'%3e%3cpath%20d='M44.9903%203.79049V0H37.7988V3.79049H39.6445V28.2095H37.7988V32H44.9903V28.2095H43.1446V3.79049H44.9903Z'%20fill='black'/%3e%3cpath%20d='M147.5%203.79049V0H140.305V3.79049H142.151V25.1238L126.499%203.79049H127.85V0H119.298V3.79049H121.144V28.2095H119.298V32H126.493V28.2095H124.647V6.89978L140.282%2028.2095H138.948V32H147.5V28.2095H145.654V3.79049H147.5Z'%20fill='black'/%3e%3cpath%20d='M111.688%203.79049V0H87.7104V3.79049H89.5133V28.2095H87.7104V32H111.688V28.2095H93.0167V17.7351H109.232V14.1537H93.0167V3.79049H111.688Z'%20fill='black'/%3e%3cpath%20d='M28.4057%2028.2095L18.1095%203.79049H19.0357V0H11.8409V3.79049H12.767L2.47419%2028.2095H0.5V32H7.69477V28.2095H6.16221L8.34075%2023.0397H22.5358L24.7144%2028.2095H23.0137V32H30.2085V28.2095H28.4024H28.4057ZM9.88649%2019.3774L15.4399%206.20508L20.9934%2019.3774H9.88649Z'%20fill='black'/%3e%3cpath%20d='M80.1699%203.79049V0H71.6173V3.79049H73.7134L64.5972%2014.4841C64.6631%2014.8416%2064.7026%2015.2092%2064.7026%2015.5869C64.7026%2016.1703%2064.6169%2016.7301%2064.4653%2017.2629L73.7991%2028.2129H71.6205V32.0034H80.1732V28.2129H78.2484L67.9853%2015.9545L78.1693%203.79387H80.1732L80.1699%203.79049Z'%20fill='black'/%3e%3cpath%20d='M62.4711%200H55.2764V3.79049H62.4711V0Z'%20fill='black'/%3e%3cpath%20d='M62.4711%2028.2095H55.2764V32H62.4711V28.2095Z'%20fill='black'/%3e%3cpath%20d='M58.8723%209.62109C55.6589%209.62109%2053.042%2012.2953%2053.042%2015.5867C53.042%2018.8781%2055.6556%2021.5524%2058.8723%2021.5524C62.089%2021.5524%2064.7026%2018.8781%2064.7026%2015.5867C64.7026%2012.2953%2062.089%209.62109%2058.8723%209.62109Z'%20fill='%23E60012'/%3e%3c/g%3e%3cdefs%3e%3cclipPath%20id='clip0_2422_1755'%3e%3crect%20width='147'%20height='32'%20fill='white'%20transform='translate(0.5)'/%3e%3c/clipPath%3e%3c/defs%3e%3c/svg%3e)
LLMの出力が信頼できるかを判定する手法 Google DeepMindが新しく考案 | AIDB
生成AI全般
更新日2024.06.14
高橋智也 (Tomoya Takahashi)
Google DeepMindの研究者らは、LLMの「不確実性」について調べています。簡単に言うと、モデルが質問に対してどの程度自信を持って答えているかを見極める方法を探っています。
この記事をシェアする
LLMの出力が信頼できるかを判定する手法 Google DeepMindが新しく考案 | AIDB
更新日2024.06.14
高橋智也 (Tomoya Takahashi)
Google DeepMindの研究者らは、LLMの「不確実性」について調べています。簡単に言うと、モデルが質問に対してどの程度自信を持って答えているかを見極める方法を探っています。
この記事をシェアする
