opensearch-project/opensearch-neural-sparse-encoding-doc-v3-distill
The model should be selected considering search relevance, model inference and retrieval efficiency(FLOPS). We benchmark models' 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 33.5K tok/s
Corpus p50 58.4ms
Query 4.1K tok/s
Query p50 41.1ms
CosQA
Code search with natural language queries
Corpus: 6,267 Queries: 500
Performance L4 b1 c16
Corpus 15.6K tok/s
Corpus p50 51.5ms
Query 1.9K tok/s
Query p50 47.9ms
FiQA2018
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
Performance L4 b1 c16
Corpus 40.5K tok/s
Corpus p50 59.2ms
Query 4.6K tok/s
Query p50 39.5ms
LegalBenchConsumerContractsQA
Question answering on consumer contracts
Corpus: 153 Queries: 396
Performance L4 b1 c16
Corpus 97.4K tok/s
Corpus p50 76.3ms
Query 6.3K tok/s
Query p50 40.7ms
NFCorpus
Biomedical literature search from NutritionFacts.org
Corpus: 3,593 Queries: 323
Quality
ndcg at 10 0.3399
map at 10 0.1314
mrr at 10 0.5457
Performance L4 b1 c16
Corpus 71.1K tok/s
Corpus p50 65.9ms
Query 1.9K tok/s
Query p50 41.1ms
SCIDOCS
Citation prediction, document classification, and recommendation for scientific papers
Corpus: 25,656 Queries: 1,000
Performance L4 b1 c16
Corpus 47.1K tok/s
Corpus p50 59.3ms
Query 4.4K tok/s
Query p50 39.9ms
SciFact
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
Performance L4 b1 c16
Corpus 59.3K tok/s
Corpus p50 67.2ms
Query 6.1K tok/s
Query p50 42.4ms
StackOverflowQA
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
Performance L4 b1 c16
Corpus 53.1K tok/s
Corpus p50 62.1ms
Query 97.0K tok/s
Query p50 46.1ms