Datasets:
Tasks:
Text Retrieval
Modalities:
Text
Formats:
json
Sub-tasks:
document-retrieval
Languages:
English
Size:
1K - 10K
License:
File size: 445 Bytes
0e11747 830cdf4 0e11747 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 | name: ParseEmbed
description: >
ParseEmbed is a hard retrieval benchmark for embedding models. It measures
semantic scope, formatting-sensitive meaning, and table cell grounding under
near-duplicate hard negatives.
evaluation_framework: parseembed
tasks:
- id: mean
config: parse-embed
split: mean
- id: text_formatting
config: parse-embed
split: text_formatting
- id: table
config: parse-embed
split: table
|