ãChatGPTããªã©åºã使ããã¦ããAIãµã¼ãã¹ã¯ãé常ã§ããã°ã»ã¼ããã£ãããããã¦ãã¦ããäººãæ®ºãæ¹æ³ããçå¼¾ãã¤ããæ¹æ³ãã¨ãã£ãå«ççã«åé¡ã®ãã質åã«ã¯çããªãããã«ãªã£ã¦ãã¾ããã¨ãããããã¾ãã«ãå¤ã質åãä¸åº¦ã«ã¶ã¤ãã¦ãã¾ããã¨ã«ããã»ã¼ããã£ãå¤ããAIãåé¡ã®ããåçãè¡ã£ã¦ãã¾ãå¯è½æ§ããããã¨ããããã¾ããã Many-shot jailbreaking \ Anthropic https://www.anthropic.com/research/many-shot-jailbreaking å¤§è¦æ¨¡è¨èªã¢ãã«(LLM)ã¯ãã¢ãã«ã®å·æ°ã¨å ±ã«ã³ã³ããã¹ãã¦ã£ã³ãã¦(æ±ããæ å ±é)ãå¢å ãã¦ãããè¨äºä½ææç¹ã§ã¯é·ç·¨å°èª¬æ°åå(100ä¸ãã¼ã¯ã³ä»¥ä¸)ãåãæ±ããã¢ãã«ãåå¨ãã¾ãã 大éã®æ å ±ãæ±ããã¨ããã®ã¯ã¦ã¼ã¶ã¼ã«ã¨ã£ã¦å©ç¹ã«ãªãã¾ããã大éã®æ å ±ãæ±ããã¨ã«ã


{{#tags}}- {{label}}
{{/tags}}