BAAI/bge-m3
For more details please refer to our github repo: https://github.com/FlagOpen/FlagEmbedding
Benchmarks
CQADupstackPhysicsRetrieval
Duplicate question retrieval from StackExchange Physics
Corpus: 38,314 Queries: 1,039
multivector
Performance L4 b1 c16
Corpus TPS 28.0K
Corpus p50 70.7ms
Query TPS 3.2K
Query p50 44.7ms
dense
Performance L4 b1 c16
Corpus TPS 27.3K
Corpus p50 74.8ms
Query TPS 3.1K
Query p50 45.2ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 27.0K
Corpus p50 75.3ms
Query TPS 2.6K
Query p50 55.2ms
sparse
Performance L4 b1 c16
Corpus TPS 27.8K
Corpus p50 70.8ms
Query TPS 3.3K
Query p50 44.9ms
CosQA
Code search with natural language queries
Corpus: 6,267 Queries: 500
multivector
Performance L4 b1 c16
Corpus TPS 15.3K
Corpus p50 51.4ms
Query TPS 1.8K
Query p50 45.8ms
dense
Performance L4 b1 c16
Corpus TPS 15.8K
Corpus p50 51.3ms
Query TPS 1.7K
Query p50 47.9ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 13.5K
Corpus p50 58.4ms
Query TPS 1.5K
Query p50 57.6ms
sparse
Performance L4 b1 c16
Corpus TPS 16.9K
Corpus p50 51.0ms
Query TPS 1.9K
Query p50 43.7ms
FiQA2018
Financial opinion mining and question answering
Corpus: 57,599 Queries: 648
multivector
Performance L4 b1 c16
Corpus TPS 33.8K
Corpus p50 77.8ms
Query TPS 3.6K
Query p50 44.8ms
dense
Performance L4 b1 c16
Corpus TPS 28.6K
Corpus p50 87.3ms
Query TPS 3.1K
Query p50 48.4ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 26.8K
Corpus p50 89.3ms
Query TPS 2.9K
Query p50 55.8ms
sparse
Performance L4 b1 c16
Corpus TPS 30.2K
Corpus p50 81.3ms
Query TPS 3.3K
Query p50 46.2ms
LegalBenchConsumerContractsQA
Question answering on consumer contracts
Corpus: 153 Queries: 396
multivector
Performance L4 b1 c16
Corpus TPS 39.1K
Corpus p50 216.2ms
Query TPS 5.1K
Query p50 44.0ms
dense
Performance L4 b1 c16
Corpus TPS 39.5K
Corpus p50 210.7ms
Query TPS 4.2K
Query p50 51.7ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 36.7K
Corpus p50 224.8ms
Query TPS 3.7K
Query p50 59.5ms
sparse
Performance L4 b1 c16
Corpus TPS 40.6K
Corpus p50 208.1ms
Query TPS 5.1K
Query p50 45.0ms
NFCorpus
Biomedical literature search from NutritionFacts.org
Corpus: 3,593 Queries: 323
default
Quality
ndcg at 10 0.3144
map at 10 0.1174
mrr at 10 0.5243
Performance A10G b1 c4
Corpus TPS 127
Corpus p50 9.4s
Query TPS 39
Query p50 537.6ms
Performance L4 b1 c16
Corpus TPS 38.1K
Corpus p50 100.0ms
Query TPS 1.5K
Query p50 41.0ms
multivector
Performance L4 b1 c16
Corpus TPS 37.9K
Corpus p50 126.6ms
Query TPS 1.5K
Query p50 46.1ms
dense
Performance L4 b1 c16
Corpus TPS 34.2K
Corpus p50 134.0ms
Query TPS 937
Query p50 64.2ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 34.4K
Corpus p50 137.4ms
Query TPS 1.2K
Query p50 56.8ms
sparse
Performance L4 b1 c16
Corpus TPS 40.0K
Corpus p50 124.4ms
Query TPS 1.4K
Query p50 45.0ms
NanoFiQA2018Retrieval
Smaller subset of the FiQA financial QA dataset
multivector
Performance L4 b1 c16
Corpus TPS 30.6K
Corpus p50 86.6ms
Query TPS 2.7K
Query p50 54.1ms
dense
Performance L4 b1 c16
Corpus TPS 28.1K
Corpus p50 89.9ms
Query TPS 2.5K
Query p50 59.2ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 30.9K
Corpus p50 86.4ms
Query TPS 2.8K
Query p50 54.9ms
sparse
Performance L4 b1 c16
Corpus TPS 30.3K
Corpus p50 87.5ms
Query TPS 2.7K
Query p50 53.3ms
default
Quality
ndcg at 10 0.5726
map at 10 0.4957
mrr at 10 0.6467
Performance L4 b1 c16
Corpus TPS 31.2K
Corpus p50 68.9ms
Query TPS 2.9K
Query p50 43.3ms
SCIDOCS
Citation prediction, document classification, and recommendation for scientific papers
Corpus: 25,656 Queries: 1,000
multivector
Performance L4 b1 c16
Corpus TPS 34.4K
Corpus p50 90.7ms
Query TPS 3.4K
Query p50 43.0ms
dense
Performance L4 b1 c16
Corpus TPS 33.2K
Corpus p50 94.8ms
Query TPS 2.5K
Query p50 54.2ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 31.0K
Corpus p50 101.1ms
Query TPS 2.7K
Query p50 55.7ms
sparse
Performance L4 b1 c16
Corpus TPS 32.5K
Corpus p50 93.4ms
Query TPS 3.4K
Query p50 44.7ms
SciFact
Scientific claim verification using research literature
Corpus: 5,183 Queries: 300
multivector
Performance L4 b1 c16
Corpus TPS 38.2K
Corpus p50 116.5ms
Query TPS 4.9K
Query p50 44.4ms
dense
Performance L4 b1 c16
Corpus TPS 37.9K
Corpus p50 118.2ms
Query TPS 4.2K
Query p50 48.5ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 32.2K
Corpus p50 126.9ms
Query TPS 3.9K
Query p50 57.5ms
sparse
Performance L4 b1 c16
Corpus TPS 37.8K
Corpus p50 117.6ms
Query TPS 4.2K
Query p50 47.7ms
StackOverflowQA
Programming question answering from Stack Overflow
Corpus: 19,931 Queries: 1,994
multivector
Performance L4 b1 c16
Corpus TPS 33.2K
Corpus p50 107.6ms
Query TPS 33.8K
Query p50 137.6ms
dense
Performance L4 b1 c16
Corpus TPS 33.4K
Corpus p50 104.7ms
Query TPS 33.8K
Query p50 138.5ms
multivector_candidates-k-50_candidates-model-Alibaba-NLP__gte-multilingual-base
Performance L4 b1 c16
Corpus TPS 29.5K
Corpus p50 118.6ms
Query TPS 31.5K
Query p50 147.9ms
sparse
Performance L4 b1 c16
Corpus TPS 32.7K
Corpus p50 108.4ms
Query TPS 34.8K
Query p50 137.6ms