llama-lang-adapt/XL-Sum-small-v2

Name: llama-lang-adapt/XL-Sum-small-v2
Creator: llama-lang-adapt
Published: 2024-10-09 14:02:00
License: 暂无描述

Hugging Face2024-10-09 更新2025-04-12 收录

下载链接：

https://hf-mirror.com/datasets/llama-lang-adapt/XL-Sum-small-v2

下载链接

链接失效反馈

官方服务：

资源简介：

--- dataset_info: - config_name: amharic features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1569562.0551987502 num_examples: 250 - name: test num_bytes: 1200623.783031989 num_examples: 250 - name: validation num_bytes: 1249771.210013908 num_examples: 250 download_size: 2001658 dataset_size: 4019957.0482446477 - config_name: arabic features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1270655.5545190438 num_examples: 250 - name: test num_bytes: 1135653.71081254 num_examples: 250 - name: validation num_bytes: 1138176.476860738 num_examples: 250 download_size: 1790255 dataset_size: 3544485.742192322 - config_name: azerbaijani features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1149598.9888854586 num_examples: 250 - name: test num_bytes: 980183.2509270705 num_examples: 250 - name: validation num_bytes: 960512.0519159456 num_examples: 250 download_size: 1765693 dataset_size: 3090294.2917284747 - config_name: bengali features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2789469.914835843 num_examples: 250 - name: test num_bytes: 2338806.324110672 num_examples: 250 - name: validation num_bytes: 2277381.175889328 num_examples: 250 download_size: 2696871 dataset_size: 7405657.414835842 - config_name: burmese features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2055401.291311009 num_examples: 250 - name: test num_bytes: 2068167.105263158 num_examples: 250 - name: validation num_bytes: 2104032.8947368423 num_examples: 250 download_size: 2033389 dataset_size: 6227601.291311009 - config_name: chinese_simplified features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 918355.8361436754 num_examples: 250 - name: test num_bytes: 829274.5182012848 num_examples: 250 - name: validation num_bytes: 823207.0663811563 num_examples: 250 download_size: 1742219 dataset_size: 2570837.4207261167 - config_name: chinese_traditional features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 919010.4754769485 num_examples: 250 - name: test num_bytes: 826744.164882227 num_examples: 250 - name: validation num_bytes: 806565.1498929337 num_examples: 250 download_size: 1789768 dataset_size: 2552319.790252109 - config_name: english features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 753651.139233073 num_examples: 250 - name: test num_bytes: 717273.66710013 num_examples: 250 - name: validation num_bytes: 721542.6527958388 num_examples: 250 download_size: 1444665 dataset_size: 2192467.4591290415 - config_name: french features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 825153.8461538461 num_examples: 250 - name: test num_bytes: 648233.8858195212 num_examples: 250 - name: validation num_bytes: 664104.7421731123 num_examples: 250 download_size: 1258540 dataset_size: 2137492.4741464797 - config_name: gujarati features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 3118539.121614212 num_examples: 250 - name: test num_bytes: 2221719.271290606 num_examples: 250 - name: validation num_bytes: 2262670.9833187005 num_examples: 250 download_size: 2916519 dataset_size: 7602929.3762235185 - config_name: hausa features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 570238.625740106 num_examples: 250 - name: test num_bytes: 568007.481296758 num_examples: 250 - name: validation num_bytes: 578122.506234414 num_examples: 250 download_size: 1039907 dataset_size: 1716368.6132712779 - config_name: hindi features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1959523.4854050693 num_examples: 250 - name: test num_bytes: 1619873.4034135866 num_examples: 250 - name: validation num_bytes: 1611117.1866169323 num_examples: 250 download_size: 1905147 dataset_size: 5190514.075435588 - config_name: igbo features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 578673.4401147502 num_examples: 250 - name: test num_bytes: 611302.2030651341 num_examples: 250 - name: validation num_bytes: 606772.9885057472 num_examples: 250 download_size: 989262 dataset_size: 1796748.6316856313 - config_name: indonesian features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 846870.8161184038 num_examples: 250 - name: test num_bytes: 735218.2531380753 num_examples: 250 - name: validation num_bytes: 723896.338912134 num_examples: 250 download_size: 1344271 dataset_size: 2305985.408168613 - config_name: japanese features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1299743.6384085477 num_examples: 250 - name: test num_bytes: 1252634.420697413 num_examples: 250 - name: validation num_bytes: 1228125.7030371204 num_examples: 250 download_size: 2118633 dataset_size: 3780503.762143081 - config_name: kirundi features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 597250.130525583 num_examples: 250 - name: test num_bytes: 607965.1810584958 num_examples: 250 - name: validation num_bytes: 599110.3760445683 num_examples: 250 download_size: 1144753 dataset_size: 1804325.6876286468 - config_name: korean features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1103699.2852280464 num_examples: 250 - name: test num_bytes: 1029055.0 num_examples: 250 - name: validation num_bytes: 1030479.0909090909 num_examples: 250 download_size: 1844047 dataset_size: 3163233.3761371374 - config_name: kyrgyz features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1917066.7475728155 num_examples: 250 - name: test num_bytes: 1670476.5 num_examples: 250 - name: validation num_bytes: 1638172.5 num_examples: 250 download_size: 2471657 dataset_size: 5225715.747572815 - config_name: marathi features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 3179363.2486471613 num_examples: 250 - name: test num_bytes: 2608744.126284875 num_examples: 250 - name: validation num_bytes: 2563066.9970631422 num_examples: 250 download_size: 2965898 dataset_size: 8351174.371995178 - config_name: nepali features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1903332.8168044076 num_examples: 250 - name: test num_bytes: 1847913.103448276 num_examples: 250 - name: validation num_bytes: 1870868.2758620689 num_examples: 250 download_size: 2180166 dataset_size: 5622114.196114752 - config_name: oromo features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 908957.9416130629 num_examples: 250 - name: test num_bytes: 813752.642007926 num_examples: 250 - name: validation num_bytes: 815115.9180977542 num_examples: 250 download_size: 1484271 dataset_size: 2537826.501718743 - config_name: pashto features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1038053.7692468474 num_examples: 250 - name: test num_bytes: 966360.7859531773 num_examples: 250 - name: validation num_bytes: 992654.960981048 num_examples: 250 download_size: 1437478 dataset_size: 2997069.516181072 - config_name: persian features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1595828.8978011047 num_examples: 250 - name: test num_bytes: 1271854.4277006434 num_examples: 250 - name: validation num_bytes: 1256947.511005757 num_examples: 250 download_size: 2034596 dataset_size: 4124630.8365075053 - config_name: pidgin features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 481022.779105126 num_examples: 250 - name: test num_bytes: 511565.37793223286 num_examples: 250 - name: validation num_bytes: 512138.5751520417 num_examples: 250 download_size: 970867 dataset_size: 1504726.7321894006 - config_name: portuguese features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1067450.210794049 num_examples: 250 - name: test num_bytes: 821126.3414634146 num_examples: 250 - name: validation num_bytes: 826328.7456445993 num_examples: 250 download_size: 1662644 dataset_size: 2714905.297902063 - config_name: punjabi features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2917000.517346318 num_examples: 250 - name: test num_bytes: 1922143.2748538011 num_examples: 250 - name: validation num_bytes: 1950780.701754386 num_examples: 250 download_size: 2546578 dataset_size: 6789924.493954506 - config_name: russian features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1981897.1169448774 num_examples: 250 - name: test num_bytes: 1662099.614395887 num_examples: 250 - name: validation num_bytes: 1665811.9215938302 num_examples: 250 download_size: 2695179 dataset_size: 5309808.652934594 - config_name: scottish_gaelic features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 407392.9931454684 num_examples: 250 - name: test num_bytes: 466978.5 num_examples: 250 - name: validation num_bytes: 466613.0 num_examples: 250 download_size: 769177 dataset_size: 1340984.4931454684 - config_name: serbian_cyrillic features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2465827.319587629 num_examples: 250 - name: test num_bytes: 1643955.4455445544 num_examples: 250 - name: validation num_bytes: 1687471.1221122113 num_examples: 250 download_size: 2912628 dataset_size: 5797253.887244394 - config_name: serbian_latin features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1431336.5173172073 num_examples: 250 - name: test num_bytes: 992603.4103410341 num_examples: 250 - name: validation num_bytes: 963041.804180418 num_examples: 250 download_size: 2305130 dataset_size: 3386981.7318386594 - config_name: sinhala features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1599412.511542013 num_examples: 250 - name: test num_bytes: 1475039.0 num_examples: 250 - name: validation num_bytes: 1454452.0 num_examples: 250 download_size: 1888645 dataset_size: 4528903.511542013 - config_name: somali features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 712048.6413955048 num_examples: 250 - name: test num_bytes: 723985.5704697986 num_examples: 250 - name: validation num_bytes: 730702.0134228187 num_examples: 250 download_size: 1351143 dataset_size: 2166736.2252881224 - config_name: spanish features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1376380.3529257413 num_examples: 250 - name: test num_bytes: 1272710.2666386731 num_examples: 250 - name: validation num_bytes: 1276252.781860172 num_examples: 250 download_size: 2325049 dataset_size: 3925343.401424587 - config_name: swahili features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 663463.2818435046 num_examples: 250 - name: test num_bytes: 691559.5238095238 num_examples: 250 - name: validation num_bytes: 702616.0081053698 num_examples: 250 download_size: 1218240 dataset_size: 2057638.8137583982 - config_name: tamil features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2785898.024288004 num_examples: 250 - name: test num_bytes: 2528173.9023186974 num_examples: 250 - name: validation num_bytes: 2569812.284163789 num_examples: 250 download_size: 2672942 dataset_size: 7883884.210770491 - config_name: telugu features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 3051882.9526916803 num_examples: 250 - name: test num_bytes: 2608776.305683564 num_examples: 250 - name: validation num_bytes: 2591115.975422427 num_examples: 250 download_size: 3076441 dataset_size: 8251775.233797671 - config_name: thai features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 3003638.754534462 num_examples: 250 - name: test num_bytes: 2859638.014527845 num_examples: 250 - name: validation num_bytes: 2809091.1016949154 num_examples: 250 download_size: 3340890 dataset_size: 8672367.870757222 - config_name: tigrinya features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1497765.1807007887 num_examples: 250 - name: test num_bytes: 1068058.7371512481 num_examples: 250 - name: validation num_bytes: 1056244.1262848752 num_examples: 250 download_size: 1885889 dataset_size: 3622068.0441369116 - config_name: turkish features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 939307.2012069473 num_examples: 250 - name: test num_bytes: 841502.5758021784 num_examples: 250 - name: validation num_bytes: 860196.6440977333 num_examples: 250 download_size: 1605948 dataset_size: 2641006.421106859 - config_name: ukrainian features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1494510.5263767042 num_examples: 250 - name: test num_bytes: 1321049.7314317466 num_examples: 250 - name: validation num_bytes: 1326951.1020559364 num_examples: 250 download_size: 2316327 dataset_size: 4142511.3598643867 - config_name: urdu features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1238063.7072341684 num_examples: 250 - name: test num_bytes: 1101365.3050366517 num_examples: 250 - name: validation num_bytes: 1102391.8183967841 num_examples: 250 download_size: 1613554 dataset_size: 3441820.8306676047 - config_name: uzbek features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 2129719.1201353637 num_examples: 250 - name: test num_bytes: 1647984.7457627119 num_examples: 250 - name: validation num_bytes: 1691362.7118644067 num_examples: 250 download_size: 2445963 dataset_size: 5469066.577762483 - config_name: vietnamese features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 1628064.8920930522 num_examples: 250 - name: test num_bytes: 922911.4129080488 num_examples: 250 - name: validation num_bytes: 921885.2479441814 num_examples: 250 download_size: 1882304 dataset_size: 3472861.5529452823 - config_name: welsh features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 669700.8579942458 num_examples: 250 - name: test num_bytes: 663179.4819078947 num_examples: 250 - name: validation num_bytes: 650467.927631579 num_examples: 250 download_size: 1242582 dataset_size: 1983348.2675337195 - config_name: yoruba features: - name: id dtype: string - name: url dtype: string - name: title dtype: string - name: summary dtype: string - name: text dtype: string splits: - name: train num_bytes: 649711.3779527559 num_examples: 250 - name: test num_bytes: 601998.7389659521 num_examples: 250 - name: validation num_bytes: 616177.4905422446 num_examples: 250 download_size: 1049633 dataset_size: 1867887.6074609526 configs: - config_name: amharic data_files: - split: train path: amharic/train-* - split: test path: amharic/test-* - split: validation path: amharic/validation-* - config_name: arabic data_files: - split: train path: arabic/train-* - split: test path: arabic/test-* - split: validation path: arabic/validation-* - config_name: azerbaijani data_files: - split: train path: azerbaijani/train-* - split: test path: azerbaijani/test-* - split: validation path: azerbaijani/validation-* - config_name: bengali data_files: - split: train path: bengali/train-* - split: test path: bengali/test-* - split: validation path: bengali/validation-* - config_name: burmese data_files: - split: train path: burmese/train-* - split: test path: burmese/test-* - split: validation path: burmese/validation-* - config_name: chinese_simplified data_files: - split: train path: chinese_simplified/train-* - split: test path: chinese_simplified/test-* - split: validation path: chinese_simplified/validation-* - config_name: chinese_traditional data_files: - split: train path: chinese_traditional/train-* - split: test path: chinese_traditional/test-* - split: validation path: chinese_traditional/validation-* - config_name: english data_files: - split: train path: english/train-* - split: test path: english/test-* - split: validation path: english/validation-* - config_name: french data_files: - split: train path: french/train-* - split: test path: french/test-* - split: validation path: french/validation-* - config_name: gujarati data_files: - split: train path: gujarati/train-* - split: test path: gujarati/test-* - split: validation path: gujarati/validation-* - config_name: hausa data_files: - split: train path: hausa/train-* - split: test path: hausa/test-* - split: validation path: hausa/validation-* - config_name: hindi data_files: - split: train path: hindi/train-* - split: test path: hindi/test-* - split: validation path: hindi/validation-* - config_name: igbo data_files: - split: train path: igbo/train-* - split: test path: igbo/test-* - split: validation path: igbo/validation-* - config_name: indonesian data_files: - split: train path: indonesian/train-* - split: test path: indonesian/test-* - split: validation path: indonesian/validation-* - config_name: japanese data_files: - split: train path: japanese/train-* - split: test path: japanese/test-* - split: validation path: japanese/validation-* - config_name: kirundi data_files: - split: train path: kirundi/train-* - split: test path: kirundi/test-* - split: validation path: kirundi/validation-* - config_name: korean data_files: - split: train path: korean/train-* - split: test path: korean/test-* - split: validation path: korean/validation-* - config_name: kyrgyz data_files: - split: train path: kyrgyz/train-* - split: test path: kyrgyz/test-* - split: validation path: kyrgyz/validation-* - config_name: marathi data_files: - split: train path: marathi/train-* - split: test path: marathi/test-* - split: validation path: marathi/validation-* - config_name: nepali data_files: - split: train path: nepali/train-* - split: test path: nepali/test-* - split: validation path: nepali/validation-* - config_name: oromo data_files: - split: train path: oromo/train-* - split: test path: oromo/test-* - split: validation path: oromo/validation-* - config_name: pashto data_files: - split: train path: pashto/train-* - split: test path: pashto/test-* - split: validation path: pashto/validation-* - config_name: persian data_files: - split: train path: persian/train-* - split: test path: persian/test-* - split: validation path: persian/validation-* - config_name: pidgin data_files: - split: train path: pidgin/train-* - split: test path: pidgin/test-* - split: validation path: pidgin/validation-* - config_name: portuguese data_files: - split: train path: portuguese/train-* - split: test path: portuguese/test-* - split: validation path: portuguese/validation-* - config_name: punjabi data_files: - split: train path: punjabi/train-* - split: test path: punjabi/test-* - split: validation path: punjabi/validation-* - config_name: russian data_files: - split: train path: russian/train-* - split: test path: russian/test-* - split: validation path: russian/validation-* - config_name: scottish_gaelic data_files: - split: train path: scottish_gaelic/train-* - split: test path: scottish_gaelic/test-* - split: validation path: scottish_gaelic/validation-* - config_name: serbian_cyrillic data_files: - split: train path: serbian_cyrillic/train-* - split: test path: serbian_cyrillic/test-* - split: validation path: serbian_cyrillic/validation-* - config_name: serbian_latin data_files: - split: train path: serbian_latin/train-* - split: test path: serbian_latin/test-* - split: validation path: serbian_latin/validation-* - config_name: sinhala data_files: - split: train path: sinhala/train-* - split: test path: sinhala/test-* - split: validation path: sinhala/validation-* - config_name: somali data_files: - split: train path: somali/train-* - split: test path: somali/test-* - split: validation path: somali/validation-* - config_name: spanish data_files: - split: train path: spanish/train-* - split: test path: spanish/test-* - split: validation path: spanish/validation-* - config_name: swahili data_files: - split: train path: swahili/train-* - split: test path: swahili/test-* - split: validation path: swahili/validation-* - config_name: tamil data_files: - split: train path: tamil/train-* - split: test path: tamil/test-* - split: validation path: tamil/validation-* - config_name: telugu data_files: - split: train path: telugu/train-* - split: test path: telugu/test-* - split: validation path: telugu/validation-* - config_name: thai data_files: - split: train path: thai/train-* - split: test path: thai/test-* - split: validation path: thai/validation-* - config_name: tigrinya data_files: - split: train path: tigrinya/train-* - split: test path: tigrinya/test-* - split: validation path: tigrinya/validation-* - config_name: turkish data_files: - split: train path: turkish/train-* - split: test path: turkish/test-* - split: validation path: turkish/validation-* - config_name: ukrainian data_files: - split: train path: ukrainian/train-* - split: test path: ukrainian/test-* - split: validation path: ukrainian/validation-* - config_name: urdu data_files: - split: train path: urdu/train-* - split: test path: urdu/test-* - split: validation path: urdu/validation-* - config_name: uzbek data_files: - split: train path: uzbek/train-* - split: test path: uzbek/test-* - split: validation path: uzbek/validation-* - config_name: vietnamese data_files: - split: train path: vietnamese/train-* - split: test path: vietnamese/test-* - split: validation path: vietnamese/validation-* - config_name: welsh data_files: - split: train path: welsh/train-* - split: test path: welsh/test-* - split: validation path: welsh/validation-* - config_name: yoruba data_files: - split: train path: yoruba/train-* - split: test path: yoruba/test-* - split: validation path: yoruba/validation-* ---

提供机构：

llama-lang-adapt

搜集汇总

数据集介绍

构建方式

在自然语言处理领域，高质量的多语言摘要数据集对于推动跨语言文本理解研究至关重要。XL-Sum-small-v2 数据集作为 XL-Sum 的精简版本，其构建遵循了严谨的采样策略。该数据集从原始大规模语料库中，为每个语言配置均匀抽取了 250 条训练样本、250 条测试样本以及 250 条验证样本，确保了各语言子集在数据量上的绝对均衡。每条数据均包含唯一的标识符、原始文章链接、新闻标题、人工撰写的摘要以及完整的新闻正文，形成了结构化的四元组。这种设计不仅保留了原始数据的核心信息，还通过标准化处理，使得不同语言的数据具有统一的数据模式，便于跨语言模型的训练与评估。

特点

该数据集最显著的特点在于其广泛的语言覆盖范围与精致的规模控制。它涵盖了从阿姆哈拉语到祖鲁语等 44 种语言，横跨多个语系与书写系统，为多语言摘要研究提供了丰富的语言多样性。每个语言配置均包含 750 个样本，总样本量约为 33,000 条，这种小样本设计特别适合用于低资源语言的模型微调与零样本迁移学习。此外，数据集中每条样本的摘要与文本均保持原始语言的完整性，避免了机器翻译带来的噪声，确保了跨语言评估的可靠性。数据按语言划分为独立配置，用户可灵活选择单个或多个语言子集进行研究。

使用方法

使用该数据集时，研究者可通过 Hugging Face Datasets 库便捷加载。加载特定语言子集需指定配置名称，例如加载英语数据可执行 `load_dataset("llama-lang-adapt/XL-Sum-small-v2", "english")`。数据集已预分割为训练、测试和验证三部分，可直接用于监督学习中的序列到序列模型训练。每个样本以字典形式提供，包含 `id`、`url`、`title`、`summary` 和 `text` 字段，其中 `summary` 可作为目标输出，`text` 作为输入。对于多语言研究，可通过循环遍历不同配置名称，构建统一的训练管道，实现跨语言摘要模型的开发与评测。

背景与挑战

背景概述

在自然语言处理领域，多语言文本摘要任务长期受限于高质量标注数据的匮乏，尤其是低资源语言的代表性严重不足。XL-Sum-small-v2数据集由llama-lang-adapt团队于近年构建，旨在为跨语言摘要研究提供标准化评估基准。该数据集覆盖阿姆哈拉语、阿拉伯语、孟加拉语等44种语言，每种语言均包含250条训练、测试与验证样本，结构统一为新闻文章及其对应摘要。其核心研究问题在于验证模型在资源极不均衡的语言环境下的泛化能力，并推动多语言摘要系统向更公平、更全面的方向发展。作为XL-Sum的轻量级变体，该数据集因其语言覆盖广度与精炼规模，已成为评估低资源语言摘要性能的重要参考，对推动跨语言自然语言处理研究具有显著影响力。

当前挑战

XL-Sum-small-v2所面临的挑战主要体现在领域问题与构建过程两个层面。在领域问题上，多语言摘要的核心难点在于如何在不同语言间保持语义一致性与摘要质量，尤其是对于形态丰富或语序差异巨大的语言（如日语、泰语），模型常因缺乏语言特异性知识而生成不连贯或信息丢失的摘要。此外，低资源语言（如基隆迪语、奥罗莫语）的样本量极为有限，易导致过拟合与泛化能力不足。在构建过程中，数据采集面临来源多样性不足、摘要质量参差不齐等问题，不同语言间摘要的压缩率与信息密度难以统一，且人工标注成本高昂，难以大规模扩展。这些问题共同制约了数据集在多语言场景下的鲁棒性与实用性。

常用场景

经典使用场景

在自然语言处理领域，XL-Sum-small-v2数据集凭借其多语言、多领域的新闻文本与对应摘要的对齐特性，成为跨语言文本摘要任务的核心基准。研究者常利用其覆盖的45种语言（涵盖阿姆哈拉语、阿拉伯语、孟加拉语等低资源语言）来训练和评估序列到序列模型，例如mT5、mBART等架构在生成式摘要上的表现。该数据集通过提供每语言250条训练样本的紧凑规模，尤其适合快速迭代的学术实验与低资源场景下的迁移学习研究。

解决学术问题

该数据集有效缓解了多语言摘要研究中长期存在的标注数据匮乏问题，特别是针对非洲、南亚等地区的低资源语言。它为跨语言知识迁移提供了标准化的评估平台，使研究者能够量化模型在不同语言对之间的泛化能力。通过统一的数据格式和分层采样策略，XL-Sum-small-v2推动了多语言文本压缩、语义保持度评估以及零样本摘要等前沿方向的探索，其发布显著促进了低资源语言自然语言处理技术的公平性与包容性发展。

衍生相关工作

基于XL-Sum-small-v2，学术界衍生了一系列重要工作，包括多语言预训练模型的微调策略优化（如XLM-R与mT5的适配）、低资源语言摘要的对比学习框架设计，以及跨语言摘要质量评估指标（如LAPIS）的提出。部分研究进一步探索了基于该数据集的提示学习与少样本学习范式，推动了多语言文本生成模型的鲁棒性提升。此外，该数据集也被作为子集融入更大的XL-Sum完整版本，支持更全面的多语言摘要研究生态建设。

以上内容由遇见数据集搜集并总结生成