opensearch-project/opensearch-neural-sparse-encoding-v2-distill
The model should be selected considering search relevance, model inference and retrieval efficiency(FLOPS). We benchmark models' zero-shot performance on a subset of BEIR benchmark: TrecCovid,NFCorpus,NQ,HotpotQA,FiQA,ArguAna,Touche,DBPedia,SCIDOCS,FEVER,Climate FEVER,SciFact,Quora.
Benchmarks
CQADupstackPhysicsRetrieval
Duplicate question retrieval from StackExchange Physics
Corpus: 38,314 Queries: 1,039
Performance L4 b1 c16
Corpus TPS 31.1K
Corpus p50 57.0ms
Query TPS 3.2K
Query p50 52.8ms
CosQA
Code search with natural language queries
Corpus: 6,267 Queries: 500
Performance L4 b1 c16
Corpus TPS 13.5K
Corpus p50 51.4ms
Query TPS 2.1K
Query p50 45.3ms
FiQA2018
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
Performance L4 b1 c16
Corpus TPS 37.0K
Corpus p50 59.3ms
Query TPS 3.8K
Query p50 47.5ms
LegalBenchConsumerContractsQA
Question answering on consumer contracts
Corpus: 153 Queries: 396
Performance L4 b1 c16
Corpus TPS 91.6K
Corpus p50 73.7ms
Query TPS 5.2K
Query p50 49.9ms
NFCorpus
Biomedical literature search from NutritionFacts.org
Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3373
map at 10 0.1278
mrr at 10 0.5548
Performance L4 b1 c16
Corpus TPS 61.6K
Corpus p50 68.8ms
Query TPS 1.6K
Query p50 49.3ms
SCIDOCS
Citation prediction, document classification, and recommendation for scientific papers
Corpus: 25,656 Queries: 1,000
Performance L4 b1 c16
Corpus TPS 42.4K
Corpus p50 60.4ms
Query TPS 3.5K
Query p50 49.6ms
SciFact
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
Performance L4 b1 c16
Corpus TPS 54.4K
Corpus p50 68.2ms
Query TPS 5.1K
Query p50 50.1ms
StackOverflowQA
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
Performance L4 b1 c16
Corpus TPS 46.0K
Corpus p50 66.2ms
Query TPS 69.8K
Query p50 66.0ms