metadata
language:
- en
license: apache-2.0
library_name: transformers
tags:
- language
- granite
- embeddings
model-index:
- name: ibm-granite/granite-embedding-125m-english
results:
- dataset:
type: mteb/arguana
name: MTEB ArguaAna
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.33642
- type: map_at_10
value: 0.49716
- type: map_at_100
value: 0.50519
- type: map_at_1000
value: 0.50521
- type: map_at_3
value: 0.45057
- type: map_at_5
value: 0.47774
- type: mrr_at_1
value: 0.34922
- type: mrr_at_10
value: 0.50197
- type: mrr_at_100
value: 0.50992
- type: mrr_at_1000
value: 0.50994
- type: mrr_at_3
value: 0.45484
- type: mrr_at_5
value: 0.48272
- type: ndcg_at_1
value: 0.33642
- type: ndcg_at_10
value: 0.58401
- type: ndcg_at_100
value: 0.6157
- type: ndcg_at_1000
value: 0.61608
- type: ndcg_at_3
value: 0.48825
- type: ndcg_at_5
value: 0.53689
- type: precision_at_1
value: 0.33642
- type: precision_at_10
value: 0.08606
- type: precision_at_100
value: 0.00994
- type: precision_at_1000
value: 0.001
- type: precision_at_3
value: 0.19915
- type: precision_at_5
value: 0.14296
- type: recall_at_1
value: 0.33642
- type: recall_at_10
value: 0.8606
- type: recall_at_100
value: 0.9936
- type: recall_at_1000
value: 0.99644
- type: recall_at_3
value: 0.59744
- type: recall_at_5
value: 0.71479
- dataset:
type: mteb/climate-fever
name: MTEB ClimateFEVER
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.1457
- type: map_at_10
value: 0.24102
- type: map_at_100
value: 0.25826
- type: map_at_1000
value: 0.26021
- type: map_at_3
value: 0.20346
- type: map_at_5
value: 0.22228
- type: mrr_at_1
value: 0.32573
- type: mrr_at_10
value: 0.44411
- type: mrr_at_100
value: 0.45176
- type: mrr_at_1000
value: 0.45209
- type: mrr_at_3
value: 0.4126
- type: mrr_at_5
value: 0.43312
- type: ndcg_at_1
value: 0.32573
- type: ndcg_at_10
value: 0.3315
- type: ndcg_at_100
value: 0.39898
- type: ndcg_at_1000
value: 0.43151
- type: ndcg_at_3
value: 0.27683
- type: ndcg_at_5
value: 0.29538
- type: precision_at_1
value: 0.32573
- type: precision_at_10
value: 0.10176
- type: precision_at_100
value: 0.01754
- type: precision_at_1000
value: 0.00236
- type: precision_at_3
value: 0.20347
- type: precision_at_5
value: 0.15505
- type: recall_at_1
value: 0.1457
- type: recall_at_10
value: 0.38825
- type: recall_at_100
value: 0.62237
- type: recall_at_1000
value: 0.8022
- type: recall_at_3
value: 0.25245
- type: recall_at_5
value: 0.30821
- dataset:
type: mteb/cqadupstack-android
name: MTEB CQADupstackAndroidRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.36964
- type: map_at_10
value: 0.5043
- type: map_at_100
value: 0.52066
- type: map_at_1000
value: 0.52175
- type: map_at_3
value: 0.46001
- type: map_at_5
value: 0.48312
- type: mrr_at_1
value: 0.45923
- type: mrr_at_10
value: 0.56733
- type: mrr_at_100
value: 0.57292
- type: mrr_at_1000
value: 0.57321
- type: mrr_at_3
value: 0.54053
- type: mrr_at_5
value: 0.55556
- type: ndcg_at_1
value: 0.45923
- type: ndcg_at_10
value: 0.57667
- type: ndcg_at_100
value: 0.62373
- type: ndcg_at_1000
value: 0.6368
- type: ndcg_at_3
value: 0.51843
- type: ndcg_at_5
value: 0.54257
- type: precision_at_1
value: 0.45923
- type: precision_at_10
value: 0.11316
- type: precision_at_100
value: 0.01705
- type: precision_at_1000
value: 0.00216
- type: precision_at_3
value: 0.2537
- type: precision_at_5
value: 0.1814
- type: recall_at_1
value: 0.36964
- type: recall_at_10
value: 0.71234
- type: recall_at_100
value: 0.90421
- type: recall_at_1000
value: 0.98296
- type: recall_at_3
value: 0.53655
- type: recall_at_5
value: 0.60996
- dataset:
type: mteb/cqadupstack-english
name: MTEB CQADupstackEnglishRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.36198
- type: map_at_10
value: 0.49199
- type: map_at_100
value: 0.50602
- type: map_at_1000
value: 0.50736
- type: map_at_3
value: 0.45678
- type: map_at_5
value: 0.47605
- type: mrr_at_1
value: 0.45478
- type: mrr_at_10
value: 0.55075
- type: mrr_at_100
value: 0.55656
- type: mrr_at_1000
value: 0.55688
- type: mrr_at_3
value: 0.52887
- type: mrr_at_5
value: 0.54282
- type: ndcg_at_1
value: 0.45478
- type: ndcg_at_10
value: 0.55505
- type: ndcg_at_100
value: 0.59606
- type: ndcg_at_1000
value: 0.61255
- type: ndcg_at_3
value: 0.51124
- type: ndcg_at_5
value: 0.53166
- type: precision_at_1
value: 0.45478
- type: precision_at_10
value: 0.10752
- type: precision_at_100
value: 0.01666
- type: precision_at_1000
value: 0.00211
- type: precision_at_3
value: 0.25053
- type: precision_at_5
value: 0.17694
- type: recall_at_1
value: 0.36198
- type: recall_at_10
value: 0.66465
- type: recall_at_100
value: 0.83632
- type: recall_at_1000
value: 0.93276
- type: recall_at_3
value: 0.53207
- type: recall_at_5
value: 0.59169
- dataset:
type: mteb/cqadupstack-gaming
name: MTEB CQADupstackGamingRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.44157
- type: map_at_10
value: 0.57753
- type: map_at_100
value: 0.58698
- type: map_at_1000
value: 0.5874
- type: map_at_3
value: 0.54223
- type: map_at_5
value: 0.56307
- type: mrr_at_1
value: 0.50094
- type: mrr_at_10
value: 0.607
- type: mrr_at_100
value: 0.6126
- type: mrr_at_1000
value: 0.6128
- type: mrr_at_3
value: 0.58265
- type: mrr_at_5
value: 0.59817
- type: ndcg_at_1
value: 0.50094
- type: ndcg_at_10
value: 0.63641
- type: ndcg_at_100
value: 0.67055
- type: ndcg_at_1000
value: 0.67855
- type: ndcg_at_3
value: 0.58022
- type: ndcg_at_5
value: 0.6097
- type: precision_at_1
value: 0.50094
- type: precision_at_10
value: 0.10182
- type: precision_at_100
value: 0.01278
- type: precision_at_1000
value: 0.00138
- type: precision_at_3
value: 0.2581
- type: precision_at_5
value: 0.17755
- type: recall_at_1
value: 0.44157
- type: recall_at_10
value: 0.7778
- type: recall_at_100
value: 0.92244
- type: recall_at_1000
value: 0.9781
- type: recall_at_3
value: 0.63087
- type: recall_at_5
value: 0.70172
- dataset:
type: mteb/cqadupstack-gis
name: MTEB CQADupstackGisRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.29532
- type: map_at_10
value: 0.40214
- type: map_at_100
value: 0.41289
- type: map_at_1000
value: 0.41359
- type: map_at_3
value: 0.37086
- type: map_at_5
value: 0.38889
- type: mrr_at_1
value: 0.3209
- type: mrr_at_10
value: 0.42423
- type: mrr_at_100
value: 0.43342
- type: mrr_at_1000
value: 0.43395
- type: mrr_at_3
value: 0.39736
- type: mrr_at_5
value: 0.41307
- type: ndcg_at_1
value: 0.3209
- type: ndcg_at_10
value: 0.46075
- type: ndcg_at_100
value: 0.5103
- type: ndcg_at_1000
value: 0.52668
- type: ndcg_at_3
value: 0.40149
- type: ndcg_at_5
value: 0.43111
- type: precision_at_1
value: 0.3209
- type: precision_at_10
value: 0.07141
- type: precision_at_100
value: 0.01018
- type: precision_at_1000
value: 0.00118
- type: precision_at_3
value: 0.17175
- type: precision_at_5
value: 0.12068
- type: recall_at_1
value: 0.29532
- type: recall_at_10
value: 0.62025
- type: recall_at_100
value: 0.83829
- type: recall_at_1000
value: 0.95995
- type: recall_at_3
value: 0.4603
- type: recall_at_5
value: 0.53089
- dataset:
type: mteb/cqadupstack-mathematica
name: MTEB CQADupstackMathematicaRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.18944
- type: map_at_10
value: 0.29611
- type: map_at_100
value: 0.31063
- type: map_at_1000
value: 0.31174
- type: map_at_3
value: 0.26098
- type: map_at_5
value: 0.28151
- type: mrr_at_1
value: 0.23756
- type: mrr_at_10
value: 0.34491
- type: mrr_at_100
value: 0.35457
- type: mrr_at_1000
value: 0.35512
- type: mrr_at_3
value: 0.3126
- type: mrr_at_5
value: 0.3317
- type: ndcg_at_1
value: 0.23756
- type: ndcg_at_10
value: 0.36015
- type: ndcg_at_100
value: 0.42175
- type: ndcg_at_1000
value: 0.44607
- type: ndcg_at_3
value: 0.29725
- type: ndcg_at_5
value: 0.32879
- type: precision_at_1
value: 0.23756
- type: precision_at_10
value: 0.06928
- type: precision_at_100
value: 0.01153
- type: precision_at_1000
value: 0.00149
- type: precision_at_3
value: 0.14635
- type: precision_at_5
value: 0.1107
- type: recall_at_1
value: 0.18944
- type: recall_at_10
value: 0.50691
- type: recall_at_100
value: 0.76503
- type: recall_at_1000
value: 0.93624
- type: recall_at_3
value: 0.33611
- type: recall_at_5
value: 0.41427
- dataset:
type: mteb/cqadupstack-physics
name: MTEB CQADupstackPhysicsRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.33824
- type: map_at_10
value: 0.46868
- type: map_at_100
value: 0.48306
- type: map_at_1000
value: 0.48406
- type: map_at_3
value: 0.43335
- type: map_at_5
value: 0.45279
- type: mrr_at_1
value: 0.42348
- type: mrr_at_10
value: 0.52972
- type: mrr_at_100
value: 0.53707
- type: mrr_at_1000
value: 0.53734
- type: mrr_at_3
value: 0.50722
- type: mrr_at_5
value: 0.52012
- type: ndcg_at_1
value: 0.42348
- type: ndcg_at_10
value: 0.53504
- type: ndcg_at_100
value: 0.58899
- type: ndcg_at_1000
value: 0.60323
- type: ndcg_at_3
value: 0.48478
- type: ndcg_at_5
value: 0.5079
- type: precision_at_1
value: 0.42348
- type: precision_at_10
value: 0.0975
- type: precision_at_100
value: 0.01466
- type: precision_at_1000
value: 0.00177
- type: precision_at_3
value: 0.23741
- type: precision_at_5
value: 0.16439
- type: recall_at_1
value: 0.33824
- type: recall_at_10
value: 0.67142
- type: recall_at_100
value: 0.89134
- type: recall_at_1000
value: 0.97816
- type: recall_at_3
value: 0.52305
- type: recall_at_5
value: 0.58804
- dataset:
type: mteb/cqadupstack-programmers
name: MTEB CQADupstackProgrammersRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.30125
- type: map_at_10
value: 0.42119
- type: map_at_100
value: 0.43599
- type: map_at_1000
value: 0.4369
- type: map_at_3
value: 0.38018
- type: map_at_5
value: 0.40368
- type: mrr_at_1
value: 0.37557
- type: mrr_at_10
value: 0.47573
- type: mrr_at_100
value: 0.4846
- type: mrr_at_1000
value: 0.48499
- type: mrr_at_3
value: 0.44654
- type: mrr_at_5
value: 0.4644
- type: ndcg_at_1
value: 0.37557
- type: ndcg_at_10
value: 0.48743
- type: ndcg_at_100
value: 0.54458
- type: ndcg_at_1000
value: 0.56076
- type: ndcg_at_3
value: 0.42573
- type: ndcg_at_5
value: 0.45528
- type: precision_at_1
value: 0.37557
- type: precision_at_10
value: 0.09269
- type: precision_at_100
value: 0.01401
- type: precision_at_1000
value: 0.0017
- type: precision_at_3
value: 0.20624
- type: precision_at_5
value: 0.15068
- type: recall_at_1
value: 0.30125
- type: recall_at_10
value: 0.62619
- type: recall_at_100
value: 0.86574
- type: recall_at_1000
value: 0.97102
- type: recall_at_3
value: 0.45437
- type: recall_at_5
value: 0.53197
- dataset:
type: mteb/cqadupstack-stats
name: MTEB CQADupstackStatsRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.29193
- type: map_at_10
value: 0.37529
- type: map_at_100
value: 0.38614
- type: map_at_1000
value: 0.38714
- type: map_at_3
value: 0.34897
- type: map_at_5
value: 0.36273
- type: mrr_at_1
value: 0.32669
- type: mrr_at_10
value: 0.40288
- type: mrr_at_100
value: 0.41177
- type: mrr_at_1000
value: 0.41241
- type: mrr_at_3
value: 0.38037
- type: mrr_at_5
value: 0.39195
- type: ndcg_at_1
value: 0.32669
- type: ndcg_at_10
value: 0.42353
- type: ndcg_at_100
value: 0.47424
- type: ndcg_at_1000
value: 0.4959
- type: ndcg_at_3
value: 0.37604
- type: ndcg_at_5
value: 0.39682
- type: precision_at_1
value: 0.32669
- type: precision_at_10
value: 0.06871
- type: precision_at_100
value: 0.01008
- type: precision_at_1000
value: 0.00126
- type: precision_at_3
value: 0.16309
- type: precision_at_5
value: 0.11288
- type: recall_at_1
value: 0.29193
- type: recall_at_10
value: 0.54159
- type: recall_at_100
value: 0.77267
- type: recall_at_1000
value: 0.92805
- type: recall_at_3
value: 0.41014
- type: recall_at_5
value: 0.46248
- dataset:
type: mteb/cqadupstack-tex
name: MTEB CQADupstackTexRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.21217
- type: map_at_10
value: 0.30848
- type: map_at_100
value: 0.32173
- type: map_at_1000
value: 0.32296
- type: map_at_3
value: 0.27882
- type: map_at_5
value: 0.29537
- type: mrr_at_1
value: 0.25946
- type: mrr_at_10
value: 0.35091
- type: mrr_at_100
value: 0.36047
- type: mrr_at_1000
value: 0.36111
- type: mrr_at_3
value: 0.32485
- type: mrr_at_5
value: 0.33964
- type: ndcg_at_1
value: 0.25946
- type: ndcg_at_10
value: 0.3655
- type: ndcg_at_100
value: 0.42328
- type: ndcg_at_1000
value: 0.44783
- type: ndcg_at_3
value: 0.31463
- type: ndcg_at_5
value: 0.33803
- type: precision_at_1
value: 0.25946
- type: precision_at_10
value: 0.06793
- type: precision_at_100
value: 0.01138
- type: precision_at_1000
value: 0.00155
- type: precision_at_3
value: 0.1513
- type: precision_at_5
value: 0.10991
- type: recall_at_1
value: 0.21217
- type: recall_at_10
value: 0.49327
- type: recall_at_100
value: 0.7472
- type: recall_at_1000
value: 0.91637
- type: recall_at_3
value: 0.34993
- type: recall_at_5
value: 0.41029
- dataset:
type: mteb/cqadupstack-unix
name: MTEB CQADupstackUnixRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.34303
- type: map_at_10
value: 0.45312
- type: map_at_100
value: 0.46563
- type: map_at_1000
value: 0.4664
- type: map_at_3
value: 0.4143
- type: map_at_5
value: 0.43633
- type: mrr_at_1
value: 0.40112
- type: mrr_at_10
value: 0.49097
- type: mrr_at_100
value: 0.49966
- type: mrr_at_1000
value: 0.50006
- type: mrr_at_3
value: 0.46129
- type: mrr_at_5
value: 0.47901
- type: ndcg_at_1
value: 0.40112
- type: ndcg_at_10
value: 0.513
- type: ndcg_at_100
value: 0.56534
- type: ndcg_at_1000
value: 0.58048
- type: ndcg_at_3
value: 0.4491
- type: ndcg_at_5
value: 0.48048
- type: precision_at_1
value: 0.40112
- type: precision_at_10
value: 0.08806
- type: precision_at_100
value: 0.01266
- type: precision_at_1000
value: 0.00149
- type: precision_at_3
value: 0.20211
- type: precision_at_5
value: 0.14496
- type: recall_at_1
value: 0.34303
- type: recall_at_10
value: 0.65508
- type: recall_at_100
value: 0.8753
- type: recall_at_1000
value: 0.9742
- type: recall_at_3
value: 0.48465
- type: recall_at_5
value: 0.56374
- dataset:
type: mteb/cqadupstack-webmasters
name: MTEB CQADupstackWebmastersRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.30312
- type: map_at_10
value: 0.40931
- type: map_at_100
value: 0.42893
- type: map_at_1000
value: 0.4312
- type: map_at_3
value: 0.37527
- type: map_at_5
value: 0.3936
- type: mrr_at_1
value: 0.36364
- type: mrr_at_10
value: 0.45677
- type: mrr_at_100
value: 0.46753
- type: mrr_at_1000
value: 0.46787
- type: mrr_at_3
value: 0.42918
- type: mrr_at_5
value: 0.4443
- type: ndcg_at_1
value: 0.36364
- type: ndcg_at_10
value: 0.47301
- type: ndcg_at_100
value: 0.53698
- type: ndcg_at_1000
value: 0.55503
- type: ndcg_at_3
value: 0.41875
- type: ndcg_at_5
value: 0.44316
- type: precision_at_1
value: 0.36364
- type: precision_at_10
value: 0.09032
- type: precision_at_100
value: 0.01806
- type: precision_at_1000
value: 0.00258
- type: precision_at_3
value: 0.19499
- type: precision_at_5
value: 0.1415
- type: recall_at_1
value: 0.30312
- type: recall_at_10
value: 0.59418
- type: recall_at_100
value: 0.8656
- type: recall_at_1000
value: 0.97412
- type: recall_at_3
value: 0.44251
- type: recall_at_5
value: 0.50457
- dataset:
type: mteb/cqadupstack-wordpress
name: MTEB CQADupstackWordpressRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.23851
- type: map_at_10
value: 0.33429
- type: map_at_100
value: 0.34482
- type: map_at_1000
value: 0.3457
- type: map_at_3
value: 0.30271
- type: map_at_5
value: 0.31905
- type: mrr_at_1
value: 0.25693
- type: mrr_at_10
value: 0.35383
- type: mrr_at_100
value: 0.36295
- type: mrr_at_1000
value: 0.36346
- type: mrr_at_3
value: 0.32532
- type: mrr_at_5
value: 0.3402
- type: ndcg_at_1
value: 0.25693
- type: ndcg_at_10
value: 0.39196
- type: ndcg_at_100
value: 0.44501
- type: ndcg_at_1000
value: 0.46482
- type: ndcg_at_3
value: 0.33
- type: ndcg_at_5
value: 0.35736
- type: precision_at_1
value: 0.25693
- type: precision_at_10
value: 0.06433
- type: precision_at_100
value: 0.00989
- type: precision_at_1000
value: 0.00128
- type: precision_at_3
value: 0.14295
- type: precision_at_5
value: 0.10277
- type: recall_at_1
value: 0.23851
- type: recall_at_10
value: 0.55036
- type: recall_at_100
value: 0.79592
- type: recall_at_1000
value: 0.94283
- type: recall_at_3
value: 0.38435
- type: recall_at_5
value: 0.44872
- dataset:
type: mteb/dbpedia
name: MTEB DBPedia
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.0871
- type: map_at_10
value: 0.19218
- type: map_at_100
value: 0.26291
- type: map_at_1000
value: 0.27985
- type: map_at_3
value: 0.13974
- type: map_at_5
value: 0.16104
- type: mrr_at_1
value: 0.6725
- type: mrr_at_10
value: 0.75037
- type: mrr_at_100
value: 0.75318
- type: mrr_at_1000
value: 0.75325
- type: mrr_at_3
value: 0.73833
- type: mrr_at_5
value: 0.74308
- type: ndcg_at_1
value: 0.54375
- type: ndcg_at_10
value: 0.39409
- type: ndcg_at_100
value: 0.44382
- type: ndcg_at_1000
value: 0.52485
- type: ndcg_at_3
value: 0.44463
- type: ndcg_at_5
value: 0.41276
- type: precision_at_1
value: 0.6725
- type: precision_at_10
value: 0.3055
- type: precision_at_100
value: 0.09588
- type: precision_at_1000
value: 0.02118
- type: precision_at_3
value: 0.48167
- type: precision_at_5
value: 0.394
- type: recall_at_1
value: 0.0871
- type: recall_at_10
value: 0.2527
- type: recall_at_100
value: 0.5185
- type: recall_at_1000
value: 0.76491
- type: recall_at_3
value: 0.15516
- type: recall_at_5
value: 0.18907
- dataset:
type: mteb/fever
name: MTEB FEVER
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.78993
- type: map_at_10
value: 0.8502
- type: map_at_100
value: 0.85186
- type: map_at_1000
value: 0.852
- type: map_at_3
value: 0.8437
- type: map_at_5
value: 0.84812
- type: mrr_at_1
value: 0.85179
- type: mrr_at_10
value: 0.90744
- type: mrr_at_100
value: 0.90799
- type: mrr_at_1000
value: 0.90801
- type: mrr_at_3
value: 0.90322
- type: mrr_at_5
value: 0.90622
- type: ndcg_at_1
value: 0.85179
- type: ndcg_at_10
value: 0.88229
- type: ndcg_at_100
value: 0.8884
- type: ndcg_at_1000
value: 0.89116
- type: ndcg_at_3
value: 0.87304
- type: ndcg_at_5
value: 0.87862
- type: precision_at_1
value: 0.85179
- type: precision_at_10
value: 0.10129
- type: precision_at_100
value: 0.0106
- type: precision_at_1000
value: 0.0011
- type: precision_at_3
value: 0.32543
- type: precision_at_5
value: 0.19931
- type: recall_at_1
value: 0.78993
- type: recall_at_10
value: 0.92685
- type: recall_at_100
value: 0.9516
- type: recall_at_1000
value: 0.96943
- type: recall_at_3
value: 0.89965
- type: recall_at_5
value: 0.91562
- dataset:
type: mteb/fiqa
name: MTEB FiQA2018
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.22586
- type: map_at_10
value: 0.36836
- type: map_at_100
value: 0.38863
- type: map_at_1000
value: 0.39041
- type: map_at_3
value: 0.32445
- type: map_at_5
value: 0.34951
- type: mrr_at_1
value: 0.44599
- type: mrr_at_10
value: 0.53471
- type: mrr_at_100
value: 0.54186
- type: mrr_at_1000
value: 0.54223
- type: mrr_at_3
value: 0.51157
- type: mrr_at_5
value: 0.52423
- type: ndcg_at_1
value: 0.44599
- type: ndcg_at_10
value: 0.44931
- type: ndcg_at_100
value: 0.51914
- type: ndcg_at_1000
value: 0.54674
- type: ndcg_at_3
value: 0.41597
- type: ndcg_at_5
value: 0.42611
- type: precision_at_1
value: 0.44599
- type: precision_at_10
value: 0.12346
- type: precision_at_100
value: 0.01951
- type: precision_at_1000
value: 0.00244
- type: precision_at_3
value: 0.27623
- type: precision_at_5
value: 0.20093
- type: recall_at_1
value: 0.22586
- type: recall_at_10
value: 0.5152
- type: recall_at_100
value: 0.77251
- type: recall_at_1000
value: 0.93503
- type: recall_at_3
value: 0.37802
- type: recall_at_5
value: 0.4386
- dataset:
type: mteb/hotpotqa
name: MTEB HotpotQA
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.38177
- type: map_at_10
value: 0.59021
- type: map_at_100
value: 0.59924
- type: map_at_1000
value: 0.59989
- type: map_at_3
value: 0.55553
- type: map_at_5
value: 0.57773
- type: mrr_at_1
value: 0.76354
- type: mrr_at_10
value: 0.827
- type: mrr_at_100
value: 0.82887
- type: mrr_at_1000
value: 0.82896
- type: mrr_at_3
value: 0.8172
- type: mrr_at_5
value: 0.82338
- type: ndcg_at_1
value: 0.76354
- type: ndcg_at_10
value: 0.67775
- type: ndcg_at_100
value: 0.70849
- type: ndcg_at_1000
value: 0.7215
- type: ndcg_at_3
value: 0.629
- type: ndcg_at_5
value: 0.65679
- type: precision_at_1
value: 0.76354
- type: precision_at_10
value: 0.14176
- type: precision_at_100
value: 0.01656
- type: precision_at_1000
value: 0.00183
- type: precision_at_3
value: 0.40113
- type: precision_at_5
value: 0.26255
- type: recall_at_1
value: 0.38177
- type: recall_at_10
value: 0.70878
- type: recall_at_100
value: 0.82822
- type: recall_at_1000
value: 0.91472
- type: recall_at_3
value: 0.60169
- type: recall_at_5
value: 0.65638
- dataset:
type: mteb/msmarco
name: MTEB MSMARCO
config: default
split: dev
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.15062
- type: map_at_10
value: 0.26008
- type: map_at_100
value: 0.27305
- type: map_at_1000
value: 0.27373
- type: map_at_3
value: 0.22236
- type: map_at_5
value: 0.24362
- type: mrr_at_1
value: 0.15444
- type: mrr_at_10
value: 0.26458
- type: mrr_at_100
value: 0.27718
- type: mrr_at_1000
value: 0.2778
- type: mrr_at_3
value: 0.22701
- type: mrr_at_5
value: 0.24844
- type: ndcg_at_1
value: 0.15444
- type: ndcg_at_10
value: 0.32495
- type: ndcg_at_100
value: 0.38957
- type: ndcg_at_1000
value: 0.40684
- type: ndcg_at_3
value: 0.24745
- type: ndcg_at_5
value: 0.2856
- type: precision_at_1
value: 0.15444
- type: precision_at_10
value: 0.05486
- type: precision_at_100
value: 0.00875
- type: precision_at_1000
value: 0.00102
- type: precision_at_3
value: 0.1086
- type: precision_at_5
value: 0.08441
- type: recall_at_1
value: 0.15062
- type: recall_at_10
value: 0.5272
- type: recall_at_100
value: 0.83006
- type: recall_at_1000
value: 0.96263
- type: recall_at_3
value: 0.31556
- type: recall_at_5
value: 0.40706
- dataset:
type: mteb/nfcorpus
name: MTEB NFCorpus
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.06126
- type: map_at_10
value: 0.14152
- type: map_at_100
value: 0.1827
- type: map_at_1000
value: 0.1988
- type: map_at_3
value: 0.10301
- type: map_at_5
value: 0.12085
- type: mrr_at_1
value: 0.47988
- type: mrr_at_10
value: 0.5692
- type: mrr_at_100
value: 0.57428
- type: mrr_at_1000
value: 0.57482
- type: mrr_at_3
value: 0.55315
- type: mrr_at_5
value: 0.56352
- type: ndcg_at_1
value: 0.45356
- type: ndcg_at_10
value: 0.3725
- type: ndcg_at_100
value: 0.34496
- type: ndcg_at_1000
value: 0.43374
- type: ndcg_at_3
value: 0.42643
- type: ndcg_at_5
value: 0.40882
- type: precision_at_1
value: 0.47368
- type: precision_at_10
value: 0.2774
- type: precision_at_100
value: 0.09071
- type: precision_at_1000
value: 0.02226
- type: precision_at_3
value: 0.40144
- type: precision_at_5
value: 0.35913
- type: recall_at_1
value: 0.06126
- type: recall_at_10
value: 0.18427
- type: recall_at_100
value: 0.35018
- type: recall_at_1000
value: 0.6766
- type: recall_at_3
value: 0.11706
- type: recall_at_5
value: 0.14419
- dataset:
type: mteb/nq
name: MTEB NQ
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.33053
- type: map_at_10
value: 0.49739
- type: map_at_100
value: 0.50626
- type: map_at_1000
value: 0.50647
- type: map_at_3
value: 0.4491
- type: map_at_5
value: 0.4783
- type: mrr_at_1
value: 0.37254
- type: mrr_at_10
value: 0.52222
- type: mrr_at_100
value: 0.52855
- type: mrr_at_1000
value: 0.52869
- type: mrr_at_3
value: 0.48445
- type: mrr_at_5
value: 0.50834
- type: ndcg_at_1
value: 0.37254
- type: ndcg_at_10
value: 0.58044
- type: ndcg_at_100
value: 0.61613
- type: ndcg_at_1000
value: 0.62046
- type: ndcg_at_3
value: 0.49219
- type: ndcg_at_5
value: 0.54037
- type: precision_at_1
value: 0.37254
- type: precision_at_10
value: 0.09655
- type: precision_at_100
value: 0.01167
- type: precision_at_1000
value: 0.00121
- type: precision_at_3
value: 0.22538
- type: precision_at_5
value: 0.16344
- type: recall_at_1
value: 0.33053
- type: recall_at_10
value: 0.8076
- type: recall_at_100
value: 0.95862
- type: recall_at_1000
value: 0.99044
- type: recall_at_3
value: 0.58157
- type: recall_at_5
value: 0.69235
- dataset:
type: mteb/quora
name: MTEB QuoraRetrieval
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.70056
- type: map_at_10
value: 0.84009
- type: map_at_100
value: 0.84661
- type: map_at_1000
value: 0.84678
- type: map_at_3
value: 0.81036
- type: map_at_5
value: 0.82923
- type: mrr_at_1
value: 0.8062
- type: mrr_at_10
value: 0.86971
- type: mrr_at_100
value: 0.87079
- type: mrr_at_1000
value: 0.8708
- type: mrr_at_3
value: 0.85943
- type: mrr_at_5
value: 0.86664
- type: ndcg_at_1
value: 0.8064
- type: ndcg_at_10
value: 0.87821
- type: ndcg_at_100
value: 0.89091
- type: ndcg_at_1000
value: 0.89202
- type: ndcg_at_3
value: 0.849
- type: ndcg_at_5
value: 0.86544
- type: precision_at_1
value: 0.8064
- type: precision_at_10
value: 0.13347
- type: precision_at_100
value: 0.01527
- type: precision_at_1000
value: 0.00157
- type: precision_at_3
value: 0.37153
- type: precision_at_5
value: 0.2448
- type: recall_at_1
value: 0.70056
- type: recall_at_10
value: 0.95148
- type: recall_at_100
value: 0.99474
- type: recall_at_1000
value: 0.99977
- type: recall_at_3
value: 0.86773
- type: recall_at_5
value: 0.91396
- dataset:
type: mteb/scidocs
name: MTEB SCIDOCS
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.05737
- type: map_at_10
value: 0.14896
- type: map_at_100
value: 0.17646
- type: map_at_1000
value: 0.1803
- type: map_at_3
value: 0.10474
- type: map_at_5
value: 0.12656
- type: mrr_at_1
value: 0.281
- type: mrr_at_10
value: 0.39579
- type: mrr_at_100
value: 0.40687
- type: mrr_at_1000
value: 0.40722
- type: mrr_at_3
value: 0.35917
- type: mrr_at_5
value: 0.38097
- type: ndcg_at_1
value: 0.281
- type: ndcg_at_10
value: 0.24146
- type: ndcg_at_100
value: 0.339
- type: ndcg_at_1000
value: 0.39728
- type: ndcg_at_3
value: 0.22721
- type: ndcg_at_5
value: 0.20015
- type: precision_at_1
value: 0.281
- type: precision_at_10
value: 0.1254
- type: precision_at_100
value: 0.02651
- type: precision_at_1000
value: 0.00404
- type: precision_at_3
value: 0.212
- type: precision_at_5
value: 0.176
- type: recall_at_1
value: 0.05737
- type: recall_at_10
value: 0.254
- type: recall_at_100
value: 0.53772
- type: recall_at_1000
value: 0.82013
- type: recall_at_3
value: 0.12897
- type: recall_at_5
value: 0.17855
- dataset:
type: mteb/scifact
name: MTEB SciFact
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.60011
- type: map_at_10
value: 0.70101
- type: map_at_100
value: 0.70687
- type: map_at_1000
value: 0.70699
- type: map_at_3
value: 0.67135
- type: map_at_5
value: 0.6878
- type: mrr_at_1
value: 0.62667
- type: mrr_at_10
value: 0.71022
- type: mrr_at_100
value: 0.71484
- type: mrr_at_1000
value: 0.71496
- type: mrr_at_3
value: 0.68944
- type: mrr_at_5
value: 0.69961
- type: ndcg_at_1
value: 0.62667
- type: ndcg_at_10
value: 0.7472
- type: ndcg_at_100
value: 0.76961
- type: ndcg_at_1000
value: 0.77294
- type: ndcg_at_3
value: 0.69776
- type: ndcg_at_5
value: 0.71964
- type: precision_at_1
value: 0.62667
- type: precision_at_10
value: 0.09933
- type: precision_at_100
value: 0.01103
- type: precision_at_1000
value: 0.00113
- type: precision_at_3
value: 0.27
- type: precision_at_5
value: 0.178
- type: recall_at_1
value: 0.60011
- type: recall_at_10
value: 0.878
- type: recall_at_100
value: 0.97333
- type: recall_at_1000
value: 1
- type: recall_at_3
value: 0.74839
- type: recall_at_5
value: 0.80028
- dataset:
type: mteb/touche2020
name: MTEB Touche2020
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.02152
- type: map_at_10
value: 0.07747
- type: map_at_100
value: 0.1364
- type: map_at_1000
value: 0.15235
- type: map_at_3
value: 0.04103
- type: map_at_5
value: 0.05482
- type: mrr_at_1
value: 0.26531
- type: mrr_at_10
value: 0.41399
- type: mrr_at_100
value: 0.43047
- type: mrr_at_1000
value: 0.43047
- type: mrr_at_3
value: 0.38776
- type: mrr_at_5
value: 0.40612
- type: ndcg_at_1
value: 0.23469
- type: ndcg_at_10
value: 0.20147
- type: ndcg_at_100
value: 0.3279
- type: ndcg_at_1000
value: 0.45324
- type: ndcg_at_3
value: 0.22555
- type: ndcg_at_5
value: 0.2097
- type: precision_at_1
value: 0.26531
- type: precision_at_10
value: 0.17755
- type: precision_at_100
value: 0.07082
- type: precision_at_1000
value: 0.01547
- type: precision_at_3
value: 0.2449
- type: precision_at_5
value: 0.21633
- type: recall_at_1
value: 0.02152
- type: recall_at_10
value: 0.13331
- type: recall_at_100
value: 0.4535
- type: recall_at_1000
value: 0.83447
- type: recall_at_3
value: 0.05531
- type: recall_at_5
value: 0.08029
- dataset:
type: mteb/trec-covid
name: MTEB TRECCOVID
config: default
split: test
task:
type: Retrieval
metrics:
- type: map_at_1
value: 0.00202
- type: map_at_10
value: 0.01727
- type: map_at_100
value: 0.10906
- type: map_at_1000
value: 0.2894
- type: map_at_3
value: 0.00553
- type: map_at_5
value: 0.00924
- type: mrr_at_1
value: 0.74
- type: mrr_at_10
value: 0.85667
- type: mrr_at_100
value: 0.85667
- type: mrr_at_1000
value: 0.85667
- type: mrr_at_3
value: 0.85667
- type: mrr_at_5
value: 0.85667
- type: ndcg_at_1
value: 0.66
- type: ndcg_at_10
value: 0.69259
- type: ndcg_at_100
value: 0.57274
- type: ndcg_at_1000
value: 0.55462
- type: ndcg_at_3
value: 0.70654
- type: ndcg_at_5
value: 0.71611
- type: precision_at_1
value: 0.74
- type: precision_at_10
value: 0.748
- type: precision_at_100
value: 0.5962
- type: precision_at_1000
value: 0.24842
- type: precision_at_3
value: 0.77333
- type: precision_at_5
value: 0.788
- type: recall_at_1
value: 0.00202
- type: recall_at_10
value: 0.02001
- type: recall_at_100
value: 0.14801
- type: recall_at_1000
value: 0.53939
- type: recall_at_3
value: 0.00609
- type: recall_at_5
value: 0.01048
pipeline_tag: sentence-similarity
Granite-Embedding-125m-English
Model Summary: Granite-Embedding-125m-English is a 125M parameter dense biencoder embedding model from the Granite Embeddings suite that can be used to generate high quality text embeddings. This model produces embedding vectors of size 768. Compared to most other open-source models, this model was only trained using open-source relevance-pair datasets with permissive, enterprise-friendly license, plus IBM collected and generated datasets. While maintaining competitive scores on academic benchmarks such as BEIR, this model also performs well on many enterprise use cases. This model is developed using retrieval oriented pretraining, contrastive finetuning and knowledge distillation.
- Developers: Granite Embedding Team, IBM
- GitHub Repository: ibm-granite/granite-embedding-models
- Website: Granite Docs
- Paper: Coming Soon
- Release Date: December 18th, 2024
- License: Apache 2.0
Supported Languages: English.
Intended use: The model is designed to produce fixed length vector representations for a given text, which can be used for text similarity, retrieval, and search applications.
Usage with Sentence Transformers: The model is compatible with SentenceTransformer library and is very easy to use:
First, install the sentence transformers library
pip install sentence_transformers
The model can then be used to encode pairs of text and find the similarity between their representations
from sentence_transformers import SentenceTransformer, util
model_path = "ibm-granite/granite-embedding-125m-english"
# Load the Sentence Transformer model
model = SentenceTransformer(model_path)
input_queries = [
' Who made the song My achy breaky heart? ',
'summit define'
]
input_passages = [
"Achy Breaky Heart is a country song written by Don Von Tress. Originally titled Don't Tell My Heart and performed by The Marcy Brothers in 1991. ",
"Definition of summit for English Language Learners. : 1 the highest point of a mountain : the top of a mountain. : 2 the highest level. : 3 a meeting or series of meetings between the leaders of two or more governments."
]
# encode queries and passages
query_embeddings = model.encode(input_queries)
passage_embeddings = model.encode(input_passages)
# calculate cosine similarity
print(util.cos_sim(query_embeddings, passage_embeddings))
Usage with Huggingface Transformers: This is a simple example of how to use the Granite-Embedding-125m-English model with the Transformers library and PyTorch.
First, install the required libraries
pip install transformers torch
The model can then be used to encode pairs of text
import torch
from transformers import AutoModel, AutoTokenizer
model_path = "ibm-granite/granite-embedding-125m-english"
# Load the model and tokenizer
model = AutoModel.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)
model.eval()
input_queries = [
' Who made the song My achy breaky heart? ',
'summit define'
]
# tokenize inputs
tokenized_queries = tokenizer(input_queries, padding=True, truncation=True, return_tensors='pt')
# encode queries
with torch.no_grad():
# Queries
model_output = model(**tokenized_queries)
# Perform pooling. granite-embedding-125m-english uses CLS Pooling
query_embeddings = model_output[0][:, 0]
# normalize the embeddings
query_embeddings = torch.nn.functional.normalize(query_embeddings, dim=1)
Evaluation:
The performance of the Granite-Embedding-125M-English model on MTEB Retrieval (i.e., BEIR) and code retrieval (CoIR) benchmarks is reported below.
| Model | Paramters (M) | Embedding Dimension | MTEB Retrieval (15) | CoIR (10) |
|---|---|---|---|---|
| granite-embedding-125m-english | 125 | 768 | 52.3 | 50.3 |
Model Architecture: Granite-Embedding-125m-English is based on an encoder-only RoBERTa like transformer architecture, trained internally at IBM Research.
| Model | granite-embedding-30m-english | granite-embedding-125m-english | granite-embedding-107m-multilingual | granite-embedding-278m-multilingual |
|---|---|---|---|---|
| Embedding size | 384 | 768 | 384 | 768 |
| Number of layers | 6 | 12 | 6 | 12 |
| Number of attention heads | 12 | 12 | 12 | 12 |
| Intermediate size | 1536 | 3072 | 1536 | 3072 |
| Activation Function | GeLU | GeLU | GeLU | GeLU |
| Vocabulary Size | 50265 | 50265 | 250002 | 250002 |
| Max. Sequence Length | 512 | 512 | 512 | 512 |
| # Parameters | 30M | 125M | 107M | 278M |
Training Data: Overall, the training data consists of four key sources: (1) unsupervised title-body paired data scraped from the web, (2) publicly available paired with permissive, enterprise-friendly license, (3) IBM-internal paired data targetting specific technical domains, and (4) IBM-generated synthetic data. The data is listed below:
| Dataset | Num. Pairs |
|---|---|
| SPECTER citation triplets | 684,100 |
| Stack Exchange Duplicate questions (titles) | 304,525 |
| Stack Exchange Duplicate questions (bodies) | 250,519 |
| Stack Exchange Duplicate questions (titles+bodies) | 250,460 |
| Natural Questions (NQ) | 100,231 |
| SQuAD2.0 | 87,599 |
| PAQ (Question, Answer) pairs | 64,371,441 |
| Stack Exchange (Title, Answer) pairs | 4,067,139 |
| Stack Exchange (Title, Body) pairs | 23,978,013 |
| Stack Exchange (Title+Body, Answer) pairs | 187,195 |
| S2ORC Citation pairs (Titles) | 52,603,982 |
| S2ORC (Title, Abstract) | 41,769,185 |
| S2ORC (Citations, abstracts) | 52,603,982 |
| WikiAnswers Duplicate question pairs | 77,427,422 |
| SearchQA | 582,261 |
| HotpotQA | 85,000 |
| Fever | 109,810 |
| Arxiv | 2,358,545 |
| Wikipedia | 20,745,403 |
| PubMed | 20,000,000 |
| Miracl En Pairs | 9,016 |
| DBPedia Title-Body Pairs | 4,635,922 |
| Synthetic: Query-Wikipedia Passage | 1,879,093 |
| Synthetic: Fact Verification | 9,888 |
| IBM Internal Triples | 40,290 |
| IBM Internal Title-Body Pairs | 1,524,586 |
Notably, we do not use the popular MS-MARCO retrieval dataset in our training corpus due to its non-commercial license, while other open-source models train on this dataset due to its high quality.
Infrastructure: We train Granite Embedding Models using IBM's computing cluster, Cognitive Compute Cluster, which is outfitted with NVIDIA A100 80gb GPUs. This cluster provides a scalable and efficient infrastructure for training our models over multiple GPUs.
Ethical Considerations and Limitations: The data used to train the base language model was filtered to remove text containing hate, abuse, and profanity. Granite-Embedding-125m-English is trained only for English texts, and has a context length of 512 tokens (longer texts will be truncated to this size).
Resources
- ⭐️ Learn about the latest updates with Granite: https://www.ibm.com/granite
- 📄 Get started with tutorials, best practices, and prompt engineering advice: https://www.ibm.com/granite/docs/
- 💡 Learn about the latest Granite learning resources: https://ibm.biz/granite-learning-resources