Retrieval tasks, published encode vs BF16
| Task | BF16 | exlr8 K4 | Ξ | Top-10 overlap with BF16 |
|---|---|---|---|---|
| SciFact | 0.8514 | 0.8520 | +0.0005 | 88% |
| NFCorpus | 0.3691 | 0.3646 | β0.0045 | 88% |
| ArguAna | 0.5887 | 0.5881 | β0.0006 | 89% |
| XPQA Arabic | 0.4657 | 0.4723 | +0.0066 | 86% |
| XPQA Chinese | 0.6307 | 0.6363 | +0.0056 | 91% |
| XPQA Hindi | 0.7437 | 0.7460 | +0.0023 | 91% |
| XPQA German | 0.7526 | 0.7514 | β0.0013 | 88% |
| CSN Python | 0.9550 | 0.9539 | β0.0012 | 92% |
| CSN Java | 0.9163 | 0.9111 | β0.0052 | 93% |
| CSN Go | 0.9667 | 0.9679 | +0.0012 | 93% |
| STS-B (Spearman) | 0.8750 | 0.8744 | β0.0005 | sim corr 0.999 |
- Downloads last month
- 39
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support
Model tree for kindlingai/embeddinggemma-2-exlr8-k4
Base model
google/embeddinggemma-2