Hybrid Retrieval Benchmark — dbpedia
500 sampled queries, seed 42, warmup 10, concurrency 8. Dense: all-MiniLM-L6-v2.
Sparse: Qdrant/bm25. Colbert: answerai-colbert-small-v1. Cross-encoder: local fastembed
rerank (Xenova/ms-marco-MiniLM-L-6-v2). Dataset: BeIR/dbpedia-entity-generated-queries,
100K passages, one generated query per passage.
recall@20 by quantization config × rescorer (hybrid prefetch, prefetch_limit=100)
| config | colbert | cross-encoder | rrf |
| no-quant (f32) | 96.4 | 96.0 | 95.0 |
| no-quant (f16) | 96.0 | 95.8 | 95.0 |
| no-quant (turbo4) | 96.2 | 96.0 | 95.2 |
| scalar int8 | 96.4 | 96.0 | 95.0 |
| product x16 | 96.4 | 96.0 | 95.2 |
| binary (f32) | 96.0 | 95.8 | 95.2 |
| binary (f16) | 95.6 | 96.0 | 95.0 |
| turbo4 tq1 | 96.4 | 95.8 | 95.0 |
recall / mrr / ndcg by rescorer
recall@20
mrr@20
ndcg@20
(opacity = metric; hue = rescorer)
throughput by rescorer
recall@20 vs prefetch_limit, by rescorer (hybrid, k=20)
colbert · recall@20
cross-encoder · recall@20
rrf · recall@20
throughput vs prefetch_limit, by rescorer (hybrid, k=20)
colbert · qps (log)
cross-encoder · qps (log)
rrf · qps (log)
colbert & cross-encoder, hybrid vs sparse-only prefetch (k=20, prefetch_limit=100)
recall@20
mrr@20
ndcg@20
(opacity = metric; hue = rescorer)
hybrid prefetch
colbertmrr 0.884ndcg 0.903
cross-encodermrr 0.886ndcg 0.904
sparse-only prefetch
colbertmrr 0.853ndcg 0.867
cross-encodermrr 0.856ndcg 0.869
recall@20 vs prefetch_limit, colbert & cross-encoder (sparse-only, k=20)
colbert · recall@20
cross-encoder · recall@20
throughput vs prefetch_limit, colbert & cross-encoder (sparse-only, k=20)
colbert · qps (log)
cross-encoder · qps (log)
mrr@20 by config, colbert rescorer (bold outline = turbo4 datatype)
hybridsparse-only
ndcg@20 by config, colbert rescorer (bold outline = turbo4 datatype)
hybridsparse-only
exact values, colbert rescorer, k=20 prefetch_limit=100
| config | hybrid mrr | hybrid ndcg | sparse-only mrr | sparse-only ndcg |
| no-quant (f32) | 0.8860 | 0.9046 | 0.8540 | 0.8673 |
| no-quant (f16) | 0.8857 | 0.9037 | 0.8540 | 0.8673 |
| no-quant (turbo4) | 0.8840 | 0.9027 | 0.8546 | 0.8676 |
| scalar int8 | 0.8839 | 0.9031 | 0.8520 | 0.8657 |
| product x16 | 0.8796 | 0.8998 | 0.8501 | 0.8643 |
| binary (f32) | 0.8847 | 0.9028 | 0.8532 | 0.8665 |
| binary (f16) | 0.8839 | 0.9013 | 0.8540 | 0.8669 |
| turbo4 tq1 | 0.8841 | 0.9033 | 0.8546 | 0.8677 |