Retrieval-Augmented Question Answering on PopQA sampled queries from Augmented dataset
0.8725Average Latency (s)HaS
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HaSCategory=LLM for Validation2026.04 | 0.8725 | 45.78 | 29.06 | 27.2 | 26.38 | |
| SafeRadiusCategory=Reuse-based Methods2026.04 | 0.9773 | 42.45 | 26.49 | 24.77 | 23.38 | |
| MinCacheCategory=Reuse-based Methods2026.04 | 1.0437 | 42.54 | 26.76 | 24.52 | 23.6 | |
| ProximityCategory=Reuse-based Methods2026.04 | 1.1328 | 44.15 | 28.02 | 26.14 | 25.22 | |
| Full-Database RetrievalRetrieval Type=Full-Database2026.04 | 1.3847 | 46.52 | 29.7 | 27.8 | 27.03 | |
| CRAGCategory=LLM for Validation, Configuration=Replaces homologous validation with LLM-based evaluator2026.04 | 1.8186 | 45.57 | 28.85 | 27.06 | 25.42 |