Benchmark
VN-MTEB (vie, v1) - Reranking
Rerank benchmarks measure how well models put the most relevant search results first. Scores are percentages.
- Models measured
- 8
- Scores recorded
- 8
- Sources
- 1
- IQ weight
- 1.25
Score distribution
- 61%
- Mean
- 64.7%
- Median
- 6.3%
- Std dev
- 52.3%
- 10th percentile
- 66.4%
- 90th percentile
- 51.1%
- Min
This benchmark has a weight of 1.25 for embedding models, measured over 8 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Jina Embeddings v5 Text SmallReranking· MTEB Extra Text Boards data | 67.5% |
| 2 | Multilingual E5 Large InstructReranking· MTEB Extra Text Boards data | 66.4% |
| 3 | E5 Mistral 7B InstructReranking· MTEB Extra Text Boards data | 66.1% |
| 4 | BGE M3Reranking· MTEB Extra Text Boards data | 64.7% |
| 5 | Multilingual E5 LargeReranking· MTEB Extra Text Boards data | 64.1% |
| 6 | BGE multilingual gemma2Reranking· MTEB Extra Text Boards data | 56.1% |
| 7 | All MiniLM L12 V2Reranking· MTEB Extra Text Boards data | 52.3% |
| 8 | All MiniLM L6 V2Reranking· MTEB Extra Text Boards data | 51.1% |