Skip to content

Benchmark

MLEB - Singaporean Judicial Keywords

Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.

Models measured
18
Scores recorded
18
Sources
1
IQ weight
0.81

Score distribution

81.2%
Mean
84.4%
Median
12.3%
Std dev
63.4%
10th percentile
94.8%
90th percentile
57.5%
Min
57.5%96.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.81 for embedding models, measured over 18 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1voyage-4-largeSingaporean Judicial Keywords· MLEB data96.8%
2voyage-4Singaporean Judicial Keywords· MLEB data96.2%
3voyage-4-liteSingaporean Judicial Keywords· MLEB data94.8%
4Jina Embeddings v4Singaporean Judicial Keywords· MLEB data92.7%
5Qwen3 Embedding 8BSingaporean Judicial Keywords· MLEB data92.5%
6Qwen3 Embedding 4BSingaporean Judicial Keywords· MLEB data90.4%
7Jina Embeddings v5 Text SmallSingaporean Judicial Keywords· MLEB data88.5%
8Jina Embeddings v5 Text NanoSingaporean Judicial Keywords· MLEB data85.6%
9Qwen3 Embedding 0.6BSingaporean Judicial Keywords· MLEB data84.4%
10Text Embedding 3 LargeSingaporean Judicial Keywords· MLEB data83.5%
11Gemini Embedding 001Singaporean Judicial Keywords· MLEB data78.3%
12Text Embedding 3 SmallSingaporean Judicial Keywords· MLEB data77.7%
13snowflake-arctic-embed-l-v2.0Singaporean Judicial Keywords· MLEB data77.2%
14snowflake-arctic-embed-m-v2.0Singaporean Judicial Keywords· MLEB data73.8%
15Text Embedding Ada 002Singaporean Judicial Keywords· MLEB data69.6%
16Granite Embedding Small English R2Singaporean Judicial Keywords· MLEB data63.4%
17BGE M3Singaporean Judicial Keywords· MLEB data58.3%
18Multilingual E5 Large InstructSingaporean Judicial Keywords· MLEB data57.5%

Data credits: MLEB (Isaacus), MIT (licence)