asmhatre
74e6334704
Ingest time, recall@10 against exact search, latency, metadata filtering and running cost, with the defaults that cost recall reproduced and every figure written to output/.
8 lines
398 B
Plaintext
8 lines
398 B
Plaintext
Spring AI 2.0.1 sends num_candidates = (int)(1.5 * topK) = 15 for topK=10 (ElasticsearchVectorStore bytecode: ldc2_w 1.5d, dmul, d2i)
|
|
same index, same 100 queries, raw _search with the num_candidates varied:
|
|
num_candidates recall@10 p50 ms
|
|
15 0.472 5.0
|
|
50 0.477 3.9
|
|
100 0.479 3.6
|
|
500 0.484 3.8
|