protocol in the original measurement, a different definition of the metric, or a checkpoint that is not the one the numbers were measured on.
| metric | this row | obtained here | agree |
|---|---|---|---|
cosine_matched |
0.97 | 0.771 | no |
l2_matched |
0.24 | 0.676 | no |
Also obtained in that run: mae_heat_all = 0.392, r2_heat_all = 0.497, mae_dir_gap = 2.79, r2_dir_gap = -35.8, retrieval_top1 = 0.0253, retrieval_top5 = 0.0908, n_evaluated = 3,787
Notes¶
Numbers quoted from the paper (npj Computational Materials, 2026). reproduce re-evaluates the shipped checkpoint on the CDVAE validation split and adds MAE, R2 and retrieval metrics; cosine_matched is compared with the quoted value.