update metrics and citation
Browse files
README.md
CHANGED
@@ -56,6 +56,36 @@ SparseEncoder(
|
|
56 |
)
|
57 |
```
|
58 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
59 |
## Usage
|
60 |
|
61 |
### Direct Usage (Sentence Transformers)
|
@@ -314,6 +344,40 @@ You can finetune this model on your own dataset.
|
|
314 |
}
|
315 |
```
|
316 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
317 |
#### SpladeLoss
|
318 |
```bibtex
|
319 |
@misc{formal2022distillationhardnegativesampling,
|
|
|
56 |
)
|
57 |
```
|
58 |
|
59 |
+
### Metrics
|
60 |
+
```
|
61 |
+
{
|
62 |
+
"NDCG": {
|
63 |
+
"NDCG@2": 0.82678,
|
64 |
+
"NDCG@10": 0.86919,
|
65 |
+
"NDCG@20": 0.87429,
|
66 |
+
"NDCG@100": 0.87886
|
67 |
+
},
|
68 |
+
"MAP": {
|
69 |
+
"MAP@2": 0.8145,
|
70 |
+
"MAP@10": 0.84849,
|
71 |
+
"MAP@20": 0.85042,
|
72 |
+
"MAP@100": 0.85119
|
73 |
+
},
|
74 |
+
"Recall": {
|
75 |
+
"Recall@2": 0.8235,
|
76 |
+
"Recall@10": 0.905,
|
77 |
+
"Recall@20": 0.9215,
|
78 |
+
"Recall@100": 0.942
|
79 |
+
},
|
80 |
+
"Precision": {
|
81 |
+
"P@2": 0.8235,
|
82 |
+
"P@10": 0.181,
|
83 |
+
"P@20": 0.09215,
|
84 |
+
"P@100": 0.01884
|
85 |
+
}
|
86 |
+
}
|
87 |
+
```
|
88 |
+
|
89 |
## Usage
|
90 |
|
91 |
### Direct Usage (Sentence Transformers)
|
|
|
344 |
}
|
345 |
```
|
346 |
|
347 |
+
#### LIMIT
|
348 |
+
@misc{weller2025theoreticallimit,
|
349 |
+
title={On the Theoretical Limitations of Embedding-Based Retrieval},
|
350 |
+
author={Orion Weller and Michael Boratko and Iftekhar Naim and Jinhyuk Lee},
|
351 |
+
year={2025},
|
352 |
+
eprint={2508.21038},
|
353 |
+
archivePrefix={arXiv},
|
354 |
+
primaryClass={cs.IR},
|
355 |
+
url={https://arxiv.org/abs/2508.21038},
|
356 |
+
}
|
357 |
+
|
358 |
+
#### OpenSearch Models
|
359 |
+
@inproceedings{Shen_2025, series={SIGIR ’25},
|
360 |
+
title={Exploring $\ell_0$ parsification for Inference-free Sparse Retrievers},
|
361 |
+
url={http://dx.doi.org/10.1145/3726302.3730192},
|
362 |
+
DOI={10.1145/3726302.3730192},
|
363 |
+
booktitle={Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval},
|
364 |
+
publisher={ACM},
|
365 |
+
author={Shen, Xinjie and Geng, Zhichao and Yang, Yang},
|
366 |
+
year={2025},
|
367 |
+
month=jul, pages={2572–2576},
|
368 |
+
collection={SIGIR ’25}
|
369 |
+
}
|
370 |
+
|
371 |
+
@misc{geng2025competitivesearchrelevanceinferencefree,
|
372 |
+
title={Towards Competitive Search Relevance For Inference-Free Learned Sparse Retrievers},
|
373 |
+
author={Zhichao Geng and Yiwen Wang and Dongyu Ru and Yang Yang},
|
374 |
+
year={2025},
|
375 |
+
eprint={2411.04403},
|
376 |
+
archivePrefix={arXiv},
|
377 |
+
primaryClass={cs.IR},
|
378 |
+
url={https://arxiv.org/abs/2411.04403},
|
379 |
+
}
|
380 |
+
|
381 |
#### SpladeLoss
|
382 |
```bibtex
|
383 |
@misc{formal2022distillationhardnegativesampling,
|