Semantic Answer Caching on HotpotQA
62Answer Hit Rate (aHR)Naive
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| NaiveRegime=near_miss, Evidence validation=None2026.05 | 62 | 0 | 0 | |
| NaiveRegime=document_drift, Evidence validation=None2026.05 | 62 | 35 | 56 | |
| NaiveRegime=exact_repeat, Evidence validation=None2026.05 | 41 | 15 | 37 | |
| NaiveRegime=long_shared_doc, Evidence validation=None2026.05 | 41 | 15 | 37 | |
| NaiveRegime=bounded_kb_cag, Evidence validation=None2026.05 | 41 | 15 | 37 | |
| NaiveRegime=paraphrase, Evidence validation=None2026.05 | 40 | 22 | 56 | |
| GroundedCacheRegime=long_shared_doc, Evidence validation=full2026.05 | 6 | 0 | 0 | |
| GroundedCacheRegime=bounded_kb_cag, Evidence validation=full2026.05 | 6 | 0 | 0 | |
| GroundedCacheRegime=exact_repeat, Evidence validation=full2026.05 | 4 | 0 | 0 | |
| GroundedCacheRegime=paraphrase, Evidence validation=full2026.05 | 4 | 0 | 0 | |
| GroundedCacheRegime=near_miss, Evidence validation=full2026.05 | 4 | 0 | 0 | |
| GroundedCacheRegime=document_drift, Evidence validation=full2026.05 | 1 | 0 | 0 |