---

tags:

• mteb

base_model: mixedbread-ai/mxbai-embed-mini-v1 library_name: sentence-transformers model-index:

• name: mxbai-embed-xsmall-v1

results: - task: type: Retrieval dataset: type: arguana name: MTEB ArguAna config: default split: test revision: None metrics: - type: ndcgat1 value: 25.18 - type: ndcgat3 value: 39.22 - type: ndcgat5 value: 43.93 - type: ndcgat10 value: 49.58 - type: ndcgat30 value: 53.41 - type: ndcgat100 value: 54.11 - type: mapat1 value: 25.18 - type: mapat3 value: 35.66 - type: mapat5 value: 38.25 - type: mapat10 value: 40.58 - type: mapat30 value: 41.6 - type: mapat100 value: 41.69 - type: recallat1 value: 25.18 - type: recallat3 value: 49.57 - type: recallat5 value: 61.09 - type: recallat10 value: 78.59 - type: recallat30 value: 94.03 - type: recallat100 value: 97.94 - type: precisionat1 value: 25.18 - type: precisionat3 value: 16.52 - type: precisionat5 value: 12.22 - type: precisionat10 value: 7.86 - type: precisionat30 value: 3.13 - type: precisionat100 value: 0.98 - type: accuracyat3 value: 49.57 - type: accuracyat5 value: 61.09 - type: accuracyat10 value: 78.59 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackAndroidRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 44.35 - type: ndcgat3 value: 49.64 - type: ndcgat5 value: 51.73 - type: ndcgat10 value: 54.82 - type: ndcgat30 value: 57.64 - type: ndcgat100 value: 59.77 - type: mapat1 value: 36.26 - type: mapat3 value: 44.35 - type: mapat5 value: 46.26 - type: mapat10 value: 48.24 - type: mapat30 value: 49.34 - type: mapat100 value: 49.75 - type: recallat1 value: 36.26 - type: recallat3 value: 51.46 - type: recallat5 value: 57.78 - type: recallat10 value: 66.5 - type: recallat30 value: 77.19 - type: recallat100 value: 87.53 - type: precisionat1 value: 44.35 - type: precisionat3 value: 23.65 - type: precisionat5 value: 16.88 - type: precisionat10 value: 10.7 - type: precisionat30 value: 4.53 - type: precisionat100 value: 1.65 - type: accuracyat3 value: 60.51 - type: accuracyat5 value: 67.67 - type: accuracyat10 value: 74.68 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackEnglishRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 39.43 - type: ndcgat3 value: 44.13 - type: ndcgat5 value: 46.06 - type: ndcgat10 value: 48.31 - type: ndcgat30 value: 51.06 - type: ndcgat100 value: 53.07 - type: mapat1 value: 31.27 - type: mapat3 value: 39.07 - type: mapat5 value: 40.83 - type: mapat10 value: 42.23 - type: mapat30 value: 43.27 - type: mapat100 value: 43.66 - type: recallat1 value: 31.27 - type: recallat3 value: 45.89 - type: recallat5 value: 51.44 - type: recallat10 value: 58.65 - type: recallat30 value: 69.12 - type: recallat100 value: 78.72 - type: precisionat1 value: 39.43 - type: precisionat3 value: 21.61 - type: precisionat5 value: 15.34 - type: precisionat10 value: 9.27 - type: precisionat30 value: 4.01 - type: precisionat100 value: 1.52 - type: accuracyat3 value: 55.48 - type: accuracyat5 value: 60.76 - type: accuracyat10 value: 67.45 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackGamingRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 45.58 - type: ndcgat3 value: 52.68 - type: ndcgat5 value: 55.28 - type: ndcgat10 value: 57.88 - type: ndcgat30 value: 60.6 - type: ndcgat100 value: 62.03 - type: mapat1 value: 39.97 - type: mapat3 value: 49.06 - type: mapat5 value: 50.87 - type: mapat10 value: 52.2 - type: mapat30 value: 53.06 - type: mapat100 value: 53.28 - type: recallat1 value: 39.97 - type: recallat3 value: 57.4 - type: recallat5 value: 63.83 - type: recallat10 value: 71.33 - type: recallat30 value: 81.81 - type: recallat100 value: 89.0 - type: precisionat1 value: 45.58 - type: precisionat3 value: 23.55 - type: precisionat5 value: 16.01 - type: precisionat10 value: 9.25 - type: precisionat30 value: 3.67 - type: precisionat100 value: 1.23 - type: accuracyat3 value: 62.76 - type: accuracyat5 value: 68.84 - type: accuracyat10 value: 75.8 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackGisRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 27.35 - type: ndcgat3 value: 34.23 - type: ndcgat5 value: 37.1 - type: ndcgat10 value: 40.26 - type: ndcgat30 value: 43.54 - type: ndcgat100 value: 45.9 - type: mapat1 value: 25.28 - type: mapat3 value: 31.68 - type: mapat5 value: 33.38 - type: mapat10 value: 34.79 - type: mapat30 value: 35.67 - type: mapat100 value: 35.96 - type: recallat1 value: 25.28 - type: recallat3 value: 38.95 - type: recallat5 value: 45.82 - type: recallat10 value: 55.11 - type: recallat30 value: 68.13 - type: recallat100 value: 80.88 - type: precisionat1 value: 27.35 - type: precisionat3 value: 14.65 - type: precisionat5 value: 10.44 - type: precisionat10 value: 6.37 - type: precisionat30 value: 2.65 - type: precisionat100 value: 0.97 - type: accuracyat3 value: 42.15 - type: accuracyat5 value: 49.15 - type: accuracyat10 value: 58.53 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackMathematicaRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 18.91 - type: ndcgat3 value: 24.37 - type: ndcgat5 value: 26.11 - type: ndcgat10 value: 29.37 - type: ndcgat30 value: 33.22 - type: ndcgat100 value: 35.73 - type: mapat1 value: 15.23 - type: mapat3 value: 21.25 - type: mapat5 value: 22.38 - type: mapat10 value: 23.86 - type: mapat30 value: 24.91 - type: mapat100 value: 25.24 - type: recallat1 value: 15.23 - type: recallat3 value: 28.28 - type: recallat5 value: 32.67 - type: recallat10 value: 42.23 - type: recallat30 value: 56.87 - type: recallat100 value: 69.44 - type: precisionat1 value: 18.91 - type: precisionat3 value: 11.9 - type: precisionat5 value: 8.48 - type: precisionat10 value: 5.63 - type: precisionat30 value: 2.64 - type: precisionat100 value: 1.02 - type: accuracyat3 value: 33.95 - type: accuracyat5 value: 38.81 - type: accuracyat10 value: 49.13 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackPhysicsRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 36.96 - type: ndcgat3 value: 42.48 - type: ndcgat5 value: 44.57 - type: ndcgat10 value: 47.13 - type: ndcgat30 value: 50.65 - type: ndcgat100 value: 53.14 - type: mapat1 value: 30.1 - type: mapat3 value: 37.97 - type: mapat5 value: 39.62 - type: mapat10 value: 41.06 - type: mapat30 value: 42.13 - type: mapat100 value: 42.53 - type: recallat1 value: 30.1 - type: recallat3 value: 45.98 - type: recallat5 value: 51.58 - type: recallat10 value: 59.24 - type: recallat30 value: 72.47 - type: recallat100 value: 84.53 - type: precisionat1 value: 36.96 - type: precisionat3 value: 20.5 - type: precisionat5 value: 14.4 - type: precisionat10 value: 8.62 - type: precisionat30 value: 3.67 - type: precisionat100 value: 1.38 - type: accuracyat3 value: 54.09 - type: accuracyat5 value: 60.25 - type: accuracyat10 value: 67.37 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackProgrammersRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 28.65 - type: ndcgat3 value: 34.3 - type: ndcgat5 value: 36.8 - type: ndcgat10 value: 39.92 - type: ndcgat30 value: 42.97 - type: ndcgat100 value: 45.45 - type: mapat1 value: 23.35 - type: mapat3 value: 30.36 - type: mapat5 value: 32.15 - type: mapat10 value: 33.74 - type: mapat30 value: 34.69 - type: mapat100 value: 35.02 - type: recallat1 value: 23.35 - type: recallat3 value: 37.71 - type: recallat5 value: 44.23 - type: recallat10 value: 53.6 - type: recallat30 value: 64.69 - type: recallat100 value: 77.41 - type: precisionat1 value: 28.65 - type: precisionat3 value: 16.74 - type: precisionat5 value: 12.21 - type: precisionat10 value: 7.61 - type: precisionat30 value: 3.29 - type: precisionat100 value: 1.22 - type: accuracyat3 value: 44.86 - type: accuracyat5 value: 52.4 - type: accuracyat10 value: 61.07 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackStatsRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 26.07 - type: ndcgat3 value: 31.62 - type: ndcgat5 value: 33.23 - type: ndcgat10 value: 35.62 - type: ndcgat30 value: 38.41 - type: ndcgat100 value: 40.81 - type: mapat1 value: 22.96 - type: mapat3 value: 28.85 - type: mapat5 value: 29.97 - type: mapat10 value: 31.11 - type: mapat30 value: 31.86 - type: mapat100 value: 32.15 - type: recallat1 value: 22.96 - type: recallat3 value: 35.14 - type: recallat5 value: 39.22 - type: recallat10 value: 46.52 - type: recallat30 value: 57.58 - type: recallat100 value: 70.57 - type: precisionat1 value: 26.07 - type: precisionat3 value: 14.11 - type: precisionat5 value: 9.69 - type: precisionat10 value: 5.81 - type: precisionat30 value: 2.45 - type: precisionat100 value: 0.92 - type: accuracyat3 value: 39.42 - type: accuracyat5 value: 43.41 - type: accuracyat10 value: 50.92 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackTexRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 21.78 - type: ndcgat3 value: 25.74 - type: ndcgat5 value: 27.86 - type: ndcgat10 value: 30.3 - type: ndcgat30 value: 33.51 - type: ndcgat100 value: 36.12 - type: mapat1 value: 17.63 - type: mapat3 value: 22.7 - type: mapat5 value: 24.14 - type: mapat10 value: 25.31 - type: mapat30 value: 26.22 - type: mapat100 value: 26.56 - type: recallat1 value: 17.63 - type: recallat3 value: 28.37 - type: recallat5 value: 33.99 - type: recallat10 value: 41.23 - type: recallat30 value: 53.69 - type: recallat100 value: 67.27 - type: precisionat1 value: 21.78 - type: precisionat3 value: 12.41 - type: precisionat5 value: 9.07 - type: precisionat10 value: 5.69 - type: precisionat30 value: 2.61 - type: precisionat100 value: 1.03 - type: accuracyat3 value: 33.62 - type: accuracyat5 value: 39.81 - type: accuracyat10 value: 47.32 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackUnixRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 30.97 - type: ndcgat3 value: 36.13 - type: ndcgat5 value: 39.0 - type: ndcgat10 value: 41.78 - type: ndcgat30 value: 44.96 - type: ndcgat100 value: 47.52 - type: mapat1 value: 26.05 - type: mapat3 value: 32.77 - type: mapat5 value: 34.6 - type: mapat10 value: 35.93 - type: mapat30 value: 36.88 - type: mapat100 value: 37.22 - type: recallat1 value: 26.05 - type: recallat3 value: 40.0 - type: recallat5 value: 47.34 - type: recallat10 value: 55.34 - type: recallat30 value: 67.08 - type: recallat100 value: 80.2 - type: precisionat1 value: 30.97 - type: precisionat3 value: 16.6 - type: precisionat5 value: 12.03 - type: precisionat10 value: 7.3 - type: precisionat30 value: 3.08 - type: precisionat100 value: 1.15 - type: accuracyat3 value: 45.62 - type: accuracyat5 value: 53.64 - type: accuracyat10 value: 61.66 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackWebmastersRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 29.64 - type: ndcgat3 value: 35.49 - type: ndcgat5 value: 37.77 - type: ndcgat10 value: 40.78 - type: ndcgat30 value: 44.59 - type: ndcgat100 value: 46.97 - type: mapat1 value: 24.77 - type: mapat3 value: 31.33 - type: mapat5 value: 32.95 - type: mapat10 value: 34.47 - type: mapat30 value: 35.7 - type: mapat100 value: 36.17 - type: recallat1 value: 24.77 - type: recallat3 value: 38.16 - type: recallat5 value: 44.1 - type: recallat10 value: 53.31 - type: recallat30 value: 68.43 - type: recallat100 value: 80.24 - type: precisionat1 value: 29.64 - type: precisionat3 value: 16.8 - type: precisionat5 value: 12.21 - type: precisionat10 value: 7.83 - type: precisionat30 value: 3.89 - type: precisionat100 value: 1.63 - type: accuracyat3 value: 45.45 - type: accuracyat5 value: 51.58 - type: accuracyat10 value: 61.07 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackWordpressRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 23.47 - type: ndcgat3 value: 27.98 - type: ndcgat5 value: 30.16 - type: ndcgat10 value: 32.97 - type: ndcgat30 value: 36.3 - type: ndcgat100 value: 38.47 - type: mapat1 value: 21.63 - type: mapat3 value: 26.02 - type: mapat5 value: 27.32 - type: mapat10 value: 28.51 - type: mapat30 value: 29.39 - type: mapat100 value: 29.66 - type: recallat1 value: 21.63 - type: recallat3 value: 31.47 - type: recallat5 value: 36.69 - type: recallat10 value: 44.95 - type: recallat30 value: 58.2 - type: recallat100 value: 69.83 - type: precisionat1 value: 23.47 - type: precisionat3 value: 11.71 - type: precisionat5 value: 8.32 - type: precisionat10 value: 5.23 - type: precisionat30 value: 2.29 - type: precisionat100 value: 0.86 - type: accuracyat3 value: 34.01 - type: accuracyat5 value: 39.37 - type: accuracyat10 value: 48.24 - task: type: Retrieval dataset: type: climate-fever name: MTEB ClimateFEVER config: default split: test revision: None metrics: - type: ndcgat1 value: 19.8 - type: ndcgat3 value: 17.93 - type: ndcgat5 value: 19.39 - type: ndcgat10 value: 22.42 - type: ndcgat30 value: 26.79 - type: ndcgat100 value: 29.84 - type: mapat1 value: 9.09 - type: mapat3 value: 12.91 - type: mapat5 value: 14.12 - type: mapat10 value: 15.45 - type: mapat30 value: 16.73 - type: mapat100 value: 17.21 - type: recallat1 value: 9.09 - type: recallat3 value: 16.81 - type: recallat5 value: 20.9 - type: recallat10 value: 27.65 - type: recallat30 value: 41.23 - type: recallat100 value: 53.57 - type: precisionat1 value: 19.8 - type: precisionat3 value: 13.36 - type: precisionat5 value: 10.33 - type: precisionat10 value: 7.15 - type: precisionat30 value: 3.66 - type: precisionat100 value: 1.49 - type: accuracyat3 value: 36.22 - type: accuracyat5 value: 44.1 - type: accuracyat10 value: 55.11 - task: type: Retrieval dataset: type: dbpedia-entity name: MTEB DBPedia config: default split: test revision: None metrics: - type: ndcgat1 value: 42.75 - type: ndcgat3 value: 35.67 - type: ndcgat5 value: 33.58 - type: ndcgat10 value: 32.19 - type: ndcgat30 value: 31.82 - type: ndcgat100 value: 35.87 - type: mapat1 value: 7.05 - type: mapat3 value: 10.5 - type: mapat5 value: 12.06 - type: mapat10 value: 14.29 - type: mapat30 value: 17.38 - type: mapat100 value: 19.58 - type: recallat1 value: 7.05 - type: recallat3 value: 11.89 - type: recallat5 value: 14.7 - type: recallat10 value: 19.78 - type: recallat30 value: 29.88 - type: recallat100 value: 42.4 - type: precisionat1 value: 54.25 - type: precisionat3 value: 39.42 - type: precisionat5 value: 33.15 - type: precisionat10 value: 25.95 - type: precisionat30 value: 15.51 - type: precisionat100 value: 7.9 - type: accuracyat3 value: 72.0 - type: accuracyat5 value: 77.75 - type: accuracyat10 value: 83.5 - task: type: Retrieval dataset: type: fever name: MTEB FEVER config: default split: test revision: None metrics: - type: ndcgat1 value: 40.19 - type: ndcgat3 value: 50.51 - type: ndcgat5 value: 53.51 - type: ndcgat10 value: 56.45 - type: ndcgat30 value: 58.74 - type: ndcgat100 value: 59.72 - type: mapat1 value: 37.56 - type: mapat3 value: 46.74 - type: mapat5 value: 48.46 - type: mapat10 value: 49.7 - type: mapat30 value: 50.31 - type: mapat100 value: 50.43 - type: recallat1 value: 37.56 - type: recallat3 value: 58.28 - type: recallat5 value: 65.45 - type: recallat10 value: 74.28 - type: recallat30 value: 83.42 - type: recallat100 value: 88.76 - type: precisionat1 value: 40.19 - type: precisionat3 value: 20.99 - type: precisionat5 value: 14.24 - type: precisionat10 value: 8.12 - type: precisionat30 value: 3.06 - type: precisionat100 value: 0.98 - type: accuracyat3 value: 62.3 - type: accuracyat5 value: 69.94 - type: accuracyat10 value: 79.13 - task: type: Retrieval dataset: type: fiqa name: MTEB FiQA2018 config: default split: test revision: None metrics: - type: ndcgat1 value: 34.41 - type: ndcgat3 value: 33.2 - type: ndcgat5 value: 34.71 - type: ndcgat10 value: 37.1 - type: ndcgat30 value: 40.88 - type: ndcgat100 value: 44.12 - type: mapat1 value: 17.27 - type: mapat3 value: 25.36 - type: mapat5 value: 27.76 - type: mapat10 value: 29.46 - type: mapat30 value: 30.74 - type: mapat100 value: 31.29 - type: recallat1 value: 17.27 - type: recallat3 value: 30.46 - type: recallat5 value: 36.91 - type: recallat10 value: 44.47 - type: recallat30 value: 56.71 - type: recallat100 value: 70.72 - type: precisionat1 value: 34.41 - type: precisionat3 value: 22.32 - type: precisionat5 value: 16.91 - type: precisionat10 value: 10.53 - type: precisionat30 value: 4.62 - type: precisionat100 value: 1.79 - type: accuracyat3 value: 50.77 - type: accuracyat5 value: 57.56 - type: accuracyat10 value: 65.12 - task: type: Retrieval dataset: type: hotpotqa name: MTEB HotpotQA config: default split: test revision: None metrics: - type: ndcgat1 value: 57.93 - type: ndcgat3 value: 44.21 - type: ndcgat5 value: 46.4 - type: ndcgat10 value: 48.37 - type: ndcgat30 value: 50.44 - type: ndcgat100 value: 51.86 - type: mapat1 value: 28.97 - type: mapat3 value: 36.79 - type: mapat5 value: 38.31 - type: mapat10 value: 39.32 - type: mapat30 value: 39.99 - type: mapat100 value: 40.2 - type: recallat1 value: 28.97 - type: recallat3 value: 41.01 - type: recallat5 value: 45.36 - type: recallat10 value: 50.32 - type: recallat30 value: 57.38 - type: recallat100 value: 64.06 - type: precisionat1 value: 57.93 - type: precisionat3 value: 27.34 - type: precisionat5 value: 18.14 - type: precisionat10 value: 10.06 - type: precisionat30 value: 3.82 - type: precisionat100 value: 1.28 - type: accuracyat3 value: 71.03 - type: accuracyat5 value: 75.14 - type: accuracyat10 value: 79.84 - task: type: Retrieval dataset: type: msmarco name: MTEB MSMARCO config: default split: dev revision: None metrics: - type: ndcgat1 value: 19.74 - type: ndcgat3 value: 29.47 - type: ndcgat5 value: 32.99 - type: ndcgat10 value: 36.76 - type: ndcgat30 value: 40.52 - type: ndcgat100 value: 42.78 - type: mapat1 value: 19.2 - type: mapat3 value: 26.81 - type: mapat5 value: 28.78 - type: mapat10 value: 30.35 - type: mapat30 value: 31.3 - type: mapat100 value: 31.57 - type: recallat1 value: 19.2 - type: recallat3 value: 36.59 - type: recallat5 value: 45.08 - type: recallat10 value: 56.54 - type: recallat30 value: 72.05 - type: recallat100 value: 84.73 - type: precisionat1 value: 19.74 - type: precisionat3 value: 12.61 - type: precisionat5 value: 9.37 - type: precisionat10 value: 5.89 - type: precisionat30 value: 2.52 - type: precisionat100 value: 0.89 - type: accuracyat3 value: 37.38 - type: accuracyat5 value: 46.06 - type: accuracyat10 value: 57.62 - task: type: Retrieval dataset: type: nq name: MTEB NQ config: default split: test revision: None metrics: - type: ndcgat1 value: 25.9 - type: ndcgat3 value: 35.97 - type: ndcgat5 value: 40.27 - type: ndcgat10 value: 44.44 - type: ndcgat30 value: 48.31 - type: ndcgat100 value: 50.14 - type: mapat1 value: 23.03 - type: mapat3 value: 32.45 - type: mapat5 value: 34.99 - type: mapat10 value: 36.84 - type: mapat30 value: 37.92 - type: mapat100 value: 38.16 - type: recallat1 value: 23.03 - type: recallat3 value: 43.49 - type: recallat5 value: 53.41 - type: recallat10 value: 65.65 - type: recallat30 value: 80.79 - type: recallat100 value: 90.59 - type: precisionat1 value: 25.9 - type: precisionat3 value: 16.76 - type: precisionat5 value: 12.54 - type: precisionat10 value: 7.78 - type: precisionat30 value: 3.23 - type: precisionat100 value: 1.1 - type: accuracyat3 value: 47.31 - type: accuracyat5 value: 57.16 - type: accuracyat10 value: 69.09 - task: type: Retrieval dataset: type: nfcorpus name: MTEB NFCorpus config: default split: test revision: None metrics: - type: ndcgat1 value: 40.87 - type: ndcgat3 value: 36.79 - type: ndcgat5 value: 34.47 - type: ndcgat10 value: 32.05 - type: ndcgat30 value: 29.23 - type: ndcgat100 value: 29.84 - type: mapat1 value: 5.05 - type: mapat3 value: 8.5 - type: mapat5 value: 9.87 - type: mapat10 value: 11.71 - type: mapat30 value: 13.48 - type: mapat100 value: 14.86 - type: recallat1 value: 5.05 - type: recallat3 value: 9.55 - type: recallat5 value: 11.91 - type: recallat10 value: 16.07 - type: recallat30 value: 22.13 - type: recallat100 value: 30.7 - type: precisionat1 value: 42.72 - type: precisionat3 value: 34.78 - type: precisionat5 value: 30.03 - type: precisionat10 value: 23.93 - type: precisionat30 value: 14.61 - type: precisionat100 value: 7.85 - type: accuracyat3 value: 58.2 - type: accuracyat5 value: 64.09 - type: accuracyat10 value: 69.35 - task: type: Retrieval dataset: type: quora name: MTEB QuoraRetrieval config: default split: test revision: None metrics: - type: ndcgat1 value: 80.62 - type: ndcgat3 value: 84.62 - type: ndcgat5 value: 86.25 - type: ndcgat10 value: 87.7 - type: ndcgat30 value: 88.63 - type: ndcgat100 value: 88.95 - type: mapat1 value: 69.91 - type: mapat3 value: 80.7 - type: mapat5 value: 82.57 - type: mapat10 value: 83.78 - type: mapat30 value: 84.33 - type: mapat100 value: 84.44 - type: recallat1 value: 69.91 - type: recallat3 value: 86.36 - type: recallat5 value: 90.99 - type: recallat10 value: 95.19 - type: recallat30 value: 98.25 - type: recallat100 value: 99.47 - type: precisionat1 value: 80.62 - type: precisionat3 value: 37.03 - type: precisionat5 value: 24.36 - type: precisionat10 value: 13.4 - type: precisionat30 value: 4.87 - type: precisionat100 value: 1.53 - type: accuracyat3 value: 92.25 - type: accuracyat5 value: 95.29 - type: accuracyat10 value: 97.74 - task: type: Retrieval dataset: type: scidocs name: MTEB SCIDOCS config: default split: test revision: None metrics: - type: ndcgat1 value: 24.1 - type: ndcgat3 value: 20.18 - type: ndcgat5 value: 17.72 - type: ndcgat10 value: 21.5 - type: ndcgat30 value: 26.66 - type: ndcgat100 value: 30.95 - type: mapat1 value: 4.88 - type: mapat3 value: 9.09 - type: mapat5 value: 10.99 - type: mapat10 value: 12.93 - type: mapat30 value: 14.71 - type: mapat100 value: 15.49 - type: recallat1 value: 4.88 - type: recallat3 value: 11.55 - type: recallat5 value: 15.91 - type: recallat10 value: 22.82 - type: recallat30 value: 35.7 - type: recallat100 value: 50.41 - type: precisionat1 value: 24.1 - type: precisionat3 value: 19.0 - type: precisionat5 value: 15.72 - type: precisionat10 value: 11.27 - type: precisionat30 value: 5.87 - type: precisionat100 value: 2.49 - type: accuracyat3 value: 43.0 - type: accuracyat5 value: 51.6 - type: accuracyat10 value: 62.7 - task: type: Retrieval dataset: type: scifact name: MTEB SciFact config: default split: test revision: None metrics: - type: ndcgat1 value: 52.33 - type: ndcgat3 value: 61.47 - type: ndcgat5 value: 63.82 - type: ndcgat10 value: 65.81 - type: ndcgat30 value: 67.75 - type: ndcgat100 value: 68.96 - type: mapat1 value: 50.46 - type: mapat3 value: 58.51 - type: mapat5 value: 60.12 - type: mapat10 value: 61.07 - type: mapat30 value: 61.64 - type: mapat100 value: 61.8 - type: recallat1 value: 50.46 - type: recallat3 value: 67.81 - type: recallat5 value: 73.6 - type: recallat10 value: 79.31 - type: recallat30 value: 86.8 - type: recallat100 value: 93.5 - type: precisionat1 value: 52.33 - type: precisionat3 value: 24.56 - type: precisionat5 value: 16.27 - type: precisionat10 value: 8.9 - type: precisionat30 value: 3.28 - type: precisionat100 value: 1.06 - type: accuracyat3 value: 69.67 - type: accuracyat5 value: 75.0 - type: accuracyat10 value: 80.67 - task: type: Retrieval dataset: type: trec-covid name: MTEB TRECCOVID config: default split: test revision: None metrics: - type: ndcgat1 value: 57.0 - type: ndcgat3 value: 53.78 - type: ndcgat5 value: 52.62 - type: ndcgat10 value: 48.9 - type: ndcgat30 value: 44.2 - type: ndcgat100 value: 36.53 - type: mapat1 value: 0.16 - type: mapat3 value: 0.41 - type: mapat5 value: 0.62 - type: mapat10 value: 1.07 - type: mapat30 value: 2.46 - type: mapat100 value: 5.52 - type: recallat1 value: 0.16 - type: recallat3 value: 0.45 - type: recallat5 value: 0.72 - type: recallat10 value: 1.33 - type: recallat30 value: 3.46 - type: recallat100 value: 8.73 - type: precisionat1 value: 62.0 - type: precisionat3 value: 57.33 - type: precisionat5 value: 56.0 - type: precisionat10 value: 52.0 - type: precisionat30 value: 46.2 - type: precisionat100 value: 37.22 - type: accuracyat3 value: 82.0 - type: accuracyat5 value: 90.0 - type: accuracyat10 value: 92.0 - task: type: Retrieval dataset: type: webis-touche2020 name: MTEB Touche2020 config: default split: test revision: None metrics: - type: ndcgat1 value: 20.41 - type: ndcgat3 value: 17.62 - type: ndcgat5 value: 17.16 - type: ndcgat10 value: 17.09 - type: ndcgat30 value: 20.1 - type: ndcgat100 value: 26.33 - type: mapat1 value: 2.15 - type: mapat3 value: 3.59 - type: mapat5 value: 5.07 - type: mapat10 value: 6.95 - type: mapat30 value: 9.01 - type: mapat100 value: 10.54 - type: recallat1 value: 2.15 - type: recallat3 value: 4.5 - type: recallat5 value: 7.54 - type: recallat10 value: 12.46 - type: recallat30 value: 21.9 - type: recallat100 value: 36.58 - type: precisionat1 value: 22.45 - type: precisionat3 value: 19.05 - type: precisionat5 value: 17.55 - type: precisionat10 value: 15.51 - type: precisionat30 value: 10.07 - type: precisionat100 value: 5.57 - type: accuracyat3 value: 42.86 - type: accuracyat5 value: 53.06 - type: accuracyat10 value: 69.39 - task: type: Retrieval dataset: type: BeIR/cqadupstack name: MTEB CQADupstackRetrieval config: default split: test revision: None metrics: - type: ndcgat10 value: 41.59 license: apache-2.0 language:

• en

pipeline_tag: feature-extraction

---

The crispy sentence embedding family from Mixedbread.

^{🍞 Looking for a simple end-to-end retrieval solution? Meet Omni, our multimodal and multilingual model. Get in touch for access.}

mixedbread-ai/mxbai-embed-xsmall-v1

This model is an open-source English embedding model developed by Mixedbread. It's built upon sentence-transformers/all-MiniLM-L6-v2 and trained with the AnglE loss and Espresso. Read more details in our blog post.

In a bread loaf:

• State-of-the-art performance
• Supports both binary quantization and Matryoshka Representation Learning (MRL).
• Optimized for retrieval tasks
• 4096 context support

Performance

Binary Quantization and Matryoshka

Our model supports both binary quantization and Matryoshka Representation Learning (MRL), allowing for significant efficiency gains:

• Binary quantization: Retains 93.9% of performance while increasing efficiency by a factor of 32
• MRL: A 33% reduction in vector size still leaves 96.2% of model performance

These optimizations can lead to substantial reductions in infrastructure costs for cloud computing and vector databases. Read more here.

Quickstart

Here are several ways to produce German sentence embeddings using our model.

angle-emb

pip install -U angle-emb

from angle_emb import AnglE
  from angleemb.utils import cosinesimilarity

# 1. Specify preferred dimensions
  dimensions = 384

# 2. Load model and set pooling strategy to avg
  model = AnglE.from_pretrained(
      "mixedbread-ai/mxbai-embed-xsmall-v1",
      pooling_strategy='avg').cuda()

query = 'A man is eating a piece of bread'

docs = [
      query,
      "A man is eating food.",
      "A man is eating pasta.",
      "The girl is carrying a baby.",
      "A man is riding a horse.",
  ]

# 3. Encode
  embeddings = model.encode(docs, embedding_size=dimensions)

for doc, emb in zip(docs[1:], embeddings[1:]):
      print(f'{query} ||| {doc}', cosine_similarity(embeddings[0], emb))

Sentence Transformers

python -m pip install -U sentence-transformers

from sentence_transformers import SentenceTransformer
  from sentencetransformers.util import cossim

# 1. Specify preferred dimensions
  dimensions = 384

# 2. Load model
  model = SentenceTransformer("mixedbread-ai/mxbai-embed-xsmall-v1", truncate_dim=dimensions)

query = 'A man is eating a piece of bread'

docs = [
      query,
      "A man is eating food.",
      "A man is eating pasta.",
      "The girl is carrying a baby.",
      "A man is riding a horse.",
  ]

# 3. Encode
  embeddings = model.encode(docs)

similarities = cos_sim(embeddings[0], embeddings[1:])
  print('similarities:', similarities)

transformers

pip install -U transformers

from typing import Dict

import torch
  import numpy as np
  from transformers import AutoModel, AutoTokenizer
  from sentencetransformers.util import cossim

def pooling(outputs: torch.Tensor, inputs: Dict) -> np.ndarray:
      outputs = torch.sum(
        outputs * inputs["attentionmask"][:, :, None], dim=1) / torch.sum(inputs["attentionmask"])
      return outputs.detach().cpu().numpy()

# 1. Load model
  model_id = 'mixedbread-ai/mxbai-embed-xsmall-v1'
  tokenizer = AutoTokenizer.frompretrained(modelid)
  model = AutoModel.frompretrained(modelid).cuda()

query = 'A man is eating a piece of bread'

docs = [
      query,
      "A man is eating food.",
      "A man is eating pasta.",
      "The girl is carrying a baby.",
      "A man is riding a horse.",
  ]

# 2. Encode
  inputs = tokenizer(docs, padding=True, return_tensors='pt')
  for k, v in inputs.items():
      inputs[k] = v.cuda()
  outputs = model(inputs).lasthiddenstate
  embeddings = pooling(outputs, inputs)

# 3. Compute similarity scores
  similarities = cos_sim(embeddings[0], embeddings[1:])
  print('similarities:', similarities)

Batched API

python -m pip install batched

import uvicorn
  import batched
  from fastapi import FastAPI
  from fastapi.responses import ORJSONResponse
  from sentence_transformers import SentenceTransformer
  from pydantic import BaseModel

app = FastAPI()

model = SentenceTransformer('mixedbread-ai/mxbai-embed-xsmall-v1')
  model.encode = batched.aio.dynamically(model.encode)

class EmbeddingsRequest(BaseModel):
      input: str | list[str]

@app.post("/embeddings")
  async def embeddings(request: EmbeddingsRequest):
      return ORJSONResponse({"embeddings": await model.encode(request.input)})

if name == "main":
      uvicorn.run(app, host="0.0.0.0", port=8000)

Community

Join our discord community to share your feedback and thoughts. We're here to help and always happy to discuss the exciting field of machine learning!

License

Apache 2.0

Citation

@online{xsmall2024mxbai,
  title={Every Byte Matters: Introducing mxbai-embed-xsmall-v1},
  author={Sean Lee and Julius Lipp and Rui Huang and Darius Koenig},
  year={2024},
  url={https://www.mixedbread.ai/blog/mxbai-embed-xsmall-v1},
}