--- tags: - mteb model-index: - name: zpoint_large_embedding_zh results: - task: type: STS dataset: type: C-MTEB/AFQMC name: MTEB AFQMC config: default split: validation revision: None metrics: - type: cos_sim_pearson value: 56.52479321107392 - type: cos_sim_spearman value: 60.72175935031135 - type: euclidean_pearson value: 59.40990657564856 - type: euclidean_spearman value: 60.72175934804556 - type: manhattan_pearson value: 59.4134322847349 - type: manhattan_spearman value: 60.724413114688225 - task: type: STS dataset: type: C-MTEB/ATEC name: MTEB ATEC config: default split: test revision: None metrics: - type: cos_sim_pearson value: 56.492631347325464 - type: cos_sim_spearman value: 58.765171687177656 - type: euclidean_pearson value: 63.236364373113844 - type: euclidean_spearman value: 58.765171686714865 - type: manhattan_pearson value: 63.22241814845751 - type: manhattan_spearman value: 58.762780342648234 - task: type: Classification dataset: type: mteb/amazon_reviews_multi name: MTEB AmazonReviewsClassification (zh) config: zh split: test revision: 1399c76144fd37290681b995c656ef9b2e06e26d metrics: - type: accuracy value: 49.72 - type: f1 value: 46.588683657317084 - task: type: STS dataset: type: C-MTEB/BQ name: MTEB BQ config: default split: test revision: None metrics: - type: cos_sim_pearson value: 73.07779128771674 - type: cos_sim_spearman value: 75.03682691328844 - type: euclidean_pearson value: 73.68098259699073 - type: euclidean_spearman value: 75.03683037648963 - type: manhattan_pearson value: 73.66963332679124 - type: manhattan_spearman value: 75.02269337817758 - task: type: Clustering dataset: type: C-MTEB/CLSClusteringP2P name: MTEB CLSClusteringP2P config: default split: test revision: None metrics: - type: v_measure value: 58.2897067752906 - task: type: Clustering dataset: type: C-MTEB/CLSClusteringS2S name: MTEB CLSClusteringS2S config: default split: test revision: None metrics: - type: v_measure value: 48.79170511177673 - task: type: Reranking dataset: type: C-MTEB/CMedQAv1 name: MTEB CMedQAv1 config: default split: test revision: None metrics: - type: map value: 91.10738371185181 - type: mrr value: 92.82496031746031 - task: type: Reranking dataset: type: C-MTEB/CMedQAv2 name: MTEB CMedQAv2 config: default split: test revision: None metrics: - type: map value: 90.06959035874831 - type: mrr value: 92.00789682539683 - task: type: Retrieval dataset: type: C-MTEB/CmedqaRetrieval name: MTEB CmedqaRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 27.132 - type: map_at_10 value: 40.400999999999996 - type: map_at_100 value: 42.246 - type: map_at_1000 value: 42.351 - type: map_at_3 value: 35.94 - type: map_at_5 value: 38.527 - type: mrr_at_1 value: 41.285 - type: mrr_at_10 value: 49.474000000000004 - type: mrr_at_100 value: 50.4 - type: mrr_at_1000 value: 50.438 - type: mrr_at_3 value: 46.891 - type: mrr_at_5 value: 48.353 - type: ndcg_at_1 value: 41.285 - type: ndcg_at_10 value: 47.159 - type: ndcg_at_100 value: 54.163 - type: ndcg_at_1000 value: 55.921 - type: ndcg_at_3 value: 41.678 - type: ndcg_at_5 value: 44.069 - type: precision_at_1 value: 41.285 - type: precision_at_10 value: 10.468 - type: precision_at_100 value: 1.611 - type: precision_at_1000 value: 0.183 - type: precision_at_3 value: 23.648 - type: precision_at_5 value: 17.229 - type: recall_at_1 value: 27.132 - type: recall_at_10 value: 57.977999999999994 - type: recall_at_100 value: 86.88 - type: recall_at_1000 value: 98.586 - type: recall_at_3 value: 41.487 - type: recall_at_5 value: 48.79 - task: type: PairClassification dataset: type: C-MTEB/CMNLI name: MTEB Cmnli config: default split: validation revision: None metrics: - type: cos_sim_accuracy value: 86.06133493686109 - type: cos_sim_ap value: 92.54288511740305 - type: cos_sim_f1 value: 86.85572811163628 - type: cos_sim_precision value: 83.72748969407681 - type: cos_sim_recall value: 90.22679448211363 - type: dot_accuracy value: 86.06133493686109 - type: dot_ap value: 92.53922591080917 - type: dot_f1 value: 86.85572811163628 - type: dot_precision value: 83.72748969407681 - type: dot_recall value: 90.22679448211363 - type: euclidean_accuracy value: 86.06133493686109 - type: euclidean_ap value: 92.54287994398305 - type: euclidean_f1 value: 86.85572811163628 - type: euclidean_precision value: 83.72748969407681 - type: euclidean_recall value: 90.22679448211363 - type: manhattan_accuracy value: 86.01322910402887 - type: manhattan_ap value: 92.53060255301997 - type: manhattan_f1 value: 86.81441683456458 - type: manhattan_precision value: 83.27249302125833 - type: manhattan_recall value: 90.67103109656301 - type: max_accuracy value: 86.06133493686109 - type: max_ap value: 92.54288511740305 - type: max_f1 value: 86.85572811163628 - task: type: Retrieval dataset: type: C-MTEB/CovidRetrieval name: MTEB CovidRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 78.899 - type: map_at_10 value: 86.232 - type: map_at_100 value: 86.331 - type: map_at_1000 value: 86.332 - type: map_at_3 value: 85.256 - type: map_at_5 value: 85.883 - type: mrr_at_1 value: 79.347 - type: mrr_at_10 value: 86.252 - type: mrr_at_100 value: 86.342 - type: mrr_at_1000 value: 86.343 - type: mrr_at_3 value: 85.283 - type: mrr_at_5 value: 85.91 - type: ndcg_at_1 value: 79.347 - type: ndcg_at_10 value: 89.143 - type: ndcg_at_100 value: 89.541 - type: ndcg_at_1000 value: 89.58 - type: ndcg_at_3 value: 87.227 - type: ndcg_at_5 value: 88.31400000000001 - type: precision_at_1 value: 79.347 - type: precision_at_10 value: 9.905 - type: precision_at_100 value: 1.0070000000000001 - type: precision_at_1000 value: 0.101 - type: precision_at_3 value: 31.261 - type: precision_at_5 value: 19.305 - type: recall_at_1 value: 78.899 - type: recall_at_10 value: 97.99799999999999 - type: recall_at_100 value: 99.684 - type: recall_at_1000 value: 100 - type: recall_at_3 value: 92.808 - type: recall_at_5 value: 95.46900000000001 - task: type: Retrieval dataset: type: C-MTEB/DuRetrieval name: MTEB DuRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 27.107999999999997 - type: map_at_10 value: 82.525 - type: map_at_100 value: 85.168 - type: map_at_1000 value: 85.194 - type: map_at_3 value: 57.74399999999999 - type: map_at_5 value: 72.53699999999999 - type: mrr_at_1 value: 92.30000000000001 - type: mrr_at_10 value: 94.705 - type: mrr_at_100 value: 94.76599999999999 - type: mrr_at_1000 value: 94.76599999999999 - type: mrr_at_3 value: 94.55 - type: mrr_at_5 value: 94.64 - type: ndcg_at_1 value: 92.30000000000001 - type: ndcg_at_10 value: 89.23100000000001 - type: ndcg_at_100 value: 91.556 - type: ndcg_at_1000 value: 91.81700000000001 - type: ndcg_at_3 value: 88.558 - type: ndcg_at_5 value: 87.316 - type: precision_at_1 value: 92.30000000000001 - type: precision_at_10 value: 42.38 - type: precision_at_100 value: 4.818 - type: precision_at_1000 value: 0.488 - type: precision_at_3 value: 79.14999999999999 - type: precision_at_5 value: 66.63 - type: recall_at_1 value: 27.107999999999997 - type: recall_at_10 value: 89.914 - type: recall_at_100 value: 97.658 - type: recall_at_1000 value: 99.00099999999999 - type: recall_at_3 value: 59.673 - type: recall_at_5 value: 76.437 - task: type: Retrieval dataset: type: C-MTEB/EcomRetrieval name: MTEB EcomRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 55.00000000000001 - type: map_at_10 value: 65.57600000000001 - type: map_at_100 value: 66.096 - type: map_at_1000 value: 66.103 - type: map_at_3 value: 63.217 - type: map_at_5 value: 64.562 - type: mrr_at_1 value: 55.00000000000001 - type: mrr_at_10 value: 65.57600000000001 - type: mrr_at_100 value: 66.096 - type: mrr_at_1000 value: 66.103 - type: mrr_at_3 value: 63.217 - type: mrr_at_5 value: 64.562 - type: ndcg_at_1 value: 55.00000000000001 - type: ndcg_at_10 value: 70.74000000000001 - type: ndcg_at_100 value: 73.001 - type: ndcg_at_1000 value: 73.223 - type: ndcg_at_3 value: 65.837 - type: ndcg_at_5 value: 68.264 - type: precision_at_1 value: 55.00000000000001 - type: precision_at_10 value: 8.7 - type: precision_at_100 value: 0.97 - type: precision_at_1000 value: 0.099 - type: precision_at_3 value: 24.467 - type: precision_at_5 value: 15.86 - type: recall_at_1 value: 55.00000000000001 - type: recall_at_10 value: 87 - type: recall_at_100 value: 97 - type: recall_at_1000 value: 98.8 - type: recall_at_3 value: 73.4 - type: recall_at_5 value: 79.3 - task: type: Classification dataset: type: C-MTEB/IFlyTek-classification name: MTEB IFlyTek config: default split: validation revision: None metrics: - type: accuracy value: 51.696806464024625 - type: f1 value: 40.02655259854763 - task: type: Classification dataset: type: C-MTEB/JDReview-classification name: MTEB JDReview config: default split: test revision: None metrics: - type: accuracy value: 88.87429643527206 - type: ap value: 59.89821610336161 - type: f1 value: 83.98100504939507 - task: type: STS dataset: type: C-MTEB/LCQMC name: MTEB LCQMC config: default split: test revision: None metrics: - type: cos_sim_pearson value: 72.59510783330644 - type: cos_sim_spearman value: 79.75022839599451 - type: euclidean_pearson value: 79.54475341768782 - type: euclidean_spearman value: 79.75021730266204 - type: manhattan_pearson value: 79.53741020350834 - type: manhattan_spearman value: 79.74152434784455 - task: type: Reranking dataset: type: C-MTEB/Mmarco-reranking name: MTEB MMarcoReranking config: default split: dev revision: None metrics: - type: map value: 38.86925357762224 - type: mrr value: 38.17460317460318 - task: type: Retrieval dataset: type: C-MTEB/MMarcoRetrieval name: MTEB MMarcoRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 68.731 - type: map_at_10 value: 78.52 - type: map_at_100 value: 78.792 - type: map_at_1000 value: 78.797 - type: map_at_3 value: 76.586 - type: map_at_5 value: 77.876 - type: mrr_at_1 value: 71.003 - type: mrr_at_10 value: 79.03 - type: mrr_at_100 value: 79.27 - type: mrr_at_1000 value: 79.274 - type: mrr_at_3 value: 77.373 - type: mrr_at_5 value: 78.46600000000001 - type: ndcg_at_1 value: 71.003 - type: ndcg_at_10 value: 82.381 - type: ndcg_at_100 value: 83.504 - type: ndcg_at_1000 value: 83.627 - type: ndcg_at_3 value: 78.78699999999999 - type: ndcg_at_5 value: 80.94 - type: precision_at_1 value: 71.003 - type: precision_at_10 value: 9.961 - type: precision_at_100 value: 1.05 - type: precision_at_1000 value: 0.106 - type: precision_at_3 value: 29.694 - type: precision_at_5 value: 18.963 - type: recall_at_1 value: 68.731 - type: recall_at_10 value: 93.697 - type: recall_at_100 value: 98.546 - type: recall_at_1000 value: 99.515 - type: recall_at_3 value: 84.328 - type: recall_at_5 value: 89.42 - task: type: Classification dataset: type: mteb/amazon_massive_intent name: MTEB MassiveIntentClassification (zh-CN) config: zh-CN split: test revision: 31efe3c427b0bae9c22cbb560b8f15491cc6bed7 metrics: - type: accuracy value: 76.79219905850707 - type: f1 value: 73.15228001501512 - task: type: Classification dataset: type: mteb/amazon_massive_scenario name: MTEB MassiveScenarioClassification (zh-CN) config: zh-CN split: test revision: 7d571f92784cd94a019292a1f45445077d0ef634 metrics: - type: accuracy value: 84.9562878278413 - type: f1 value: 84.0910677219451 - task: type: Retrieval dataset: type: C-MTEB/MedicalRetrieval name: MTEB MedicalRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 57.8 - type: map_at_10 value: 64.732 - type: map_at_100 value: 65.315 - type: map_at_1000 value: 65.347 - type: map_at_3 value: 63.14999999999999 - type: map_at_5 value: 63.934999999999995 - type: mrr_at_1 value: 57.99999999999999 - type: mrr_at_10 value: 64.852 - type: mrr_at_100 value: 65.435 - type: mrr_at_1000 value: 65.467 - type: mrr_at_3 value: 63.266999999999996 - type: mrr_at_5 value: 64.072 - type: ndcg_at_1 value: 57.8 - type: ndcg_at_10 value: 68.14 - type: ndcg_at_100 value: 71.04899999999999 - type: ndcg_at_1000 value: 71.856 - type: ndcg_at_3 value: 64.813 - type: ndcg_at_5 value: 66.241 - type: precision_at_1 value: 57.8 - type: precision_at_10 value: 7.89 - type: precision_at_100 value: 0.927 - type: precision_at_1000 value: 0.099 - type: precision_at_3 value: 23.200000000000003 - type: precision_at_5 value: 14.62 - type: recall_at_1 value: 57.8 - type: recall_at_10 value: 78.9 - type: recall_at_100 value: 92.7 - type: recall_at_1000 value: 99 - type: recall_at_3 value: 69.6 - type: recall_at_5 value: 73.1 - task: type: Classification dataset: type: C-MTEB/MultilingualSentiment-classification name: MTEB MultilingualSentiment config: default split: validation revision: None metrics: - type: accuracy value: 79.22333333333333 - type: f1 value: 79.01276765455862 - task: type: PairClassification dataset: type: C-MTEB/OCNLI name: MTEB Ocnli config: default split: validation revision: None metrics: - type: cos_sim_accuracy value: 85.32755820249052 - type: cos_sim_ap value: 90.56118966152913 - type: cos_sim_f1 value: 86.28428927680798 - type: cos_sim_precision value: 81.75803402646503 - type: cos_sim_recall value: 91.34107708553326 - type: dot_accuracy value: 85.32755820249052 - type: dot_ap value: 90.56120405888693 - type: dot_f1 value: 86.28428927680798 - type: dot_precision value: 81.75803402646503 - type: dot_recall value: 91.34107708553326 - type: euclidean_accuracy value: 85.32755820249052 - type: euclidean_ap value: 90.56118966152913 - type: euclidean_f1 value: 86.28428927680798 - type: euclidean_precision value: 81.75803402646503 - type: euclidean_recall value: 91.34107708553326 - type: manhattan_accuracy value: 85.43584190579317 - type: manhattan_ap value: 90.52296007826511 - type: manhattan_f1 value: 86.42099949520444 - type: manhattan_precision value: 82.7852998065764 - type: manhattan_recall value: 90.3907074973601 - type: max_accuracy value: 85.43584190579317 - type: max_ap value: 90.56120405888693 - type: max_f1 value: 86.42099949520444 - task: type: Classification dataset: type: C-MTEB/OnlineShopping-classification name: MTEB OnlineShopping config: default split: test revision: None metrics: - type: accuracy value: 94.87999999999998 - type: ap value: 93.12892276945414 - type: f1 value: 94.86921245385685 - task: type: STS dataset: type: C-MTEB/PAWSX name: MTEB PAWSX config: default split: test revision: None metrics: - type: cos_sim_pearson value: 38.4367277229591 - type: cos_sim_spearman value: 45.942712312151656 - type: euclidean_pearson value: 44.96055989566686 - type: euclidean_spearman value: 45.94279939044163 - type: manhattan_pearson value: 44.979762134562925 - type: manhattan_spearman value: 45.96004430328375 - task: type: STS dataset: type: C-MTEB/QBQTC name: MTEB QBQTC config: default split: test revision: None metrics: - type: cos_sim_pearson value: 41.45428416733968 - type: cos_sim_spearman value: 43.462057455255845 - type: euclidean_pearson value: 38.20089604291246 - type: euclidean_spearman value: 43.46288438624811 - type: manhattan_pearson value: 38.175045608320694 - type: manhattan_spearman value: 43.468885824666344 - task: type: STS dataset: type: mteb/sts22-crosslingual-sts name: MTEB STS22 (zh) config: zh split: test revision: 6d1ba47164174a496b7fa5d3569dae26a6813b80 metrics: - type: cos_sim_pearson value: 65.61911213187778 - type: cos_sim_spearman value: 66.70525921118497 - type: euclidean_pearson value: 65.35554462551515 - type: euclidean_spearman value: 66.70525921118497 - type: manhattan_pearson value: 65.25174169329627 - type: manhattan_spearman value: 66.6550752269368 - task: type: STS dataset: type: C-MTEB/STSB name: MTEB STSB config: default split: test revision: None metrics: - type: cos_sim_pearson value: 81.27160581568329 - type: cos_sim_spearman value: 83.34482829304406 - type: euclidean_pearson value: 82.98079434913451 - type: euclidean_spearman value: 83.34503180775212 - type: manhattan_pearson value: 82.95256917013506 - type: manhattan_spearman value: 83.31034894907503 - task: type: Reranking dataset: type: C-MTEB/T2Reranking name: MTEB T2Reranking config: default split: dev revision: None metrics: - type: map value: 69.29054152015013 - type: mrr value: 79.73472208788729 - task: type: Retrieval dataset: type: C-MTEB/T2Retrieval name: MTEB T2Retrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 27 - type: map_at_10 value: 75.871 - type: map_at_100 value: 79.664 - type: map_at_1000 value: 79.725 - type: map_at_3 value: 53.14 - type: map_at_5 value: 65.365 - type: mrr_at_1 value: 88.642 - type: mrr_at_10 value: 91.732 - type: mrr_at_100 value: 91.818 - type: mrr_at_1000 value: 91.821 - type: mrr_at_3 value: 91.217 - type: mrr_at_5 value: 91.561 - type: ndcg_at_1 value: 88.642 - type: ndcg_at_10 value: 83.815 - type: ndcg_at_100 value: 87.689 - type: ndcg_at_1000 value: 88.266 - type: ndcg_at_3 value: 84.807 - type: ndcg_at_5 value: 83.53699999999999 - type: precision_at_1 value: 88.642 - type: precision_at_10 value: 41.725 - type: precision_at_100 value: 5.024 - type: precision_at_1000 value: 0.516 - type: precision_at_3 value: 74.10600000000001 - type: precision_at_5 value: 62.192 - type: recall_at_1 value: 27 - type: recall_at_10 value: 83.292 - type: recall_at_100 value: 95.66799999999999 - type: recall_at_1000 value: 98.56 - type: recall_at_3 value: 55.111 - type: recall_at_5 value: 69.327 - task: type: Classification dataset: type: C-MTEB/TNews-classification name: MTEB TNews config: default split: validation revision: None metrics: - type: accuracy value: 54.346 - type: f1 value: 52.302508458396055 - task: type: Clustering dataset: type: C-MTEB/ThuNewsClusteringP2P name: MTEB ThuNewsClusteringP2P config: default split: test revision: None metrics: - type: v_measure value: 72.47709523787981 - task: type: Clustering dataset: type: C-MTEB/ThuNewsClusteringS2S name: MTEB ThuNewsClusteringS2S config: default split: test revision: None metrics: - type: v_measure value: 69.35293863978707 - task: type: Retrieval dataset: type: C-MTEB/VideoRetrieval name: MTEB VideoRetrieval config: default split: dev revision: None metrics: - type: map_at_1 value: 64.60000000000001 - type: map_at_10 value: 75.683 - type: map_at_100 value: 75.961 - type: map_at_1000 value: 75.96199999999999 - type: map_at_3 value: 74.083 - type: map_at_5 value: 75.03800000000001 - type: mrr_at_1 value: 64.60000000000001 - type: mrr_at_10 value: 75.683 - type: mrr_at_100 value: 75.961 - type: mrr_at_1000 value: 75.96199999999999 - type: mrr_at_3 value: 74.083 - type: mrr_at_5 value: 75.03800000000001 - type: ndcg_at_1 value: 64.60000000000001 - type: ndcg_at_10 value: 80.26299999999999 - type: ndcg_at_100 value: 81.487 - type: ndcg_at_1000 value: 81.5 - type: ndcg_at_3 value: 77.003 - type: ndcg_at_5 value: 78.708 - type: precision_at_1 value: 64.60000000000001 - type: precision_at_10 value: 9.43 - type: precision_at_100 value: 0.997 - type: precision_at_1000 value: 0.1 - type: precision_at_3 value: 28.467 - type: precision_at_5 value: 17.9 - type: recall_at_1 value: 64.60000000000001 - type: recall_at_10 value: 94.3 - type: recall_at_100 value: 99.7 - type: recall_at_1000 value: 99.8 - type: recall_at_3 value: 85.39999999999999 - type: recall_at_5 value: 89.5 - task: type: Classification dataset: type: C-MTEB/waimai-classification name: MTEB Waimai config: default split: test revision: None metrics: - type: accuracy value: 89.36 - type: ap value: 75.26507519569006 - type: f1 value: 87.89845508858562 language: - zh license: mit library_name: sentence-transformers ---

ZPoint Large Embedding for Chinese

- **[2024-06-04]** Release zpoint_large_embedding_zh, and upload model weight to huggingface - **[2024-06-05]** Add training details ### Training Details **Base Model** 1) We chose [Stella](https://huggingface.co/infgrad/stella-mrl-large-zh-v3.5-1792d) as our base model. **Training Data** 1) **Hard negative samping** - For retrieval task, We sampled 10 hard negative passages/answers from top50-top200 related passages/answers for each query. - For classification/clustering tasks, we sampled 5 hard negative samples from other classes/cluster for each sample. - For classification/clustering tasks, we also used the category names of each class and cluster as positive and negative samples. 2) **Data synthesis by LLM (ZPoint-72B)** - For retrieval tasks, we used LLM to rewrite each query, generating five different rewritten results. - For retrieval tasks, we also generated five new queries for some documents by LLM. - For non-retrieval tasks, we used LLM to rewrite the queries, generating five rewritten results for each query. - Finally, total amount of synthesized data is about 30 million. 3) **Collect more data for retrieval-type tasks** - [miracl/miracl](https://huggingface.co/datasets/miracl/miracl) - [FreedomIntelligence/Huatuo26M-Lite](https://huggingface.co/datasets/FreedomIntelligence/Huatuo26M-Lite) - [PaddlePaddle/dureader_robust](https://huggingface.co/datasets/PaddlePaddle/dureader_robust) **C-MTEB test filtered** - [THUIR/T2Ranking](https://huggingface.co/datasets/THUIR/T2Ranking) **C-MTEB test filtered** - [Shitao/bge-reranker-data](https://huggingface.co/datasets/Shitao/bge-reranker-data) - [Shitao/MLDR](https://huggingface.co/datasets/Shitao/MLDR) - ... ***We constructed a dataset of approximately 100 million training samples through collection, machine translation, and LLM synthesis. This dataset includes data from various fields such as healthcare, law, electricity, automotive, and 3C (Consumer Electronics).*** **Training loss** 1) Multi-Task loss like [Piccolo](https://huggingface.co/sensenova/piccolo-large-zh-v2) 2) Matryoshka Representation Learning ### Example ```python from sentence_transformers import SentenceTransformer sentences1 = ["这个产品真垃圾"] sentences2 = ["我太喜欢这个产品了"] model = SentenceTransformer('iampanda/zpoint_large_embedding_zh') embeddings_1 = model.encode(sentences1, normalize_embeddings=True) embeddings_2 = model.encode(sentences2, normalize_embeddings=True) similarity = embeddings_1 @ embeddings_2.T print(similarity) ```