--- 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](https://github.com/ibm-granite/granite-embedding-models) - **Website**: [Granite Docs](https://www.ibm.com/granite/docs/) - **Paper:** Coming Soon - **Release Date**: December 18th, 2024 - **License:** [Apache 2.0](https://www.apache.org/licenses/LICENSE-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 ```shell pip install sentence_transformers ``` The model can then be used to encode pairs of text and find the similarity between their representations ```python 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 ```shell pip install transformers torch ``` The model can then be used to encode pairs of text ```python 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