WaveflowDB vs. Pinecone
Biomedical Retrieval Benchmark — RAG-MINI-BIOASQ dataset (40,200 passages • 4,720 queries • 441 MB)
📊 Performance Comparison Snapshot
Top-k = 10 (High-Precision Retrieval)
| Metric | WaveflowDB | Pinecone | Improvement |
|---|---|---|---|
| Precision | 0.2901 | 0.2729 | +6.3% |
| Recall | 0.5658 | 0.5342 | +5.9% |
| F1 Score | 0.3309 | 0.3109 | +6.4% |
| MRR | 0.4423 | 0.4169 | +6.1% |
| NDCG@10 | 0.4320 | 0.4065 | +6.3% |
What this means:
WaveflowDB retrieves more relevant biomedical documents and ranks them higher, which is critical for clinical support and diagnostic contexts.
🧬 Why WaveflowDB Outperforms Pinecone
Based on the evaluation, WaveflowDB delivers:
- →Higher precision — fewer irrelevant medical documents
- →Higher recall — retrieves more clinically significant references
- →Higher ranking quality — critical information appears earlier
- →Stronger balance (F1) — consistently better retrieval across tasks
WaveflowDB's architecture handles biomedical terminology, scientific relationships, and domain-specific context with greater accuracy.