Add MTEB metrics

README.md

@@ -182,7 +182,7 @@ Then you can load this model and run inference.
 from sentence_transformers import SparseEncoder
 
 # Download from the 🤗 Hub
-model = SparseEncoder("tomaarsen/splade-robbert-dutch-base")
+model = SparseEncoder("sparse-encoder/splade-robbert-dutch-base-v1")
 # Run inference
 queries = [
     "hoe maak je een keldervloer glad",
@@ -231,6 +231,52 @@ You can finetune this model on your own dataset.
 
 ### Metrics
 
+#### MTEB
+
+To evaluate this model, we've evaluated it on [BelebeleRetrieval](https://arxiv.org/abs/2308.16884) and WikipediaRetrievalMultilingual: the two Dutch Retrieval tasks recommended by [MMTEB](https://huggingface.co/spaces/mteb/leaderboard).
+
+![image/png](https://cdn-uploads.huggingface.co/production/uploads/6317233cc92fd6fee317e030/mnrPBcvDBGImGGVi4EBdW.png)
+
+As shown in this figure, `splade-robbert-dutch-base-v1` heavily outperforms the only other Dutch-capable Sparse embedding model, and outperforms all equally sized dense embedding models, despite only using an average of ~250 active (non-zero) dimensions for documents (during training). 
+
+<details><summary>Click to see the full table</summary>
+
+| Model                                             | Number of Parameters | BelebeleRetrieval | WikipediaRetrievalMultilingual |
+|---------------------------------------------------|----------------------|-------------------|--------------------------------|
+| multilingual-e5-large-instruct                    | 560M                 | 94.725            | 92.342                         |
+| multilingual-e5-large                             | 560M                 | 94.607            | -                              |
+| Solon-embeddings-large-0.1                        | 559M                 | 93.651            | 91.239                         |
+| snowflake-arctic-embed-l-v2.0                     | 568M                 | 93.318            | 90.902                         |
+| bge-m3                                            | 568M                 | 93.859            | 90.106                         |
+| multilingual-e5-base                              | 278M                 | 93.731            | 89.905                         |
+| jina-embeddings-v3                                | 572M                 | 93.105            | 90.296                         |
+| **splade-robbert-dutch-base-v1**                  | 124M                 | 93.389            | 88.937                         |
+| multilingual-e5-small                             | 118M                 | 92.859            | 88.662                         |
+| KaLM-embedding-multilingual-mini-v1               | 494M                 | 91.453            | 88.413                         |
+| Qwen3-Embedding-0.6B                              | 595M                 | 91.686            | 88.121                         |
+| snowflake-arctic-embed-m-v2.0                     | 305M                 | 88.358            | 88.898                         |
+| granite-embedding-278m-multilingual               | 278M                 | 87.039            | 86.324                         |
+| gte-multilingual-base                             | 305M                 | 89.204            | 83.976                         |
+| KaLM-embedding-multilingual-mini-instruct-v1      | 494M                 | 85.648            | 85.877                         |
+| granite-embedding-107m-multilingual               | 107M                 | 85.068            | 85.097                         |
+| robbert-2022-dutch-sentence-transformers          | 124M                 | 86.146            | 82.553                         |
+| opensearch-neural-sparse-encoding-multilingual-v1 | 167M                 | 80.101            | 85.529                         |
+| paraphrase-multilingual-mpnet-base-v2             | 278M                 | 83.910            | 76.676                         |
+| e5-large-v2                                       | 335M                 | 76.433            | 79.711                         |
+| STS-multilingual-mpnet-base-v2                    | 278M                 | 80.625            | 73.803                         |
+| paraphrase-multilingual-MiniLM-L12-v2             | 118M                 | 81.021            | 71.091                         |
+| snowflake-arctic-embed-m                          | 109M                 | 65.511            | 74.801                         |
+| potion-multilingual-128M                          | 128M                 | 72.454            | 65.559                         |
+| static-similarity-mrl-multilingual-v1             | 108M                 | 67.375            | 69.050                         |
+| snowflake-arctic-embed-m-long                     | 137M                 | 67.947            | 65.988                         |
+| snowflake-arctic-embed-m-v1.5                     | 109M                 | 65.511            | 67.920                         |
+| bge-base-en-v1.5                                  | 109M                 | 61.073            | 72.093                         |
+| snowflake-arctic-embed-s                          | 32M                  | 58.683            | 70.887                         |
+| potion-base-8M                                    | 7M                   | 22.563            | 40.107                         |
+
+</details>
+
+
 #### Sparse Information Retrieval
 
 * Dataset: `msmarco-eval-1k`