pipeline_tag: 句子相似度 datasets:

• ms_marco
• sentence-transformers/msmarco-hard-negatives metrics:
• 召回率 tags:
• colbert
• 段落检索 library_name: colbert-ai base_model: facebook/xmod-base inference: false license: mit model-index:
• name: colbert-xm results:
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-ar config: 阿拉伯语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 74.8
    • type: recall_at_500 name: Recall@500 value: 72.1
    • type: recall_at_100 name: Recall@100 value: 60.4
    • type: recall_at_10 name: Recall@10 value: 36.5
    • type: mrr_at_10 name: MRR@10 value: 19.5
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-de config: 德语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 86.0
    • type: recall_at_500 name: Recall@500 value: 84.1
    • type: recall_at_100 name: Recall@100 value: 73.9
    • type: recall_at_10 name: Recall@10 value: 49.5
    • type: mrr_at_10 name: MRR@10 value: 27.0
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-en config: 英语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 96.5
    • type: recall_at_500 name: Recall@500 value: 95.9
    • type: recall_at_100 name: Recall@100 value: 89.3
    • type: recall_at_10 name: Recall@10 value: 65.7
    • type: mrr_at_10 name: MRR@10 value: 37.2
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-es config: 西班牙语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 88.4
    • type: recall_at_500 name: Recall@500 value: 86.8
    • type: recall_at_100 name: Recall@100 value: 77.5
    • type: recall_at_10 name: Recall@10 value: 52.0
    • type: mrr_at_10 name: MRR@10 value: 28.5
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-fr config: 法语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 87.3
    • type: recall_at_500 name: Recall@500 value: 85.7
    • type: recall_at_100 name: Recall@100 value: 75.2
    • type: recall_at_10 name: Recall@10 value: 49.2
    • type: mrr_at_10 name: MRR@10 value: 26.9
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-hi config: 印地语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 82.2
    • type: recall_at_500 name: Recall@500 value: 79.9
    • type: recall_at_100 name: Recall@100 value: 69.8
    • type: recall_at_10 name: Recall@10 value: 44.2
    • type: mrr_at_10 name: MRR@10 value: 23.8
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-id config: 印尼语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 86.7
    • type: recall_at_500 name: Recall@500 value: 84.8
    • type: recall_at_100 name: Recall@100 value: 74.5
    • type: recall_at_10 name: Recall@10 value: 48.3
    • type: mrr_at_10 name: MRR@10 value: 26.3
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-it config: 意大利语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 86.1
    • type: recall_at_500 name: Recall@500 value: 84.3
    • type: recall_at_100 name: Recall@100 value: 74.1
    • type: recall_at_10 name: Recall@10 value: 48.2
    • type: mrr_at_10 name: MRR@10 value: 26.5
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-ja config: 日语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 83.6
    • type: recall_at_500 name: Recall@500 value: 81.8
    • type: recall_at_100 name: Recall@100 value: 71.4
    • type: recall_at_10 name: Recall@10 value: 44.6
    • type: mrr_at_10 name: MRR@10 value: 24.1
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-nl config: 荷兰语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 86.8
    • type: recall_at_500 name: Recall@500 value: 85.0
    • type: recall_at_100 name: Recall@100 value: 75.2
    • type: recall_at_10 name: Recall@10 value: 49.8
    • type: mrr_at_10 name: MRR@10 value: 27.5
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-pt config: 葡萄牙语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 87.1
    • type: recall_at_500 name: Recall@500 value: 85.3
    • type: recall_at_100 name: Recall@100 value: 75.8
    • type: recall_at_10 name: Recall@10 value: 50.5
    • type: mrr_at_10 name: MRR@10 value: 27.6
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-ru config: 俄语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 85.7
    • type: recall_at_500 name: Recall@500 value: 83.8
    • type: recall_at_100 name: Recall@100 value: 73.6
    • type: recall_at_10 name: Recall@10 value: 47.3
    • type: mrr_at_10 name: MRR@10 value: 25.1
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-vi config: 越南语 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 81.6
    • type: recall_at_500 name: Recall@500 value: 79.0
    • type: recall_at_100 name: Recall@100 value: 67.5
    • type: recall_at_10 name: Recall@10 value: 42.4
    • type: mrr_at_10 name: MRR@10 value: 22.6
  • task: type: 句子相似度 name: 段落检索 dataset: type: unicamp-dl/mmarco name: mMARCO-zh config: 中文 split: 验证集 metrics:
    • type: recall_at_1000 name: Recall@1000 value: 84.8
    • type: recall_at_500 name: Recall@500 value: 83.1
    • type: recall_at_100 name: Recall@100 value: 72.2
    • type: recall_at_10 name: Recall@10 value: 46.0
    • type: mrr_at_10 name: MRR@10 value: 24.6 language:
• 多语言
• 南非荷兰语
• 阿姆哈拉语
• 阿拉伯语
• 阿塞拜疆语
• 白俄罗斯语
• 保加利亚语
• 孟加拉语
• 加泰罗尼亚语
• 捷克语
• 威尔士语
• 丹麦语
• 德语
• 希腊语
• 英语
• 世界语
• 西班牙语
• 爱沙尼亚语
• 巴斯克语
• 波斯语
• 芬兰语
• 法语
• 爱尔兰语
• 加利西亚语
• 古吉拉特语
• 豪萨语
• 希伯来语
• 印地语
• 克罗地亚语
• 匈牙利语
• 亚美尼亚语
• 印尼语
• 冰岛语
• 意大利语
• 日语
• 格鲁吉亚语
• 哈萨克语
• 高棉语
• 卡纳达语
• 韩语
• 库尔德语
• 吉尔吉斯语
• 拉丁语
• 老挝语
• 立陶宛语
• 拉脱维亚语
• 马其顿语
• 马拉雅拉姆语
• 蒙古语
• 马拉地语
• 马来语
• 缅甸语
• 尼泊尔语
• 荷兰语
• 挪威语
• 奥里亚语
• 旁遮普语
• 波兰语
• 普什图语
• 葡萄牙语
• 罗马尼亚