Long-context retrieval on QMSum LongEmbed
0.813nDCG@10BM25
Evaluation results
| Method | Links | |
|---|---|---|
| BM25MLDR Train=false2026.06 | 0.813 | |
| MPE-Rand[32, 1024]MLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 0.705 | |
| MPE Fixed-64MLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 0.652 | |
| MaxP-TrainMLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 0.57 | |
| Single-vectorMLDR Train=true, Backbone=Qwen3-Embedding-0.6B2026.06 | 0.565 | |
| MaxPMLDR Train=true, Backbone=Qwen3-Embedding-0.6B, Inference chunk size=642026.06 | 0.55 | |
| OpenAI-Ada-0022026.06 | 0.4 | |
| jina-v2-base-enMLDR Train=false2026.06 | 0.389 | |
| BGE-M3MLDR Train=true2026.06 | 0.355 |