Benchmark
Agentset Embedding nDCG@10 - SciFact
Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.
- Models measured
- 15
- Scores recorded
- 15
- Sources
- 1
- IQ weight
- 1.34
Score distribution
- 71.7%
- Mean
- 72.6%
- Median
- 6.1%
- Std dev
- 63.6%
- 10th percentile
- 77.9%
- 90th percentile
- 59.9%
- Min
This benchmark has a weight of 1.34 for embedding models, measured over 15 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 | Gemini Embedding 2SciFact· Agentset | 87.1% |
| 2 | voyage-3-largeSciFact· Agentset | 77.9% |
| 3 | voyage-4SciFact· Agentset | 75.8% |
| 4 | Qwen3 Embedding 8BSciFact· Agentset | 74.4% |
| 5 | Jina Embeddings v5 Text SmallSciFact· Agentset | 73.4% |
| 6 | Text Embedding 004SciFact· Agentset | 72.9% |
| 7 | Text Embedding 3 LargeSciFact· Agentset | 72.7% |
| 8 | voyage-3.5-liteSciFact· Agentset | 72.6% |
| 9 | voyage-3.5SciFact· Agentset | 70.5% |
| 10 | Embed Multilingual V3.0SciFact· Agentset | 70.2% |
| 11 | Qwen3 Embedding 4BSciFact· Agentset | 69.7% |
| 12 | Text Embedding 3 SmallSciFact· Agentset | 68.4% |
| 13 | Qwen3 Embedding 0.6BSciFact· Agentset | 66.6% |
| 14 | Jina Embeddings v3SciFact· Agentset | 63.6% |
| 15 | BGE M3SciFact· Agentset | 59.9% |