stella-base-en-v2 / README.md
infgrad's picture
Update README.md
c9e80ff verified
metadata
tags:
  - sentence-transformers
  - feature-extraction
  - sentence-similarity
  - mteb
model-index:
  - name: stella-base-en-v2
    results:
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_counterfactual
          name: MTEB AmazonCounterfactualClassification (en)
          config: en
          split: test
          revision: e8379541af4e31359cca9fbcf4b00f2671dba205
        metrics:
          - type: accuracy
            value: 77.19402985074628
          - type: ap
            value: 40.43267503017359
          - type: f1
            value: 71.15585210518594
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_polarity
          name: MTEB AmazonPolarityClassification
          config: default
          split: test
          revision: e2d317d38cd51312af73b3d32a06d1a08b442046
        metrics:
          - type: accuracy
            value: 93.256675
          - type: ap
            value: 90.00824833079179
          - type: f1
            value: 93.2473146151734
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_reviews_multi
          name: MTEB AmazonReviewsClassification (en)
          config: en
          split: test
          revision: 1399c76144fd37290681b995c656ef9b2e06e26d
        metrics:
          - type: accuracy
            value: 49.612
          - type: f1
            value: 48.530785631574304
      - task:
          type: Retrieval
        dataset:
          type: arguana
          name: MTEB ArguAna
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 37.411
          - type: map_at_10
            value: 52.673
          - type: map_at_100
            value: 53.410999999999994
          - type: map_at_1000
            value: 53.415
          - type: map_at_3
            value: 48.495
          - type: map_at_5
            value: 51.183
          - type: mrr_at_1
            value: 37.838
          - type: mrr_at_10
            value: 52.844
          - type: mrr_at_100
            value: 53.581999999999994
          - type: mrr_at_1000
            value: 53.586
          - type: mrr_at_3
            value: 48.672
          - type: mrr_at_5
            value: 51.272
          - type: ndcg_at_1
            value: 37.411
          - type: ndcg_at_10
            value: 60.626999999999995
          - type: ndcg_at_100
            value: 63.675000000000004
          - type: ndcg_at_1000
            value: 63.776999999999994
          - type: ndcg_at_3
            value: 52.148
          - type: ndcg_at_5
            value: 57.001999999999995
          - type: precision_at_1
            value: 37.411
          - type: precision_at_10
            value: 8.578
          - type: precision_at_100
            value: 0.989
          - type: precision_at_1000
            value: 0.1
          - type: precision_at_3
            value: 20.91
          - type: precision_at_5
            value: 14.908
          - type: recall_at_1
            value: 37.411
          - type: recall_at_10
            value: 85.775
          - type: recall_at_100
            value: 98.86200000000001
          - type: recall_at_1000
            value: 99.644
          - type: recall_at_3
            value: 62.731
          - type: recall_at_5
            value: 74.53800000000001
      - task:
          type: Clustering
        dataset:
          type: mteb/arxiv-clustering-p2p
          name: MTEB ArxivClusteringP2P
          config: default
          split: test
          revision: a122ad7f3f0291bf49cc6f4d32aa80929df69d5d
        metrics:
          - type: v_measure
            value: 47.24219029437865
      - task:
          type: Clustering
        dataset:
          type: mteb/arxiv-clustering-s2s
          name: MTEB ArxivClusteringS2S
          config: default
          split: test
          revision: f910caf1a6075f7329cdf8c1a6135696f37dbd53
        metrics:
          - type: v_measure
            value: 40.474604844291726
      - task:
          type: Reranking
        dataset:
          type: mteb/askubuntudupquestions-reranking
          name: MTEB AskUbuntuDupQuestions
          config: default
          split: test
          revision: 2000358ca161889fa9c082cb41daa8dcfb161a54
        metrics:
          - type: map
            value: 62.720542706366054
          - type: mrr
            value: 75.59633733456448
      - task:
          type: STS
        dataset:
          type: mteb/biosses-sts
          name: MTEB BIOSSES
          config: default
          split: test
          revision: d3fb88f8f02e40887cd149695127462bbcf29b4a
        metrics:
          - type: cos_sim_pearson
            value: 86.31345008397868
          - type: cos_sim_spearman
            value: 85.94292212320399
          - type: euclidean_pearson
            value: 85.03974302774525
          - type: euclidean_spearman
            value: 85.88087251659051
          - type: manhattan_pearson
            value: 84.91900996712951
          - type: manhattan_spearman
            value: 85.96701905781116
      - task:
          type: Classification
        dataset:
          type: mteb/banking77
          name: MTEB Banking77Classification
          config: default
          split: test
          revision: 0fd18e25b25c072e09e0d92ab615fda904d66300
        metrics:
          - type: accuracy
            value: 84.72727272727273
          - type: f1
            value: 84.29572512364581
      - task:
          type: Clustering
        dataset:
          type: mteb/biorxiv-clustering-p2p
          name: MTEB BiorxivClusteringP2P
          config: default
          split: test
          revision: 65b79d1d13f80053f67aca9498d9402c2d9f1f40
        metrics:
          - type: v_measure
            value: 39.55532460397536
      - task:
          type: Clustering
        dataset:
          type: mteb/biorxiv-clustering-s2s
          name: MTEB BiorxivClusteringS2S
          config: default
          split: test
          revision: 258694dd0231531bc1fd9de6ceb52a0853c6d908
        metrics:
          - type: v_measure
            value: 35.91195973591251
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackAndroidRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 32.822
          - type: map_at_10
            value: 44.139
          - type: map_at_100
            value: 45.786
          - type: map_at_1000
            value: 45.906000000000006
          - type: map_at_3
            value: 40.637
          - type: map_at_5
            value: 42.575
          - type: mrr_at_1
            value: 41.059
          - type: mrr_at_10
            value: 50.751000000000005
          - type: mrr_at_100
            value: 51.548
          - type: mrr_at_1000
            value: 51.583999999999996
          - type: mrr_at_3
            value: 48.236000000000004
          - type: mrr_at_5
            value: 49.838
          - type: ndcg_at_1
            value: 41.059
          - type: ndcg_at_10
            value: 50.573
          - type: ndcg_at_100
            value: 56.25
          - type: ndcg_at_1000
            value: 58.004
          - type: ndcg_at_3
            value: 45.995000000000005
          - type: ndcg_at_5
            value: 48.18
          - type: precision_at_1
            value: 41.059
          - type: precision_at_10
            value: 9.757
          - type: precision_at_100
            value: 1.609
          - type: precision_at_1000
            value: 0.20600000000000002
          - type: precision_at_3
            value: 22.222
          - type: precision_at_5
            value: 16.023
          - type: recall_at_1
            value: 32.822
          - type: recall_at_10
            value: 61.794000000000004
          - type: recall_at_100
            value: 85.64699999999999
          - type: recall_at_1000
            value: 96.836
          - type: recall_at_3
            value: 47.999
          - type: recall_at_5
            value: 54.376999999999995
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackEnglishRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 29.579
          - type: map_at_10
            value: 39.787
          - type: map_at_100
            value: 40.976
          - type: map_at_1000
            value: 41.108
          - type: map_at_3
            value: 36.819
          - type: map_at_5
            value: 38.437
          - type: mrr_at_1
            value: 37.516
          - type: mrr_at_10
            value: 45.822
          - type: mrr_at_100
            value: 46.454
          - type: mrr_at_1000
            value: 46.495999999999995
          - type: mrr_at_3
            value: 43.556
          - type: mrr_at_5
            value: 44.814
          - type: ndcg_at_1
            value: 37.516
          - type: ndcg_at_10
            value: 45.5
          - type: ndcg_at_100
            value: 49.707
          - type: ndcg_at_1000
            value: 51.842
          - type: ndcg_at_3
            value: 41.369
          - type: ndcg_at_5
            value: 43.161
          - type: precision_at_1
            value: 37.516
          - type: precision_at_10
            value: 8.713
          - type: precision_at_100
            value: 1.38
          - type: precision_at_1000
            value: 0.188
          - type: precision_at_3
            value: 20.233999999999998
          - type: precision_at_5
            value: 14.280000000000001
          - type: recall_at_1
            value: 29.579
          - type: recall_at_10
            value: 55.458
          - type: recall_at_100
            value: 73.49799999999999
          - type: recall_at_1000
            value: 87.08200000000001
          - type: recall_at_3
            value: 42.858000000000004
          - type: recall_at_5
            value: 48.215
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackGamingRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 40.489999999999995
          - type: map_at_10
            value: 53.313
          - type: map_at_100
            value: 54.290000000000006
          - type: map_at_1000
            value: 54.346000000000004
          - type: map_at_3
            value: 49.983
          - type: map_at_5
            value: 51.867
          - type: mrr_at_1
            value: 46.27
          - type: mrr_at_10
            value: 56.660999999999994
          - type: mrr_at_100
            value: 57.274
          - type: mrr_at_1000
            value: 57.301
          - type: mrr_at_3
            value: 54.138
          - type: mrr_at_5
            value: 55.623999999999995
          - type: ndcg_at_1
            value: 46.27
          - type: ndcg_at_10
            value: 59.192
          - type: ndcg_at_100
            value: 63.026
          - type: ndcg_at_1000
            value: 64.079
          - type: ndcg_at_3
            value: 53.656000000000006
          - type: ndcg_at_5
            value: 56.387
          - type: precision_at_1
            value: 46.27
          - type: precision_at_10
            value: 9.511
          - type: precision_at_100
            value: 1.23
          - type: precision_at_1000
            value: 0.136
          - type: precision_at_3
            value: 24.096
          - type: precision_at_5
            value: 16.476
          - type: recall_at_1
            value: 40.489999999999995
          - type: recall_at_10
            value: 73.148
          - type: recall_at_100
            value: 89.723
          - type: recall_at_1000
            value: 97.073
          - type: recall_at_3
            value: 58.363
          - type: recall_at_5
            value: 65.083
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackGisRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 26.197
          - type: map_at_10
            value: 35.135
          - type: map_at_100
            value: 36.14
          - type: map_at_1000
            value: 36.216
          - type: map_at_3
            value: 32.358
          - type: map_at_5
            value: 33.814
          - type: mrr_at_1
            value: 28.475
          - type: mrr_at_10
            value: 37.096000000000004
          - type: mrr_at_100
            value: 38.006
          - type: mrr_at_1000
            value: 38.06
          - type: mrr_at_3
            value: 34.52
          - type: mrr_at_5
            value: 35.994
          - type: ndcg_at_1
            value: 28.475
          - type: ndcg_at_10
            value: 40.263
          - type: ndcg_at_100
            value: 45.327
          - type: ndcg_at_1000
            value: 47.225
          - type: ndcg_at_3
            value: 34.882000000000005
          - type: ndcg_at_5
            value: 37.347
          - type: precision_at_1
            value: 28.475
          - type: precision_at_10
            value: 6.249
          - type: precision_at_100
            value: 0.919
          - type: precision_at_1000
            value: 0.11199999999999999
          - type: precision_at_3
            value: 14.689
          - type: precision_at_5
            value: 10.237
          - type: recall_at_1
            value: 26.197
          - type: recall_at_10
            value: 54.17999999999999
          - type: recall_at_100
            value: 77.768
          - type: recall_at_1000
            value: 91.932
          - type: recall_at_3
            value: 39.804
          - type: recall_at_5
            value: 45.660000000000004
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackMathematicaRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 16.683
          - type: map_at_10
            value: 25.013999999999996
          - type: map_at_100
            value: 26.411
          - type: map_at_1000
            value: 26.531
          - type: map_at_3
            value: 22.357
          - type: map_at_5
            value: 23.982999999999997
          - type: mrr_at_1
            value: 20.896
          - type: mrr_at_10
            value: 29.758000000000003
          - type: mrr_at_100
            value: 30.895
          - type: mrr_at_1000
            value: 30.964999999999996
          - type: mrr_at_3
            value: 27.177
          - type: mrr_at_5
            value: 28.799999999999997
          - type: ndcg_at_1
            value: 20.896
          - type: ndcg_at_10
            value: 30.294999999999998
          - type: ndcg_at_100
            value: 36.68
          - type: ndcg_at_1000
            value: 39.519
          - type: ndcg_at_3
            value: 25.480999999999998
          - type: ndcg_at_5
            value: 28.027
          - type: precision_at_1
            value: 20.896
          - type: precision_at_10
            value: 5.56
          - type: precision_at_100
            value: 1.006
          - type: precision_at_1000
            value: 0.13899999999999998
          - type: precision_at_3
            value: 12.231
          - type: precision_at_5
            value: 9.104
          - type: recall_at_1
            value: 16.683
          - type: recall_at_10
            value: 41.807
          - type: recall_at_100
            value: 69.219
          - type: recall_at_1000
            value: 89.178
          - type: recall_at_3
            value: 28.772
          - type: recall_at_5
            value: 35.167
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackPhysicsRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 30.653000000000002
          - type: map_at_10
            value: 41.21
          - type: map_at_100
            value: 42.543
          - type: map_at_1000
            value: 42.657000000000004
          - type: map_at_3
            value: 38.094
          - type: map_at_5
            value: 39.966
          - type: mrr_at_1
            value: 37.824999999999996
          - type: mrr_at_10
            value: 47.087
          - type: mrr_at_100
            value: 47.959
          - type: mrr_at_1000
            value: 48.003
          - type: mrr_at_3
            value: 45.043
          - type: mrr_at_5
            value: 46.352
          - type: ndcg_at_1
            value: 37.824999999999996
          - type: ndcg_at_10
            value: 47.158
          - type: ndcg_at_100
            value: 52.65
          - type: ndcg_at_1000
            value: 54.644999999999996
          - type: ndcg_at_3
            value: 42.632999999999996
          - type: ndcg_at_5
            value: 44.994
          - type: precision_at_1
            value: 37.824999999999996
          - type: precision_at_10
            value: 8.498999999999999
          - type: precision_at_100
            value: 1.308
          - type: precision_at_1000
            value: 0.166
          - type: precision_at_3
            value: 20.308
          - type: precision_at_5
            value: 14.283000000000001
          - type: recall_at_1
            value: 30.653000000000002
          - type: recall_at_10
            value: 58.826
          - type: recall_at_100
            value: 81.94
          - type: recall_at_1000
            value: 94.71000000000001
          - type: recall_at_3
            value: 45.965
          - type: recall_at_5
            value: 52.294
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackProgrammersRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 26.71
          - type: map_at_10
            value: 36.001
          - type: map_at_100
            value: 37.416
          - type: map_at_1000
            value: 37.522
          - type: map_at_3
            value: 32.841
          - type: map_at_5
            value: 34.515
          - type: mrr_at_1
            value: 32.647999999999996
          - type: mrr_at_10
            value: 41.43
          - type: mrr_at_100
            value: 42.433
          - type: mrr_at_1000
            value: 42.482
          - type: mrr_at_3
            value: 39.117000000000004
          - type: mrr_at_5
            value: 40.35
          - type: ndcg_at_1
            value: 32.647999999999996
          - type: ndcg_at_10
            value: 41.629
          - type: ndcg_at_100
            value: 47.707
          - type: ndcg_at_1000
            value: 49.913000000000004
          - type: ndcg_at_3
            value: 36.598000000000006
          - type: ndcg_at_5
            value: 38.696000000000005
          - type: precision_at_1
            value: 32.647999999999996
          - type: precision_at_10
            value: 7.704999999999999
          - type: precision_at_100
            value: 1.242
          - type: precision_at_1000
            value: 0.16
          - type: precision_at_3
            value: 17.314
          - type: precision_at_5
            value: 12.374
          - type: recall_at_1
            value: 26.71
          - type: recall_at_10
            value: 52.898
          - type: recall_at_100
            value: 79.08
          - type: recall_at_1000
            value: 93.94
          - type: recall_at_3
            value: 38.731
          - type: recall_at_5
            value: 44.433
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 26.510999999999996
          - type: map_at_10
            value: 35.755333333333326
          - type: map_at_100
            value: 36.97525
          - type: map_at_1000
            value: 37.08741666666667
          - type: map_at_3
            value: 32.921
          - type: map_at_5
            value: 34.45041666666667
          - type: mrr_at_1
            value: 31.578416666666666
          - type: mrr_at_10
            value: 40.06066666666667
          - type: mrr_at_100
            value: 40.93350000000001
          - type: mrr_at_1000
            value: 40.98716666666667
          - type: mrr_at_3
            value: 37.710499999999996
          - type: mrr_at_5
            value: 39.033249999999995
          - type: ndcg_at_1
            value: 31.578416666666666
          - type: ndcg_at_10
            value: 41.138666666666666
          - type: ndcg_at_100
            value: 46.37291666666666
          - type: ndcg_at_1000
            value: 48.587500000000006
          - type: ndcg_at_3
            value: 36.397083333333335
          - type: ndcg_at_5
            value: 38.539
          - type: precision_at_1
            value: 31.578416666666666
          - type: precision_at_10
            value: 7.221583333333332
          - type: precision_at_100
            value: 1.1581666666666668
          - type: precision_at_1000
            value: 0.15416666666666667
          - type: precision_at_3
            value: 16.758
          - type: precision_at_5
            value: 11.830916666666665
          - type: recall_at_1
            value: 26.510999999999996
          - type: recall_at_10
            value: 52.7825
          - type: recall_at_100
            value: 75.79675
          - type: recall_at_1000
            value: 91.10483333333335
          - type: recall_at_3
            value: 39.48233333333334
          - type: recall_at_5
            value: 45.07116666666667
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackStatsRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 24.564
          - type: map_at_10
            value: 31.235000000000003
          - type: map_at_100
            value: 32.124
          - type: map_at_1000
            value: 32.216
          - type: map_at_3
            value: 29.330000000000002
          - type: map_at_5
            value: 30.379
          - type: mrr_at_1
            value: 27.761000000000003
          - type: mrr_at_10
            value: 34.093
          - type: mrr_at_100
            value: 34.885
          - type: mrr_at_1000
            value: 34.957
          - type: mrr_at_3
            value: 32.388
          - type: mrr_at_5
            value: 33.269
          - type: ndcg_at_1
            value: 27.761000000000003
          - type: ndcg_at_10
            value: 35.146
          - type: ndcg_at_100
            value: 39.597
          - type: ndcg_at_1000
            value: 42.163000000000004
          - type: ndcg_at_3
            value: 31.674000000000003
          - type: ndcg_at_5
            value: 33.224
          - type: precision_at_1
            value: 27.761000000000003
          - type: precision_at_10
            value: 5.383
          - type: precision_at_100
            value: 0.836
          - type: precision_at_1000
            value: 0.11199999999999999
          - type: precision_at_3
            value: 13.599
          - type: precision_at_5
            value: 9.202
          - type: recall_at_1
            value: 24.564
          - type: recall_at_10
            value: 44.36
          - type: recall_at_100
            value: 64.408
          - type: recall_at_1000
            value: 83.892
          - type: recall_at_3
            value: 34.653
          - type: recall_at_5
            value: 38.589
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackTexRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 17.01
          - type: map_at_10
            value: 24.485
          - type: map_at_100
            value: 25.573
          - type: map_at_1000
            value: 25.703
          - type: map_at_3
            value: 21.953
          - type: map_at_5
            value: 23.294999999999998
          - type: mrr_at_1
            value: 20.544
          - type: mrr_at_10
            value: 28.238000000000003
          - type: mrr_at_100
            value: 29.142000000000003
          - type: mrr_at_1000
            value: 29.219
          - type: mrr_at_3
            value: 25.802999999999997
          - type: mrr_at_5
            value: 27.105
          - type: ndcg_at_1
            value: 20.544
          - type: ndcg_at_10
            value: 29.387999999999998
          - type: ndcg_at_100
            value: 34.603
          - type: ndcg_at_1000
            value: 37.564
          - type: ndcg_at_3
            value: 24.731
          - type: ndcg_at_5
            value: 26.773000000000003
          - type: precision_at_1
            value: 20.544
          - type: precision_at_10
            value: 5.509
          - type: precision_at_100
            value: 0.9450000000000001
          - type: precision_at_1000
            value: 0.13799999999999998
          - type: precision_at_3
            value: 11.757
          - type: precision_at_5
            value: 8.596
          - type: recall_at_1
            value: 17.01
          - type: recall_at_10
            value: 40.392
          - type: recall_at_100
            value: 64.043
          - type: recall_at_1000
            value: 85.031
          - type: recall_at_3
            value: 27.293
          - type: recall_at_5
            value: 32.586999999999996
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackUnixRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 27.155
          - type: map_at_10
            value: 35.92
          - type: map_at_100
            value: 37.034
          - type: map_at_1000
            value: 37.139
          - type: map_at_3
            value: 33.263999999999996
          - type: map_at_5
            value: 34.61
          - type: mrr_at_1
            value: 32.183
          - type: mrr_at_10
            value: 40.099000000000004
          - type: mrr_at_100
            value: 41.001
          - type: mrr_at_1000
            value: 41.059
          - type: mrr_at_3
            value: 37.889
          - type: mrr_at_5
            value: 39.007999999999996
          - type: ndcg_at_1
            value: 32.183
          - type: ndcg_at_10
            value: 41.127
          - type: ndcg_at_100
            value: 46.464
          - type: ndcg_at_1000
            value: 48.67
          - type: ndcg_at_3
            value: 36.396
          - type: ndcg_at_5
            value: 38.313
          - type: precision_at_1
            value: 32.183
          - type: precision_at_10
            value: 6.847
          - type: precision_at_100
            value: 1.0739999999999998
          - type: precision_at_1000
            value: 0.13699999999999998
          - type: precision_at_3
            value: 16.356
          - type: precision_at_5
            value: 11.362
          - type: recall_at_1
            value: 27.155
          - type: recall_at_10
            value: 52.922000000000004
          - type: recall_at_100
            value: 76.39
          - type: recall_at_1000
            value: 91.553
          - type: recall_at_3
            value: 39.745999999999995
          - type: recall_at_5
            value: 44.637
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackWebmastersRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 25.523
          - type: map_at_10
            value: 34.268
          - type: map_at_100
            value: 35.835
          - type: map_at_1000
            value: 36.046
          - type: map_at_3
            value: 31.662000000000003
          - type: map_at_5
            value: 32.71
          - type: mrr_at_1
            value: 31.028
          - type: mrr_at_10
            value: 38.924
          - type: mrr_at_100
            value: 39.95
          - type: mrr_at_1000
            value: 40.003
          - type: mrr_at_3
            value: 36.594
          - type: mrr_at_5
            value: 37.701
          - type: ndcg_at_1
            value: 31.028
          - type: ndcg_at_10
            value: 39.848
          - type: ndcg_at_100
            value: 45.721000000000004
          - type: ndcg_at_1000
            value: 48.424
          - type: ndcg_at_3
            value: 35.329
          - type: ndcg_at_5
            value: 36.779
          - type: precision_at_1
            value: 31.028
          - type: precision_at_10
            value: 7.51
          - type: precision_at_100
            value: 1.478
          - type: precision_at_1000
            value: 0.24
          - type: precision_at_3
            value: 16.337
          - type: precision_at_5
            value: 11.383000000000001
          - type: recall_at_1
            value: 25.523
          - type: recall_at_10
            value: 50.735
          - type: recall_at_100
            value: 76.593
          - type: recall_at_1000
            value: 93.771
          - type: recall_at_3
            value: 37.574000000000005
          - type: recall_at_5
            value: 41.602
      - task:
          type: Retrieval
        dataset:
          type: BeIR/cqadupstack
          name: MTEB CQADupstackWordpressRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 20.746000000000002
          - type: map_at_10
            value: 28.557
          - type: map_at_100
            value: 29.575000000000003
          - type: map_at_1000
            value: 29.659000000000002
          - type: map_at_3
            value: 25.753999999999998
          - type: map_at_5
            value: 27.254
          - type: mrr_at_1
            value: 22.736
          - type: mrr_at_10
            value: 30.769000000000002
          - type: mrr_at_100
            value: 31.655
          - type: mrr_at_1000
            value: 31.717000000000002
          - type: mrr_at_3
            value: 28.065
          - type: mrr_at_5
            value: 29.543999999999997
          - type: ndcg_at_1
            value: 22.736
          - type: ndcg_at_10
            value: 33.545
          - type: ndcg_at_100
            value: 38.743
          - type: ndcg_at_1000
            value: 41.002
          - type: ndcg_at_3
            value: 28.021
          - type: ndcg_at_5
            value: 30.586999999999996
          - type: precision_at_1
            value: 22.736
          - type: precision_at_10
            value: 5.416
          - type: precision_at_100
            value: 0.8710000000000001
          - type: precision_at_1000
            value: 0.116
          - type: precision_at_3
            value: 11.953
          - type: precision_at_5
            value: 8.651
          - type: recall_at_1
            value: 20.746000000000002
          - type: recall_at_10
            value: 46.87
          - type: recall_at_100
            value: 71.25200000000001
          - type: recall_at_1000
            value: 88.26
          - type: recall_at_3
            value: 32.029999999999994
          - type: recall_at_5
            value: 38.21
      - task:
          type: Retrieval
        dataset:
          type: climate-fever
          name: MTEB ClimateFEVER
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 12.105
          - type: map_at_10
            value: 20.577
          - type: map_at_100
            value: 22.686999999999998
          - type: map_at_1000
            value: 22.889
          - type: map_at_3
            value: 17.174
          - type: map_at_5
            value: 18.807
          - type: mrr_at_1
            value: 27.101
          - type: mrr_at_10
            value: 38.475
          - type: mrr_at_100
            value: 39.491
          - type: mrr_at_1000
            value: 39.525
          - type: mrr_at_3
            value: 34.886
          - type: mrr_at_5
            value: 36.922
          - type: ndcg_at_1
            value: 27.101
          - type: ndcg_at_10
            value: 29.002
          - type: ndcg_at_100
            value: 37.218
          - type: ndcg_at_1000
            value: 40.644000000000005
          - type: ndcg_at_3
            value: 23.464
          - type: ndcg_at_5
            value: 25.262
          - type: precision_at_1
            value: 27.101
          - type: precision_at_10
            value: 9.179
          - type: precision_at_100
            value: 1.806
          - type: precision_at_1000
            value: 0.244
          - type: precision_at_3
            value: 17.394000000000002
          - type: precision_at_5
            value: 13.342
          - type: recall_at_1
            value: 12.105
          - type: recall_at_10
            value: 35.143
          - type: recall_at_100
            value: 63.44499999999999
          - type: recall_at_1000
            value: 82.49499999999999
          - type: recall_at_3
            value: 21.489
          - type: recall_at_5
            value: 26.82
      - task:
          type: Retrieval
        dataset:
          type: dbpedia-entity
          name: MTEB DBPedia
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 8.769
          - type: map_at_10
            value: 18.619
          - type: map_at_100
            value: 26.3
          - type: map_at_1000
            value: 28.063
          - type: map_at_3
            value: 13.746
          - type: map_at_5
            value: 16.035
          - type: mrr_at_1
            value: 65.25
          - type: mrr_at_10
            value: 73.678
          - type: mrr_at_100
            value: 73.993
          - type: mrr_at_1000
            value: 74.003
          - type: mrr_at_3
            value: 72.042
          - type: mrr_at_5
            value: 72.992
          - type: ndcg_at_1
            value: 53.625
          - type: ndcg_at_10
            value: 39.638
          - type: ndcg_at_100
            value: 44.601
          - type: ndcg_at_1000
            value: 52.80200000000001
          - type: ndcg_at_3
            value: 44.727
          - type: ndcg_at_5
            value: 42.199
          - type: precision_at_1
            value: 65.25
          - type: precision_at_10
            value: 31.025000000000002
          - type: precision_at_100
            value: 10.174999999999999
          - type: precision_at_1000
            value: 2.0740000000000003
          - type: precision_at_3
            value: 48.083
          - type: precision_at_5
            value: 40.6
          - type: recall_at_1
            value: 8.769
          - type: recall_at_10
            value: 23.910999999999998
          - type: recall_at_100
            value: 51.202999999999996
          - type: recall_at_1000
            value: 77.031
          - type: recall_at_3
            value: 15.387999999999998
          - type: recall_at_5
            value: 18.919
      - task:
          type: Classification
        dataset:
          type: mteb/emotion
          name: MTEB EmotionClassification
          config: default
          split: test
          revision: 4f58c6b202a23cf9a4da393831edf4f9183cad37
        metrics:
          - type: accuracy
            value: 54.47
          - type: f1
            value: 48.21839043361556
      - task:
          type: Retrieval
        dataset:
          type: fever
          name: MTEB FEVER
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 63.564
          - type: map_at_10
            value: 74.236
          - type: map_at_100
            value: 74.53699999999999
          - type: map_at_1000
            value: 74.557
          - type: map_at_3
            value: 72.556
          - type: map_at_5
            value: 73.656
          - type: mrr_at_1
            value: 68.497
          - type: mrr_at_10
            value: 78.373
          - type: mrr_at_100
            value: 78.54299999999999
          - type: mrr_at_1000
            value: 78.549
          - type: mrr_at_3
            value: 77.03
          - type: mrr_at_5
            value: 77.938
          - type: ndcg_at_1
            value: 68.497
          - type: ndcg_at_10
            value: 79.12599999999999
          - type: ndcg_at_100
            value: 80.319
          - type: ndcg_at_1000
            value: 80.71199999999999
          - type: ndcg_at_3
            value: 76.209
          - type: ndcg_at_5
            value: 77.90700000000001
          - type: precision_at_1
            value: 68.497
          - type: precision_at_10
            value: 9.958
          - type: precision_at_100
            value: 1.077
          - type: precision_at_1000
            value: 0.11299999999999999
          - type: precision_at_3
            value: 29.908
          - type: precision_at_5
            value: 18.971
          - type: recall_at_1
            value: 63.564
          - type: recall_at_10
            value: 90.05199999999999
          - type: recall_at_100
            value: 95.028
          - type: recall_at_1000
            value: 97.667
          - type: recall_at_3
            value: 82.17999999999999
          - type: recall_at_5
            value: 86.388
      - task:
          type: Retrieval
        dataset:
          type: fiqa
          name: MTEB FiQA2018
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 19.042
          - type: map_at_10
            value: 30.764999999999997
          - type: map_at_100
            value: 32.678000000000004
          - type: map_at_1000
            value: 32.881
          - type: map_at_3
            value: 26.525
          - type: map_at_5
            value: 28.932000000000002
          - type: mrr_at_1
            value: 37.653999999999996
          - type: mrr_at_10
            value: 46.597
          - type: mrr_at_100
            value: 47.413
          - type: mrr_at_1000
            value: 47.453
          - type: mrr_at_3
            value: 43.775999999999996
          - type: mrr_at_5
            value: 45.489000000000004
          - type: ndcg_at_1
            value: 37.653999999999996
          - type: ndcg_at_10
            value: 38.615
          - type: ndcg_at_100
            value: 45.513999999999996
          - type: ndcg_at_1000
            value: 48.815999999999995
          - type: ndcg_at_3
            value: 34.427
          - type: ndcg_at_5
            value: 35.954
          - type: precision_at_1
            value: 37.653999999999996
          - type: precision_at_10
            value: 10.864
          - type: precision_at_100
            value: 1.7850000000000001
          - type: precision_at_1000
            value: 0.23800000000000002
          - type: precision_at_3
            value: 22.788
          - type: precision_at_5
            value: 17.346
          - type: recall_at_1
            value: 19.042
          - type: recall_at_10
            value: 45.707
          - type: recall_at_100
            value: 71.152
          - type: recall_at_1000
            value: 90.7
          - type: recall_at_3
            value: 30.814000000000004
          - type: recall_at_5
            value: 37.478
      - task:
          type: Retrieval
        dataset:
          type: hotpotqa
          name: MTEB HotpotQA
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 38.001000000000005
          - type: map_at_10
            value: 59.611000000000004
          - type: map_at_100
            value: 60.582
          - type: map_at_1000
            value: 60.646
          - type: map_at_3
            value: 56.031
          - type: map_at_5
            value: 58.243
          - type: mrr_at_1
            value: 76.003
          - type: mrr_at_10
            value: 82.15400000000001
          - type: mrr_at_100
            value: 82.377
          - type: mrr_at_1000
            value: 82.383
          - type: mrr_at_3
            value: 81.092
          - type: mrr_at_5
            value: 81.742
          - type: ndcg_at_1
            value: 76.003
          - type: ndcg_at_10
            value: 68.216
          - type: ndcg_at_100
            value: 71.601
          - type: ndcg_at_1000
            value: 72.821
          - type: ndcg_at_3
            value: 63.109
          - type: ndcg_at_5
            value: 65.902
          - type: precision_at_1
            value: 76.003
          - type: precision_at_10
            value: 14.379
          - type: precision_at_100
            value: 1.702
          - type: precision_at_1000
            value: 0.186
          - type: precision_at_3
            value: 40.396
          - type: precision_at_5
            value: 26.442
          - type: recall_at_1
            value: 38.001000000000005
          - type: recall_at_10
            value: 71.897
          - type: recall_at_100
            value: 85.105
          - type: recall_at_1000
            value: 93.133
          - type: recall_at_3
            value: 60.594
          - type: recall_at_5
            value: 66.104
      - task:
          type: Classification
        dataset:
          type: mteb/imdb
          name: MTEB ImdbClassification
          config: default
          split: test
          revision: 3d86128a09e091d6018b6d26cad27f2739fc2db7
        metrics:
          - type: accuracy
            value: 91.31280000000001
          - type: ap
            value: 87.53723467501632
          - type: f1
            value: 91.30282906596291
      - task:
          type: Retrieval
        dataset:
          type: msmarco
          name: MTEB MSMARCO
          config: default
          split: dev
          revision: None
        metrics:
          - type: map_at_1
            value: 21.917
          - type: map_at_10
            value: 34.117999999999995
          - type: map_at_100
            value: 35.283
          - type: map_at_1000
            value: 35.333999999999996
          - type: map_at_3
            value: 30.330000000000002
          - type: map_at_5
            value: 32.461
          - type: mrr_at_1
            value: 22.579
          - type: mrr_at_10
            value: 34.794000000000004
          - type: mrr_at_100
            value: 35.893
          - type: mrr_at_1000
            value: 35.937000000000005
          - type: mrr_at_3
            value: 31.091
          - type: mrr_at_5
            value: 33.173
          - type: ndcg_at_1
            value: 22.579
          - type: ndcg_at_10
            value: 40.951
          - type: ndcg_at_100
            value: 46.558
          - type: ndcg_at_1000
            value: 47.803000000000004
          - type: ndcg_at_3
            value: 33.262
          - type: ndcg_at_5
            value: 37.036
          - type: precision_at_1
            value: 22.579
          - type: precision_at_10
            value: 6.463000000000001
          - type: precision_at_100
            value: 0.928
          - type: precision_at_1000
            value: 0.104
          - type: precision_at_3
            value: 14.174000000000001
          - type: precision_at_5
            value: 10.421
          - type: recall_at_1
            value: 21.917
          - type: recall_at_10
            value: 61.885
          - type: recall_at_100
            value: 87.847
          - type: recall_at_1000
            value: 97.322
          - type: recall_at_3
            value: 41.010000000000005
          - type: recall_at_5
            value: 50.031000000000006
      - task:
          type: Classification
        dataset:
          type: mteb/mtop_domain
          name: MTEB MTOPDomainClassification (en)
          config: en
          split: test
          revision: d80d48c1eb48d3562165c59d59d0034df9fff0bf
        metrics:
          - type: accuracy
            value: 93.49521203830369
          - type: f1
            value: 93.30882341740241
      - task:
          type: Classification
        dataset:
          type: mteb/mtop_intent
          name: MTEB MTOPIntentClassification (en)
          config: en
          split: test
          revision: ae001d0e6b1228650b7bd1c2c65fb50ad11a8aba
        metrics:
          - type: accuracy
            value: 71.0579115367077
          - type: f1
            value: 51.2368258319339
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_massive_intent
          name: MTEB MassiveIntentClassification (en)
          config: en
          split: test
          revision: 31efe3c427b0bae9c22cbb560b8f15491cc6bed7
        metrics:
          - type: accuracy
            value: 73.88029589778077
          - type: f1
            value: 72.34422048584663
      - task:
          type: Classification
        dataset:
          type: mteb/amazon_massive_scenario
          name: MTEB MassiveScenarioClassification (en)
          config: en
          split: test
          revision: 7d571f92784cd94a019292a1f45445077d0ef634
        metrics:
          - type: accuracy
            value: 78.2817753866846
          - type: f1
            value: 77.87746050004304
      - task:
          type: Clustering
        dataset:
          type: mteb/medrxiv-clustering-p2p
          name: MTEB MedrxivClusteringP2P
          config: default
          split: test
          revision: e7a26af6f3ae46b30dde8737f02c07b1505bcc73
        metrics:
          - type: v_measure
            value: 33.247341454119216
      - task:
          type: Clustering
        dataset:
          type: mteb/medrxiv-clustering-s2s
          name: MTEB MedrxivClusteringS2S
          config: default
          split: test
          revision: 35191c8c0dca72d8ff3efcd72aa802307d469663
        metrics:
          - type: v_measure
            value: 31.9647477166234
      - task:
          type: Reranking
        dataset:
          type: mteb/mind_small
          name: MTEB MindSmallReranking
          config: default
          split: test
          revision: 3bdac13927fdc888b903db93b2ffdbd90b295a69
        metrics:
          - type: map
            value: 31.90698374676892
          - type: mrr
            value: 33.07523683771251
      - task:
          type: Retrieval
        dataset:
          type: nfcorpus
          name: MTEB NFCorpus
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 6.717
          - type: map_at_10
            value: 14.566
          - type: map_at_100
            value: 18.465999999999998
          - type: map_at_1000
            value: 20.033
          - type: map_at_3
            value: 10.863
          - type: map_at_5
            value: 12.589
          - type: mrr_at_1
            value: 49.845
          - type: mrr_at_10
            value: 58.385
          - type: mrr_at_100
            value: 58.989999999999995
          - type: mrr_at_1000
            value: 59.028999999999996
          - type: mrr_at_3
            value: 56.76
          - type: mrr_at_5
            value: 57.766
          - type: ndcg_at_1
            value: 47.678
          - type: ndcg_at_10
            value: 37.511
          - type: ndcg_at_100
            value: 34.537
          - type: ndcg_at_1000
            value: 43.612
          - type: ndcg_at_3
            value: 43.713
          - type: ndcg_at_5
            value: 41.303
          - type: precision_at_1
            value: 49.845
          - type: precision_at_10
            value: 27.307
          - type: precision_at_100
            value: 8.746
          - type: precision_at_1000
            value: 2.182
          - type: precision_at_3
            value: 40.764
          - type: precision_at_5
            value: 35.232
          - type: recall_at_1
            value: 6.717
          - type: recall_at_10
            value: 18.107
          - type: recall_at_100
            value: 33.759
          - type: recall_at_1000
            value: 67.31
          - type: recall_at_3
            value: 11.68
          - type: recall_at_5
            value: 14.557999999999998
      - task:
          type: Retrieval
        dataset:
          type: nq
          name: MTEB NQ
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 27.633999999999997
          - type: map_at_10
            value: 42.400999999999996
          - type: map_at_100
            value: 43.561
          - type: map_at_1000
            value: 43.592
          - type: map_at_3
            value: 37.865
          - type: map_at_5
            value: 40.650999999999996
          - type: mrr_at_1
            value: 31.286
          - type: mrr_at_10
            value: 44.996
          - type: mrr_at_100
            value: 45.889
          - type: mrr_at_1000
            value: 45.911
          - type: mrr_at_3
            value: 41.126000000000005
          - type: mrr_at_5
            value: 43.536
          - type: ndcg_at_1
            value: 31.257
          - type: ndcg_at_10
            value: 50.197
          - type: ndcg_at_100
            value: 55.062
          - type: ndcg_at_1000
            value: 55.81700000000001
          - type: ndcg_at_3
            value: 41.650999999999996
          - type: ndcg_at_5
            value: 46.324
          - type: precision_at_1
            value: 31.257
          - type: precision_at_10
            value: 8.508000000000001
          - type: precision_at_100
            value: 1.121
          - type: precision_at_1000
            value: 0.11900000000000001
          - type: precision_at_3
            value: 19.1
          - type: precision_at_5
            value: 14.16
          - type: recall_at_1
            value: 27.633999999999997
          - type: recall_at_10
            value: 71.40100000000001
          - type: recall_at_100
            value: 92.463
          - type: recall_at_1000
            value: 98.13199999999999
          - type: recall_at_3
            value: 49.382
          - type: recall_at_5
            value: 60.144
      - task:
          type: Retrieval
        dataset:
          type: quora
          name: MTEB QuoraRetrieval
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 71.17099999999999
          - type: map_at_10
            value: 85.036
          - type: map_at_100
            value: 85.67099999999999
          - type: map_at_1000
            value: 85.68599999999999
          - type: map_at_3
            value: 82.086
          - type: map_at_5
            value: 83.956
          - type: mrr_at_1
            value: 82.04
          - type: mrr_at_10
            value: 88.018
          - type: mrr_at_100
            value: 88.114
          - type: mrr_at_1000
            value: 88.115
          - type: mrr_at_3
            value: 87.047
          - type: mrr_at_5
            value: 87.73100000000001
          - type: ndcg_at_1
            value: 82.03
          - type: ndcg_at_10
            value: 88.717
          - type: ndcg_at_100
            value: 89.904
          - type: ndcg_at_1000
            value: 89.991
          - type: ndcg_at_3
            value: 85.89099999999999
          - type: ndcg_at_5
            value: 87.485
          - type: precision_at_1
            value: 82.03
          - type: precision_at_10
            value: 13.444999999999999
          - type: precision_at_100
            value: 1.533
          - type: precision_at_1000
            value: 0.157
          - type: precision_at_3
            value: 37.537
          - type: precision_at_5
            value: 24.692
          - type: recall_at_1
            value: 71.17099999999999
          - type: recall_at_10
            value: 95.634
          - type: recall_at_100
            value: 99.614
          - type: recall_at_1000
            value: 99.99
          - type: recall_at_3
            value: 87.48
          - type: recall_at_5
            value: 91.996
      - task:
          type: Clustering
        dataset:
          type: mteb/reddit-clustering
          name: MTEB RedditClustering
          config: default
          split: test
          revision: 24640382cdbf8abc73003fb0fa6d111a705499eb
        metrics:
          - type: v_measure
            value: 55.067219624685315
      - task:
          type: Clustering
        dataset:
          type: mteb/reddit-clustering-p2p
          name: MTEB RedditClusteringP2P
          config: default
          split: test
          revision: 282350215ef01743dc01b456c7f5241fa8937f16
        metrics:
          - type: v_measure
            value: 62.121822992300444
      - task:
          type: Retrieval
        dataset:
          type: scidocs
          name: MTEB SCIDOCS
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 4.153
          - type: map_at_10
            value: 11.024000000000001
          - type: map_at_100
            value: 13.233
          - type: map_at_1000
            value: 13.62
          - type: map_at_3
            value: 7.779999999999999
          - type: map_at_5
            value: 9.529
          - type: mrr_at_1
            value: 20.599999999999998
          - type: mrr_at_10
            value: 31.361
          - type: mrr_at_100
            value: 32.738
          - type: mrr_at_1000
            value: 32.792
          - type: mrr_at_3
            value: 28.15
          - type: mrr_at_5
            value: 30.085
          - type: ndcg_at_1
            value: 20.599999999999998
          - type: ndcg_at_10
            value: 18.583
          - type: ndcg_at_100
            value: 27.590999999999998
          - type: ndcg_at_1000
            value: 34.001
          - type: ndcg_at_3
            value: 17.455000000000002
          - type: ndcg_at_5
            value: 15.588
          - type: precision_at_1
            value: 20.599999999999998
          - type: precision_at_10
            value: 9.74
          - type: precision_at_100
            value: 2.284
          - type: precision_at_1000
            value: 0.381
          - type: precision_at_3
            value: 16.533
          - type: precision_at_5
            value: 14.02
          - type: recall_at_1
            value: 4.153
          - type: recall_at_10
            value: 19.738
          - type: recall_at_100
            value: 46.322
          - type: recall_at_1000
            value: 77.378
          - type: recall_at_3
            value: 10.048
          - type: recall_at_5
            value: 14.233
      - task:
          type: STS
        dataset:
          type: mteb/sickr-sts
          name: MTEB SICK-R
          config: default
          split: test
          revision: a6ea5a8cab320b040a23452cc28066d9beae2cee
        metrics:
          - type: cos_sim_pearson
            value: 85.07097501003639
          - type: cos_sim_spearman
            value: 81.05827848407056
          - type: euclidean_pearson
            value: 82.6279003372546
          - type: euclidean_spearman
            value: 81.00031515279802
          - type: manhattan_pearson
            value: 82.59338284959495
          - type: manhattan_spearman
            value: 80.97432711064945
      - task:
          type: STS
        dataset:
          type: mteb/sts12-sts
          name: MTEB STS12
          config: default
          split: test
          revision: a0d554a64d88156834ff5ae9920b964011b16384
        metrics:
          - type: cos_sim_pearson
            value: 86.28991993621685
          - type: cos_sim_spearman
            value: 78.71828082424351
          - type: euclidean_pearson
            value: 83.4881331520832
          - type: euclidean_spearman
            value: 78.51746826842316
          - type: manhattan_pearson
            value: 83.4109223774324
          - type: manhattan_spearman
            value: 78.431544382179
      - task:
          type: STS
        dataset:
          type: mteb/sts13-sts
          name: MTEB STS13
          config: default
          split: test
          revision: 7e90230a92c190f1bf69ae9002b8cea547a64cca
        metrics:
          - type: cos_sim_pearson
            value: 83.16651661072123
          - type: cos_sim_spearman
            value: 84.88094386637867
          - type: euclidean_pearson
            value: 84.3547603585416
          - type: euclidean_spearman
            value: 84.85148665860193
          - type: manhattan_pearson
            value: 84.29648369879266
          - type: manhattan_spearman
            value: 84.76074870571124
      - task:
          type: STS
        dataset:
          type: mteb/sts14-sts
          name: MTEB STS14
          config: default
          split: test
          revision: 6031580fec1f6af667f0bd2da0a551cf4f0b2375
        metrics:
          - type: cos_sim_pearson
            value: 83.40596254292149
          - type: cos_sim_spearman
            value: 83.10699573133829
          - type: euclidean_pearson
            value: 83.22794776876958
          - type: euclidean_spearman
            value: 83.22583316084712
          - type: manhattan_pearson
            value: 83.15899233935681
          - type: manhattan_spearman
            value: 83.17668293648019
      - task:
          type: STS
        dataset:
          type: mteb/sts15-sts
          name: MTEB STS15
          config: default
          split: test
          revision: ae752c7c21bf194d8b67fd573edf7ae58183cbe3
        metrics:
          - type: cos_sim_pearson
            value: 87.27977121352563
          - type: cos_sim_spearman
            value: 88.73903130248591
          - type: euclidean_pearson
            value: 88.30685958438735
          - type: euclidean_spearman
            value: 88.79755484280406
          - type: manhattan_pearson
            value: 88.30305607758652
          - type: manhattan_spearman
            value: 88.80096577072784
      - task:
          type: STS
        dataset:
          type: mteb/sts16-sts
          name: MTEB STS16
          config: default
          split: test
          revision: 4d8694f8f0e0100860b497b999b3dbed754a0513
        metrics:
          - type: cos_sim_pearson
            value: 84.08819031430218
          - type: cos_sim_spearman
            value: 86.35414445951125
          - type: euclidean_pearson
            value: 85.4683192388315
          - type: euclidean_spearman
            value: 86.2079674669473
          - type: manhattan_pearson
            value: 85.35835702257341
          - type: manhattan_spearman
            value: 86.08483380002187
      - task:
          type: STS
        dataset:
          type: mteb/sts17-crosslingual-sts
          name: MTEB STS17 (en-en)
          config: en-en
          split: test
          revision: af5e6fb845001ecf41f4c1e033ce921939a2a68d
        metrics:
          - type: cos_sim_pearson
            value: 87.36149449801478
          - type: cos_sim_spearman
            value: 87.7102980757725
          - type: euclidean_pearson
            value: 88.16457177837161
          - type: euclidean_spearman
            value: 87.6598652482716
          - type: manhattan_pearson
            value: 88.23894728971618
          - type: manhattan_spearman
            value: 87.74470156709361
      - task:
          type: STS
        dataset:
          type: mteb/sts22-crosslingual-sts
          name: MTEB STS22 (en)
          config: en
          split: test
          revision: 6d1ba47164174a496b7fa5d3569dae26a6813b80
        metrics:
          - type: cos_sim_pearson
            value: 64.54023758394433
          - type: cos_sim_spearman
            value: 66.28491960187773
          - type: euclidean_pearson
            value: 67.0853128483472
          - type: euclidean_spearman
            value: 66.10307543766307
          - type: manhattan_pearson
            value: 66.7635365592556
          - type: manhattan_spearman
            value: 65.76408004780167
      - task:
          type: STS
        dataset:
          type: mteb/stsbenchmark-sts
          name: MTEB STSBenchmark
          config: default
          split: test
          revision: b0fddb56ed78048fa8b90373c8a3cfc37b684831
        metrics:
          - type: cos_sim_pearson
            value: 85.15858398195317
          - type: cos_sim_spearman
            value: 87.44850004752102
          - type: euclidean_pearson
            value: 86.60737082550408
          - type: euclidean_spearman
            value: 87.31591549824242
          - type: manhattan_pearson
            value: 86.56187011429977
          - type: manhattan_spearman
            value: 87.23854795795319
      - task:
          type: Reranking
        dataset:
          type: mteb/scidocs-reranking
          name: MTEB SciDocsRR
          config: default
          split: test
          revision: d3c5e1fc0b855ab6097bf1cda04dd73947d7caab
        metrics:
          - type: map
            value: 86.66210488769109
          - type: mrr
            value: 96.23100664767331
      - task:
          type: Retrieval
        dataset:
          type: scifact
          name: MTEB SciFact
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 56.094
          - type: map_at_10
            value: 67.486
          - type: map_at_100
            value: 67.925
          - type: map_at_1000
            value: 67.949
          - type: map_at_3
            value: 64.857
          - type: map_at_5
            value: 66.31
          - type: mrr_at_1
            value: 58.667
          - type: mrr_at_10
            value: 68.438
          - type: mrr_at_100
            value: 68.733
          - type: mrr_at_1000
            value: 68.757
          - type: mrr_at_3
            value: 66.389
          - type: mrr_at_5
            value: 67.456
          - type: ndcg_at_1
            value: 58.667
          - type: ndcg_at_10
            value: 72.506
          - type: ndcg_at_100
            value: 74.27
          - type: ndcg_at_1000
            value: 74.94800000000001
          - type: ndcg_at_3
            value: 67.977
          - type: ndcg_at_5
            value: 70.028
          - type: precision_at_1
            value: 58.667
          - type: precision_at_10
            value: 9.767000000000001
          - type: precision_at_100
            value: 1.073
          - type: precision_at_1000
            value: 0.11299999999999999
          - type: precision_at_3
            value: 27
          - type: precision_at_5
            value: 17.666999999999998
          - type: recall_at_1
            value: 56.094
          - type: recall_at_10
            value: 86.68900000000001
          - type: recall_at_100
            value: 94.333
          - type: recall_at_1000
            value: 99.667
          - type: recall_at_3
            value: 74.522
          - type: recall_at_5
            value: 79.611
      - task:
          type: PairClassification
        dataset:
          type: mteb/sprintduplicatequestions-pairclassification
          name: MTEB SprintDuplicateQuestions
          config: default
          split: test
          revision: d66bd1f72af766a5cc4b0ca5e00c162f89e8cc46
        metrics:
          - type: cos_sim_accuracy
            value: 99.83069306930693
          - type: cos_sim_ap
            value: 95.69184662911199
          - type: cos_sim_f1
            value: 91.4027149321267
          - type: cos_sim_precision
            value: 91.91102123356926
          - type: cos_sim_recall
            value: 90.9
          - type: dot_accuracy
            value: 99.69405940594059
          - type: dot_ap
            value: 90.21674151456216
          - type: dot_f1
            value: 84.4489179667841
          - type: dot_precision
            value: 85.00506585612969
          - type: dot_recall
            value: 83.89999999999999
          - type: euclidean_accuracy
            value: 99.83069306930693
          - type: euclidean_ap
            value: 95.67760109671087
          - type: euclidean_f1
            value: 91.19754350051177
          - type: euclidean_precision
            value: 93.39622641509435
          - type: euclidean_recall
            value: 89.1
          - type: manhattan_accuracy
            value: 99.83267326732673
          - type: manhattan_ap
            value: 95.69771347732625
          - type: manhattan_f1
            value: 91.32420091324201
          - type: manhattan_precision
            value: 92.68795056642637
          - type: manhattan_recall
            value: 90
          - type: max_accuracy
            value: 99.83267326732673
          - type: max_ap
            value: 95.69771347732625
          - type: max_f1
            value: 91.4027149321267
      - task:
          type: Clustering
        dataset:
          type: mteb/stackexchange-clustering
          name: MTEB StackExchangeClustering
          config: default
          split: test
          revision: 6cbc1f7b2bc0622f2e39d2c77fa502909748c259
        metrics:
          - type: v_measure
            value: 64.47378332953092
      - task:
          type: Clustering
        dataset:
          type: mteb/stackexchange-clustering-p2p
          name: MTEB StackExchangeClusteringP2P
          config: default
          split: test
          revision: 815ca46b2622cec33ccafc3735d572c266efdb44
        metrics:
          - type: v_measure
            value: 33.79602531604151
      - task:
          type: Reranking
        dataset:
          type: mteb/stackoverflowdupquestions-reranking
          name: MTEB StackOverflowDupQuestions
          config: default
          split: test
          revision: e185fbe320c72810689fc5848eb6114e1ef5ec69
        metrics:
          - type: map
            value: 53.80707639107175
          - type: mrr
            value: 54.64886522790935
      - task:
          type: Summarization
        dataset:
          type: mteb/summeval
          name: MTEB SummEval
          config: default
          split: test
          revision: cda12ad7615edc362dbf25a00fdd61d3b1eaf93c
        metrics:
          - type: cos_sim_pearson
            value: 30.852448373051395
          - type: cos_sim_spearman
            value: 32.51821499493775
          - type: dot_pearson
            value: 30.390650062190456
          - type: dot_spearman
            value: 30.588836159667636
      - task:
          type: Retrieval
        dataset:
          type: trec-covid
          name: MTEB TRECCOVID
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 0.198
          - type: map_at_10
            value: 1.51
          - type: map_at_100
            value: 8.882
          - type: map_at_1000
            value: 22.181
          - type: map_at_3
            value: 0.553
          - type: map_at_5
            value: 0.843
          - type: mrr_at_1
            value: 74
          - type: mrr_at_10
            value: 84.89999999999999
          - type: mrr_at_100
            value: 84.89999999999999
          - type: mrr_at_1000
            value: 84.89999999999999
          - type: mrr_at_3
            value: 84
          - type: mrr_at_5
            value: 84.89999999999999
          - type: ndcg_at_1
            value: 68
          - type: ndcg_at_10
            value: 64.792
          - type: ndcg_at_100
            value: 51.37199999999999
          - type: ndcg_at_1000
            value: 47.392
          - type: ndcg_at_3
            value: 68.46900000000001
          - type: ndcg_at_5
            value: 67.084
          - type: precision_at_1
            value: 74
          - type: precision_at_10
            value: 69.39999999999999
          - type: precision_at_100
            value: 53.080000000000005
          - type: precision_at_1000
            value: 21.258
          - type: precision_at_3
            value: 76
          - type: precision_at_5
            value: 73.2
          - type: recall_at_1
            value: 0.198
          - type: recall_at_10
            value: 1.7950000000000002
          - type: recall_at_100
            value: 12.626999999999999
          - type: recall_at_1000
            value: 44.84
          - type: recall_at_3
            value: 0.611
          - type: recall_at_5
            value: 0.959
      - task:
          type: Retrieval
        dataset:
          type: webis-touche2020
          name: MTEB Touche2020
          config: default
          split: test
          revision: None
        metrics:
          - type: map_at_1
            value: 1.4949999999999999
          - type: map_at_10
            value: 8.797
          - type: map_at_100
            value: 14.889
          - type: map_at_1000
            value: 16.309
          - type: map_at_3
            value: 4.389
          - type: map_at_5
            value: 6.776
          - type: mrr_at_1
            value: 18.367
          - type: mrr_at_10
            value: 35.844
          - type: mrr_at_100
            value: 37.119
          - type: mrr_at_1000
            value: 37.119
          - type: mrr_at_3
            value: 30.612000000000002
          - type: mrr_at_5
            value: 33.163
          - type: ndcg_at_1
            value: 16.326999999999998
          - type: ndcg_at_10
            value: 21.9
          - type: ndcg_at_100
            value: 34.705000000000005
          - type: ndcg_at_1000
            value: 45.709
          - type: ndcg_at_3
            value: 22.7
          - type: ndcg_at_5
            value: 23.197000000000003
          - type: precision_at_1
            value: 18.367
          - type: precision_at_10
            value: 21.02
          - type: precision_at_100
            value: 7.714
          - type: precision_at_1000
            value: 1.504
          - type: precision_at_3
            value: 26.531
          - type: precision_at_5
            value: 26.122
          - type: recall_at_1
            value: 1.4949999999999999
          - type: recall_at_10
            value: 15.504000000000001
          - type: recall_at_100
            value: 47.978
          - type: recall_at_1000
            value: 81.56
          - type: recall_at_3
            value: 5.569
          - type: recall_at_5
            value: 9.821
      - task:
          type: Classification
        dataset:
          type: mteb/toxic_conversations_50k
          name: MTEB ToxicConversationsClassification
          config: default
          split: test
          revision: d7c0de2777da35d6aae2200a62c6e0e5af397c4c
        metrics:
          - type: accuracy
            value: 72.99279999999999
          - type: ap
            value: 15.459189680101492
          - type: f1
            value: 56.33023271441895
      - task:
          type: Classification
        dataset:
          type: mteb/tweet_sentiment_extraction
          name: MTEB TweetSentimentExtractionClassification
          config: default
          split: test
          revision: d604517c81ca91fe16a244d1248fc021f9ecee7a
        metrics:
          - type: accuracy
            value: 63.070175438596486
          - type: f1
            value: 63.28070758709465
      - task:
          type: Clustering
        dataset:
          type: mteb/twentynewsgroups-clustering
          name: MTEB TwentyNewsgroupsClustering
          config: default
          split: test
          revision: 6125ec4e24fa026cec8a478383ee943acfbd5449
        metrics:
          - type: v_measure
            value: 50.076231309703054
      - task:
          type: PairClassification
        dataset:
          type: mteb/twittersemeval2015-pairclassification
          name: MTEB TwitterSemEval2015
          config: default
          split: test
          revision: 70970daeab8776df92f5ea462b6173c0b46fd2d1
        metrics:
          - type: cos_sim_accuracy
            value: 87.21463908922931
          - type: cos_sim_ap
            value: 77.67287017966282
          - type: cos_sim_f1
            value: 70.34412955465588
          - type: cos_sim_precision
            value: 67.57413709285368
          - type: cos_sim_recall
            value: 73.35092348284961
          - type: dot_accuracy
            value: 85.04500208618943
          - type: dot_ap
            value: 70.4075203869744
          - type: dot_f1
            value: 66.18172537008678
          - type: dot_precision
            value: 64.08798813643104
          - type: dot_recall
            value: 68.41688654353561
          - type: euclidean_accuracy
            value: 87.17887584192646
          - type: euclidean_ap
            value: 77.5774128274464
          - type: euclidean_f1
            value: 70.09307972480777
          - type: euclidean_precision
            value: 71.70852884349986
          - type: euclidean_recall
            value: 68.54881266490766
          - type: manhattan_accuracy
            value: 87.28020504261787
          - type: manhattan_ap
            value: 77.57835820297892
          - type: manhattan_f1
            value: 70.23063591521131
          - type: manhattan_precision
            value: 70.97817299919159
          - type: manhattan_recall
            value: 69.49868073878628
          - type: max_accuracy
            value: 87.28020504261787
          - type: max_ap
            value: 77.67287017966282
          - type: max_f1
            value: 70.34412955465588
      - task:
          type: PairClassification
        dataset:
          type: mteb/twitterurlcorpus-pairclassification
          name: MTEB TwitterURLCorpus
          config: default
          split: test
          revision: 8b6510b0b1fa4e4c4f879467980e9be563ec1cdf
        metrics:
          - type: cos_sim_accuracy
            value: 88.96650754841464
          - type: cos_sim_ap
            value: 86.00185968965064
          - type: cos_sim_f1
            value: 77.95861256351718
          - type: cos_sim_precision
            value: 74.70712773465067
          - type: cos_sim_recall
            value: 81.50600554357868
          - type: dot_accuracy
            value: 87.36950362867233
          - type: dot_ap
            value: 82.22071181147555
          - type: dot_f1
            value: 74.85680716698488
          - type: dot_precision
            value: 71.54688377316114
          - type: dot_recall
            value: 78.48783492454572
          - type: euclidean_accuracy
            value: 88.99561454573679
          - type: euclidean_ap
            value: 86.15882097229648
          - type: euclidean_f1
            value: 78.18463125322332
          - type: euclidean_precision
            value: 74.95408956067241
          - type: euclidean_recall
            value: 81.70619032953496
          - type: manhattan_accuracy
            value: 88.96650754841464
          - type: manhattan_ap
            value: 86.13133111232099
          - type: manhattan_f1
            value: 78.10771470160115
          - type: manhattan_precision
            value: 74.05465084184377
          - type: manhattan_recall
            value: 82.63012011087157
          - type: max_accuracy
            value: 88.99561454573679
          - type: max_ap
            value: 86.15882097229648
          - type: max_f1
            value: 78.18463125322332
language:
  - en
license: mit

新闻 | News

[2024-04-06] 开源puff系列模型,专门针对检索和语义匹配任务,更多的考虑泛化性和私有通用测试集效果,向量维度可变,中英双语

[2024-02-27] 开源stella-mrl-large-zh-v3.5-1792d模型,支持向量可变维度

[2024-02-17] 开源stella v3系列、dialogue编码模型和相关训练数据。

[2023-10-19] 开源stella-base-en-v2 使用简单,不需要任何前缀文本

[2023-10-12] 开源stella-base-zh-v2和stella-large-zh-v2, 效果更好且使用简单,不需要任何前缀文本

[2023-09-11] 开源stella-base-zh和stella-large-zh

欢迎去本人主页查看最新模型,并提出您的宝贵意见!

stella model

stella是一个通用的文本编码模型,主要有以下模型:

Model Name Model Size (GB) Dimension Sequence Length Language Need instruction for retrieval?
stella-base-en-v2 0.2 768 512 English No
stella-large-zh-v2 0.65 1024 1024 Chinese No
stella-base-zh-v2 0.2 768 1024 Chinese No
stella-large-zh 0.65 1024 1024 Chinese Yes
stella-base-zh 0.2 768 1024 Chinese Yes

完整的训练思路和训练过程已记录在博客1博客2,欢迎阅读讨论。

训练数据:

  1. 开源数据(wudao_base_200GB[1]、m3e[2]和simclue[3]),着重挑选了长度大于512的文本
  2. 在通用语料库上使用LLM构造一批(question, paragraph)和(sentence, paragraph)数据

训练方法:

  1. 对比学习损失函数
  2. 带有难负例的对比学习损失函数(分别基于bm25和vector构造了难负例)
  3. EWC(Elastic Weights Consolidation)[4]
  4. cosent loss[5]
  5. 每一种类型的数据一个迭代器,分别计算loss进行更新

stella-v2在stella模型的基础上,使用了更多的训练数据,同时知识蒸馏等方法去除了前置的instruction( 比如piccolo的查询:, 结果:, e5的query:passage:)。

初始权重:
stella-base-zh和stella-large-zh分别以piccolo-base-zh[6]和piccolo-large-zh作为基础模型,512-1024的position embedding使用层次分解位置编码[7]进行初始化。
感谢商汤科技研究院开源的piccolo系列模型

stella is a general-purpose text encoder, which mainly includes the following models:

Model Name Model Size (GB) Dimension Sequence Length Language Need instruction for retrieval?
stella-base-en-v2 0.2 768 512 English No
stella-large-zh-v2 0.65 1024 1024 Chinese No
stella-base-zh-v2 0.2 768 1024 Chinese No
stella-large-zh 0.65 1024 1024 Chinese Yes
stella-base-zh 0.2 768 1024 Chinese Yes

The training data mainly includes:

  1. Open-source training data (wudao_base_200GB, m3e, and simclue), with a focus on selecting texts with lengths greater than 512.
  2. A batch of (question, paragraph) and (sentence, paragraph) data constructed on a general corpus using LLM.

The loss functions mainly include:

  1. Contrastive learning loss function
  2. Contrastive learning loss function with hard negative examples (based on bm25 and vector hard negatives)
  3. EWC (Elastic Weights Consolidation)
  4. cosent loss

Model weight initialization:
stella-base-zh and stella-large-zh use piccolo-base-zh and piccolo-large-zh as the base models, respectively, and the 512-1024 position embedding uses the initialization strategy of hierarchical decomposed position encoding.

Training strategy:
One iterator for each type of data, separately calculating the loss.

Based on stella models, stella-v2 use more training data and remove instruction by Knowledge Distillation.

Metric

C-MTEB leaderboard (Chinese)

Model Name Model Size (GB) Dimension Sequence Length Average (35) Classification (9) Clustering (4) Pair Classification (2) Reranking (4) Retrieval (8) STS (8)
stella-large-zh-v2 0.65 1024 1024 65.13 69.05 49.16 82.68 66.41 70.14 58.66
stella-base-zh-v2 0.2 768 1024 64.36 68.29 49.4 79.95 66.1 70.08 56.92
stella-large-zh 0.65 1024 1024 64.54 67.62 48.65 78.72 65.98 71.02 58.3
stella-base-zh 0.2 768 1024 64.16 67.77 48.7 76.09 66.95 71.07 56.54

MTEB leaderboard (English)

Model Name Model Size (GB) Dimension Sequence Length Average (56) Classification (12) Clustering (11) Pair Classification (3) Reranking (4) Retrieval (15) STS (10) Summarization (1)
stella-base-en-v2 0.2 768 512 62.61 75.28 44.9 86.45 58.77 50.1 83.02 32.52

Reproduce our results

C-MTEB:

import torch
import numpy as np
from typing import List
from mteb import MTEB
from sentence_transformers import SentenceTransformer


class FastTextEncoder():
    def __init__(self, model_name):
        self.model = SentenceTransformer(model_name).cuda().half().eval()
        self.model.max_seq_length = 512

    def encode(
            self,
            input_texts: List[str],
            *args,
            **kwargs
    ):
        new_sens = list(set(input_texts))
        new_sens.sort(key=lambda x: len(x), reverse=True)
        vecs = self.model.encode(
            new_sens, normalize_embeddings=True, convert_to_numpy=True, batch_size=256
        ).astype(np.float32)
        sen2arrid = {sen: idx for idx, sen in enumerate(new_sens)}
        vecs = vecs[[sen2arrid[sen] for sen in input_texts]]
        torch.cuda.empty_cache()
        return vecs


if __name__ == '__main__':
    model_name = "infgrad/stella-base-zh-v2"
    output_folder = "zh_mteb_results/stella-base-zh-v2"
    task_names = [t.description["name"] for t in MTEB(task_langs=['zh', 'zh-CN']).tasks]
    model = FastTextEncoder(model_name)
    for task in task_names:
        MTEB(tasks=[task], task_langs=['zh', 'zh-CN']).run(model, output_folder=output_folder)

MTEB:

You can use official script to reproduce our result. scripts/run_mteb_english.py

Evaluation for long text

经过实际观察发现,C-MTEB的评测数据长度基本都是小于512的, 更致命的是那些长度大于512的文本,其重点都在前半部分 这里以CMRC2018的数据为例说明这个问题:

question: 《无双大蛇z》是谁旗下ω-force开发的动作游戏?

passage:《无双大蛇z》是光荣旗下ω-force开发的动作游戏,于2009年3月12日登陆索尼playstation3,并于2009年11月27日推......

passage长度为800多,大于512,但是对于这个question而言只需要前面40个字就足以检索,多的内容对于模型而言是一种噪声,反而降低了效果。
简言之,现有数据集的2个问题:
1)长度大于512的过少
2)即便大于512,对于检索而言也只需要前512的文本内容
导致无法准确评估模型的长文本编码能力。

为了解决这个问题,搜集了相关开源数据并使用规则进行过滤,最终整理了6份长文本测试集,他们分别是:

  • CMRC2018,通用百科
  • CAIL,法律阅读理解
  • DRCD,繁体百科,已转简体
  • Military,军工问答
  • Squad,英文阅读理解,已转中文
  • Multifieldqa_zh,清华的大模型长文本理解能力评测数据[9]

处理规则是选取答案在512长度之后的文本,短的测试数据会欠采样一下,长短文本占比约为1:2,所以模型既得理解短文本也得理解长文本。 除了Military数据集,我们提供了其他5个测试数据的下载地址:https://drive.google.com/file/d/1WC6EWaCbVgz-vPMDFH4TwAMkLyh5WNcN/view?usp=sharing

评测指标为Recall@5, 结果如下:

Dataset piccolo-base-zh piccolo-large-zh bge-base-zh bge-large-zh stella-base-zh stella-large-zh
CMRC2018 94.34 93.82 91.56 93.12 96.08 95.56
CAIL 28.04 33.64 31.22 33.94 34.62 37.18
DRCD 78.25 77.9 78.34 80.26 86.14 84.58
Military 76.61 73.06 75.65 75.81 83.71 80.48
Squad 91.21 86.61 87.87 90.38 93.31 91.21
Multifieldqa_zh 81.41 83.92 83.92 83.42 79.9 80.4
Average 74.98 74.83 74.76 76.15 78.96 78.24

注意: 因为长文本评测数据数量稀少,所以构造时也使用了train部分,如果自行评测,请注意模型的训练数据以免数据泄露。

Usage

stella 中文系列模型

stella-base-zh 和 stella-large-zh: 本模型是在piccolo基础上训练的,因此用法和piccolo完全一致 ,即在检索重排任务上给query和passage加上查询: 结果: 。对于短短匹配不需要做任何操作。

stella-base-zh-v2 和 stella-large-zh-v2: 本模型使用简单,任何使用场景中都不需要加前缀文本

stella中文系列模型均使用mean pooling做为文本向量。

在sentence-transformer库中的使用方法:

from sentence_transformers import SentenceTransformer

sentences = ["数据1", "数据2"]
model = SentenceTransformer('infgrad/stella-base-zh-v2')
print(model.max_seq_length)
embeddings_1 = model.encode(sentences, normalize_embeddings=True)
embeddings_2 = model.encode(sentences, normalize_embeddings=True)
similarity = embeddings_1 @ embeddings_2.T
print(similarity)

直接使用transformers库:

from transformers import AutoModel, AutoTokenizer
from sklearn.preprocessing import normalize

model = AutoModel.from_pretrained('infgrad/stella-base-zh-v2')
tokenizer = AutoTokenizer.from_pretrained('infgrad/stella-base-zh-v2')
sentences = ["数据1", "数据ABCDEFGH"]
batch_data = tokenizer(
    batch_text_or_text_pairs=sentences,
    padding="longest",
    return_tensors="pt",
    max_length=1024,
    truncation=True,
)
attention_mask = batch_data["attention_mask"]
model_output = model(**batch_data)
last_hidden = model_output.last_hidden_state.masked_fill(~attention_mask[..., None].bool(), 0.0)
vectors = last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]
vectors = normalize(vectors, norm="l2", axis=1, )
print(vectors.shape)  # 2,768

stella models for English

Using Sentence-Transformers:

from sentence_transformers import SentenceTransformer

sentences = ["one car come", "one car go"]
model = SentenceTransformer('infgrad/stella-base-en-v2')
print(model.max_seq_length)
embeddings_1 = model.encode(sentences, normalize_embeddings=True)
embeddings_2 = model.encode(sentences, normalize_embeddings=True)
similarity = embeddings_1 @ embeddings_2.T
print(similarity)

Using HuggingFace Transformers:

from transformers import AutoModel, AutoTokenizer
from sklearn.preprocessing import normalize

model = AutoModel.from_pretrained('infgrad/stella-base-en-v2')
tokenizer = AutoTokenizer.from_pretrained('infgrad/stella-base-en-v2')
sentences = ["one car come", "one car go"]
batch_data = tokenizer(
    batch_text_or_text_pairs=sentences,
    padding="longest",
    return_tensors="pt",
    max_length=512,
    truncation=True,
)
attention_mask = batch_data["attention_mask"]
model_output = model(**batch_data)
last_hidden = model_output.last_hidden_state.masked_fill(~attention_mask[..., None].bool(), 0.0)
vectors = last_hidden.sum(dim=1) / attention_mask.sum(dim=1)[..., None]
vectors = normalize(vectors, norm="l2", axis=1, )
print(vectors.shape)  # 2,768

Training Detail

硬件: 单卡A100-80GB

环境: torch1.13.*; transformers-trainer + deepspeed + gradient-checkpointing

学习率: 1e-6

batch_size: base模型为1024,额外增加20%的难负例;large模型为768,额外增加20%的难负例

数据量: 第一版模型约100万,其中用LLM构造的数据约有200K. LLM模型大小为13b。v2系列模型到了2000万训练数据。

ToDoList

评测的稳定性: 评测过程中发现Clustering任务会和官方的结果不一致,大约有±0.0x的小差距,原因是聚类代码没有设置random_seed,差距可以忽略不计,不影响评测结论。

更高质量的长文本训练和测试数据: 训练数据多是用13b模型构造的,肯定会存在噪声。 测试数据基本都是从mrc数据整理来的,所以问题都是factoid类型,不符合真实分布。

OOD的性能: 虽然近期出现了很多向量编码模型,但是对于不是那么通用的domain,这一众模型包括stella、openai和cohere, 它们的效果均比不上BM25。

Reference

  1. https://www.scidb.cn/en/detail?dataSetId=c6a3fe684227415a9db8e21bac4a15ab
  2. https://github.com/wangyuxinwhy/uniem
  3. https://github.com/CLUEbenchmark/SimCLUE
  4. https://arxiv.org/abs/1612.00796
  5. https://kexue.fm/archives/8847
  6. https://huggingface.co/sensenova/piccolo-base-zh
  7. https://kexue.fm/archives/7947
  8. https://github.com/FlagOpen/FlagEmbedding
  9. https://github.com/THUDM/LongBench