Room Embedding Evaluation on BIM-grounded Room Dataset
64NMISBM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SBMArchitecture=Transformer, Input format=BIM-Token Bundles2025.12 | 64 | 30.7 | 0.152 | 0.145 | |
| E5-Large-v2Number of parameters=560M, Pre-training=Common Crawl and Wikipedia, Input format=serialized room text2025.12 | 37.1 | -1.9 | 0.051 | 0.315 | |
| BGE-Large-en-v1.5Number of parameters=335M, Pre-training=web-scale text data, Input format=serialized room text2025.12 | 35.8 | -0.3 | 0.069 | 0.235 | |
| GTE-Large-en-v1.5Number of parameters=335M, Pre-training=web-scale text corpora, Input format=serialized room text2025.12 | 33.5 | -0.2 | 0.067 | 0.133 |