Long-context retrieval on SummScreen LongEmbed
98.1nDCG@10Single-vector
Evaluation results
| Method | Links | |
|---|---|---|
| Single-vectorMLDR Train=true, Backbone=Qwen3-Embedding-0.6B2026.06 | 98.1 | |
| BM25MLDR Train=false2026.06 | 97.6 | |
| MPE-Rand[32, 1024]MLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 94.9 | |
| BGE-M3MLDR Train=true2026.06 | 94 | |
| jina-v2-base-enMLDR Train=false2026.06 | 93.5 | |
| MPE Fixed-64MLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 92.8 | |
| OpenAI-Ada-0022026.06 | 91.8 | |
| MaxPMLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 90.6 | |
| MaxP-TrainMLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 90.3 |