Datasets:
Tasks:
Text Retrieval
Modalities:
Text
Formats:
parquet
Sub-tasks:
document-retrieval
Languages:
English
Size:
100K - 1M
License:
card: byte-preserved text wording
Browse files
README.md
CHANGED
|
@@ -66,6 +66,7 @@ before publishing; `provenance.json` records the source file hashes, what change
|
|
| 66 |
|
| 67 |
## What changed from the source
|
| 68 |
|
|
|
|
| 69 |
- `dev`: dropped 2,913 queries that have no qrels in this split (mteb does the same at load time)
|
| 70 |
- `test`: dropped 2,914 queries that have no qrels in this split (mteb does the same at load time)
|
| 71 |
- `train`: dropped 647 queries that have no qrels in this split (mteb does the same at load time)
|
|
|
|
| 66 |
|
| 67 |
## What changed from the source
|
| 68 |
|
| 69 |
+
- **byte-preserved** all text — no whitespace, newline, or control-character normalisation
|
| 70 |
- `dev`: dropped 2,913 queries that have no qrels in this split (mteb does the same at load time)
|
| 71 |
- `test`: dropped 2,914 queries that have no qrels in this split (mteb does the same at load time)
|
| 72 |
- `train`: dropped 647 queries that have no qrels in this split (mteb does the same at load time)
|