アーカイブの重要な論文について、魅力的なポッドキャストやビデオを通じて最新情報をお届けします。この番組は洞察に満ちた要約を提供し、学術研究を身近で理解しやすくします。
[QA] CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training
7 mins • Apr 19, 2025
Charts
- 188Decreased by 2
'/%3e%3cpath%20d='M74.965,77.269%20C74.795,75.733%2074.283,74.624%2073.259,73.6%20C71.339,71.595%2067.968,70.272%2064,70.272%20C60.032,70.272%2056.661,71.552%2054.741,73.6%20C53.76,74.667%2053.205,75.733%2053.035,77.269%20C52.693,80.256%2052.907,82.816%2053.248,86.955%20C53.589,90.88%2054.229,96.128%2055.04,101.419%20C55.637,105.216%2056.107,107.264%2056.533,108.715%20C57.259,111.104%2059.861,113.152%2064,113.152%20C68.139,113.152%2070.784,111.061%2071.467,108.715%20C71.893,107.264%2072.363,105.216%2072.96,101.419%20C73.771,96.085%2074.411,90.88%2074.752,86.955%20C75.136,82.816%2075.307,80.256%2074.965,77.269%20z'%20fill='%23FFFFFF'%20id='XMLID_657_'/%3e%3cpath%20d='M74.496,55.509%20C74.496,61.312%2069.803,66.005%2064,66.005%20C58.197,66.005%2053.504,61.312%2053.504,55.509%20C53.504,49.707%2058.197,45.013%2064,45.013%20C69.803,45.013%2074.496,49.749%2074.496,55.509%20z'%20fill='%23FFFFFF'%20id='XMLID_655_'/%3e%3cpath%20d='M63.872,14.165%20C39.381,14.251%2019.328,34.133%2018.987,58.624%20C18.731,78.464%2031.403,95.445%2049.109,101.675%20C49.536,101.845%2049.963,101.461%2049.92,101.035%20C49.707,99.499%2049.451,97.963%2049.28,96.427%20C49.195,95.872%2048.853,95.445%2048.384,95.189%20C34.389,89.088%2024.619,75.051%2024.789,58.795%20C25.003,37.461%2042.411,20.096%2063.701,19.925%20C85.504,19.755%20103.296,37.419%20103.296,59.179%20C103.296,75.264%2093.568,89.088%2079.701,95.189%20C79.189,95.403%2078.848,95.872%2078.805,96.427%20C78.592,97.963%2078.379,99.499%2078.165,101.035%20C78.08,101.504%2078.549,101.845%2078.976,101.675%20C96.512,95.531%20109.099,78.805%20109.099,59.179%20C109.013,34.347%2088.747,14.123%2063.872,14.165%20z'%20fill='%23FFFFFF'%20id='XMLID_653_'/%3e%3cpath%20d='M62.848,29.099%20C46.891,29.696%2034.091,42.88%2033.877,58.837%20C33.749,69.333%2038.997,78.635%2047.061,84.139%20C47.445,84.395%2048,84.096%2048,83.627%20C47.872,81.792%2047.872,80.171%2047.957,78.464%20C48,77.909%2047.787,77.397%2047.36,77.013%20C42.453,72.405%2039.467,65.835%2039.637,58.581%20C39.979,45.781%2050.261,35.371%2063.061,34.859%20C76.971,34.304%2088.405,45.483%2088.405,59.221%20C88.405,66.219%2085.419,72.533%2080.683,77.013%20C80.299,77.397%2080.085,77.909%2080.085,78.464%20C80.171,80.128%2080.128,81.749%2080.043,83.584%20C80,84.053%2080.555,84.395%2080.981,84.096%20C88.917,78.677%2094.165,69.504%2094.165,59.179%20C94.208,42.197%2080,28.416%2062.848,29.099%20z'%20fill='%23FFFFFF'%20id='XMLID_651_'/%3e%3c/g%3e%3c/svg%3e)
- 142NEW
- 197Decreased by 11
- 189Decreased by 10
- 199Decreased by 5
最近のエピソード
Apr 19, 2025
[QA] CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training
7 mins
Apr 19, 2025
CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training
21 mins
Apr 18, 2025
[QA] Antidistillation Sampling
7 mins
Apr 18, 2025
Antidistillation Sampling
11 mins
Apr 18, 2025
[QA] Position: The Most Expensive Part of an LLM should be its Training Data
7 mins
