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 31.1K tok/s
Corpus p50 57.0ms
Query 3.2K tok/s
Query p50 52.8ms
CosQA
Code search with natural language queries
Corpus: 6,267 Queries: 500
Performance L4 b1 c16
Corpus 13.5K tok/s
Corpus p50 51.4ms
Query 2.1K tok/s
Query p50 45.3ms
FiQA2018
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
Performance L4 b1 c16
Corpus 37.0K tok/s
Corpus p50 59.3ms
Query 3.8K tok/s
Query p50 47.5ms
LegalBenchConsumerContractsQA
Question answering on consumer contracts
Corpus: 153 Queries: 396
Performance L4 b1 c16
Corpus 91.6K tok/s
Corpus p50 73.7ms
Query 5.2K tok/s
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 61.6K tok/s
Corpus p50 68.8ms
Query 1.6K tok/s
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 42.4K tok/s
Corpus p50 60.4ms
Query 3.5K tok/s
Query p50 49.6ms
SciFact
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
Performance L4 b1 c16
Corpus 54.4K tok/s
Corpus p50 68.2ms
Query 5.1K tok/s
Query p50 50.1ms
StackOverflowQA
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
Performance L4 b1 c16
Corpus 46.0K tok/s
Corpus p50 66.2ms
Query 69.8K tok/s
Query p50 66.0ms