Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -32,7 +32,38 @@ Note, that the embeddings are not normalized so you will need to normalize them
|
|
| 32 |
Retrieval performance for the TREC DL21-23, MSMARCOV2-Dev and Raggy Queries can be found below with BM25 as a baseline. For both systems, retrieval is at the segment level and Doc Score = Max (passage score).
|
| 33 |
Retrieval is done via a dot product and happens in BF16.
|
| 34 |
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
## Loading the dataset
|
| 37 |
|
| 38 |
### Loading the document embeddings
|
|
@@ -67,7 +98,7 @@ import numpy as np
|
|
| 67 |
|
| 68 |
|
| 69 |
top_k = 100
|
| 70 |
-
docs_stream = load_dataset("spacemanidol/msmarco-v2.1-gte-large-en-v1.5,split="train", streaming=True)
|
| 71 |
|
| 72 |
docs = []
|
| 73 |
doc_embeddings = []
|
|
|
|
| 32 |
Retrieval performance for the TREC DL21-23, MSMARCOV2-Dev and Raggy Queries can be found below with BM25 as a baseline. For both systems, retrieval is at the segment level and Doc Score = Max (passage score).
|
| 33 |
Retrieval is done via a dot product and happens in BF16.
|
| 34 |
|
| 35 |
+
NDCG @ 10
|
| 36 |
+
| Dataset | BM25 | GTE-Large-v1.5 |
|
| 37 |
+
|--------------------|--------|----------------|
|
| 38 |
+
| Deep Learning 2021 | 0.5778 | 0.7193 |
|
| 39 |
+
| Deep Learning 2022 | 0.3576 | 0.5358 |
|
| 40 |
+
| Deep Learning 2023 | 0.3356 | 0.4642 |
|
| 41 |
+
| msmarcov2-dev | N/A | 0.3538 |
|
| 42 |
+
| msmarcov2-dev2 | N/A | 0.3470 |
|
| 43 |
+
| Raggy Queries | 0.4227 | 0.5678 |
|
| 44 |
+
| TREC RAG (eval) | N/A | 0.5676 |
|
| 45 |
+
|
| 46 |
+
Recall @ 100
|
| 47 |
+
| Dataset | BM25 | GTE-Large-v1.5 |
|
| 48 |
+
|--------------------|--------|----------------|
|
| 49 |
+
| Deep Learning 2021 | 0.3811 | 0.4156 |
|
| 50 |
+
| Deep Learning 2022 | 0.233 | 0.31173 |
|
| 51 |
+
| Deep Learning 2023 | 0.3049 | 0.35236 |
|
| 52 |
+
| msmarcov2-dev | 0.6683 | 0.85135 |
|
| 53 |
+
| msmarcov2-dev2 | 0.6771 | 0.84333 |
|
| 54 |
+
| Raggy Queries | 0.2807 | 0.35125 |
|
| 55 |
+
| TREC RAG (eval) | N/A | 0.25223 |
|
| 56 |
+
|
| 57 |
+
Recall @ 1000
|
| 58 |
+
| Dataset | BM25 | GTE-Large-v1.5 |
|
| 59 |
+
|--------------------|--------|----------------|
|
| 60 |
+
| Deep Learning 2021 | 0.7115 | 0.73185 |
|
| 61 |
+
| Deep Learning 2022 | 0.479 | 0.55174 |
|
| 62 |
+
| Deep Learning 2023 | 0.5852 | 0.6167 |
|
| 63 |
+
| msmarcov2-dev | 0.8528 | 0.93549 |
|
| 64 |
+
| msmarcov2-dev2 | 0.8577 | 0.93997 |
|
| 65 |
+
| Raggy Queries | 0.5745 | 0.63515 |
|
| 66 |
+
| TREC RAG (eval) | N/A | 0.63133 |
|
| 67 |
## Loading the dataset
|
| 68 |
|
| 69 |
### Loading the document embeddings
|
|
|
|
| 98 |
|
| 99 |
|
| 100 |
top_k = 100
|
| 101 |
+
docs_stream = load_dataset("spacemanidol/msmarco-v2.1-gte-large-en-v1.5",split="train", streaming=True)
|
| 102 |
|
| 103 |
docs = []
|
| 104 |
doc_embeddings = []
|