Ç¥ÁØÈ­ Âü¿©¾È³»

TTAÀÇ Ç¥ÁØÇöȲ

Ȩ > Ç¥ÁØÈ­ °³¿ä > TTAÀÇ Ç¥ÁØÇöȲ

Ç¥ÁعøÈ£ TTAK.KO-10.1607 ±¸Ç¥ÁعøÈ£
Á¦°³Á¤ÀÏ 2025-12-05 ÃÑÆäÀÌÁö 14
ÇѱÛÇ¥ÁØ¸í Æ¯Â¡ º¤ÅÍ À¯»çµµ ±â¹Ý »ý¼ºÇü AIÀÇ ÄÜÅÙÃ÷ °ËÁõ Áöħ
¿µ¹®Ç¥Áظí Guideline for Generative AI Content Verification Based on Feature Vector Similarity
Çѱ۳»¿ë¿ä¾à ÀÌ Ç¥ÁØÀº »ý¼ºÇü AIÀÇ ÄÜÅÙÃ÷ »ý¼º °úÁ¤¿¡¼­ Àº´Ð °èÃþÀÇ Æ¯Â¡ º¤Å͸¦ ÃßÃâÇϰí À¯»çµµ¸¦ ºñ±³ÇÏ´Â ¿ªÃßÀû ¹æ¹ýÀ» ÅëÇØ ƯÁ¤ µ¥ÀÌÅͼÂÀÇ ÇнÀ ¿©ºÎ µî »ý¼º °úÁ¤À» °ËÁõÇÏ´Â ¹æ¹ý·ÐÀ» Á¦½ÃÇÑ´Ù.
¿µ¹®³»¿ë¿ä¾à This standard presents a back-tracing methodology that involves extracting feature vectors from hidden layers during the content generation process of Generative AI. By comparing the similarity of these vectors, it provides a method for verifying the generation process, such as whether specific datasets were learned.
±¹Á¦Ç¥ÁØ
°ü·ÃÆÄÀÏ TTAK.KO-10.1607.pdf TTAK.KO-10.1607.pdf            

ÀÌÀü
´ÙÀ½