
ç¡æãŠãŒã¶ãŒã§ã䜿ããïŒOpenAI o3-mini 解説
â»ãæ¬èšäºã¯ãOpenAI o3-mini ã®å ¬åŒæ å ±ãããšã«ãå®éã®å©çšäŸã亀ããªããäœæããŠããŸããå©çšã®éã¯ãå¿ ãå ¬åŒæ å ±ãã確èªãã ããã
https://openai.com/index/openai-o3-mini/
OpenAI o3-mini ã®æŠèŠ
OpenAI o3-mini ã¯ãææ°ã®çæ°ç³»ïŒSTEMïŒåéã«ç¹åããå°åã¢ãã«ã§ãïŒ
åŸæ¥ã®å€§åã¢ãã«ã«æ¯ã¹ãã³ã¹ãå¹çãéåžžã«é«ããæ°åŠãç§åŠãããã°ã©ãã³ã°ãªã©ã®åéã§ã®è³ªåãåé¡è§£æ±ºã«åªããæ§èœãçºæ®ããŸãã
ç¹ã«ãåŠæ ¡ã®ææ¥ã課é¡äœæã«ãããŠãçæéã§æ£ç¢ºãªåçãåŸãããšãã§ããç¹ã倧ããªé
åã§ãã
ããã«ãä»åã®ã¢ããããŒãã«ãããChatGPT ã®ç¡æçã§ããReasonãã¢ãŒããéžæããããšã§ o3-mini ãå©çšã§ããããã«ãªã£ãŠããŸãïŒ
äž»ãªæ©èœãšç¹åŸŽ
çæ°ç³»åéã«æé©åãããé«ãæèå
o3-mini with medium reasoning effort matches o1âs performance in math, coding, and science, while delivering faster responses
ãšæžããŠããã
o3-mini ã¯ãæ°åŠãç§åŠãããã°ã©ãã³ã°ã®åé¡ã«å¯ŸããŠãéãæ£ç¢ºãªåçãæäŸãããšã®ããšã§ãïŒ
ããšãã°ãè€éãªæ°åŒã®èšç®ãçç§ã®å®éšçµæã®è§£èª¬ããçãæéã§è¡ãããããææ¥æºåã®æéççž®ã«ã€ãªãããŸããã
ããšã¯ãã£ã±ããå
±éãã¹ããšããããããã¹ãç³»ã®è§£ç解説ãšãããããå¿çšããæŒç¿åé¡äœããšãã§ã䜿ãããã§ããð€

ç«¶äºæ°åŠïŒAIME 2024ïŒã§ã¯ão3-mini high ã¯ãããŸã§ãããã£ã o1 ãäžåã£ãŠããŸããïŒ

å士ã¬ãã«ã®ç§åŠåé¡ïŒGPQA ãã€ã€ã¢ã³ãïŒã§ã¯ão1 ãš o3-mini ãåããããã§ããã
ããã ãšão3 ãåºãŠãããšã©ããªããã ïŒïŒ
ã©ãã ãåªç§ãªã®ïŒïŒ
ãšãªããŸããç¬
ãããŸã§ããŸã o3-mini ã¯ãŸã ãããªã«åªç§ãããªãã®ïŒãšæãããæ¹ãããããªãšæããŸãã
æ¬¡ã®ææšã¯è¯ããªã£ãŠããããšããããå°ããããããã«ãªã£ãŠããŠã

AIã®æ§èœè©äŸ¡ãæ€çŽ¢çµæã®ã©ã³ãã³ã°ãªã©ã§èŠãããã@1ãã@4ãã@8ããšãã£ãã@kããšãã衚èšã¯ãAIãåºåããè€æ°ã®åè£ã®äžã§ãã©ãã ãã®ç¢ºçã§æ£ããçããå«ãã§ããããç€ºãææšã§ãã
äž»ã«AIã®æ€çŽ¢ãçæã¿ã¹ã¯ãåé¡è§£æ±ºèœåãè©äŸ¡ããéã«äœ¿ãããŸãã
äŸãã°ãã@1 = 70%ã ãšããçµæãããå ŽåïŒ
ãAIãåºãã1ã€ç®ã®åçãã70%ã®ç¢ºçã§æ£è§£ãå«ãã§ããããšããæå³ã§ãã
@1ïŒããã1ã®ç²ŸåºŠïŒ
AIãçæãã1çªç®ã®åçãæ£è§£ã§ãã確çã
ã1çºã§æ£è§£ãåºãããããè©äŸ¡ã
粟床ãé«ãã»ã©ãæåã®åçãä¿¡é Œã§ããã
äŸãã°ã
æ€çŽ¢ãšã³ãžã³ïŒããã®æ¬ã®èè ã¯ïŒããšè³ªåããéã1çªç®ã®åçãæ£è§£ãªãæåã
翻蚳AIïŒããã®è±èªã®æç« ãèš³ããŠããšäŸé Œããæåã®èš³ãæ£ãã確çã
ããã°ã©ãã³ã°AIïŒã³ãŒãã®ãã°ä¿®æ£ãäŸé Œããæåã«åºãä¿®æ£ã³ãŒããæ£ãã確çã
@1ãé«ãAIã¯ãäžçºã§æ£ç¢ºãªçããåºããåªç§ãªã¢ãã«ïŒ
@4ïŒããã4ã®ç²ŸåºŠïŒ
AIãçæããäžäœ4ã€ã®åçã®ãã¡ã1ã€ã§ãæ£è§£ãå«ã確çã
ã1çºã§åœãããªããŠããéžæè¢ãå¢ããã°æ£è§£ã«ãã©ãçãããããè©äŸ¡ã
äŸãã°ã
質åå¿çAIïŒãããããã®æ°åŠã®åèæžã¯ïŒããšèããéãäžäœ4åã®ãã¡ãå°ãªããšã1åãæ¬åœã«ããããã§ããæ¬ã§ãã確çã
ããã°ã©ãã³ã°AIïŒããã®ãã°ã®è§£æ±ºçãæããŠããšäŸé ŒãããšããAIãææ¡ãã4ã€ã®ã³ãŒãã®ãã¡ã1ã€ã§ãæ£ããåã確çã
ç»åèªèAIïŒããã®åçã®åç©ã¯ïŒããšAIã«å€å¥ããããšããäžäœ4ã€ã®åè£ïŒäŸãã°ãç¬ãããªãªã«ãããæŽç¬ãããã¹ããŒãïŒã®äžã«æ£è§£ãå«ãŸããŠãã確çã
@4ãé«ãAIã¯ã1çºã§åœãŠãã®ã¯é£ãããŠããéžæè¢ã®äžã«ã¯æ£è§£ãå«ãŸããŠããå¯èœæ§ãé«ãïŒ
äŸãã°çåŸãAIã«è³ªåãããšãããæåã®åçãæ£ãããïŒ@1ïŒãã
ãããã€ãéžæè¢ãèŠãã°æ£è§£ããããïŒ@4, @8ïŒãã§AIã®å®åãæž¬ããŸãã
äŸïŒããã®æ°åŒã®è§£ãæ¹ãæããŠïŒã
ã¢ãã«ã®é床ãšããã©ãŒãã³ã¹
With intelligence comparable to OpenAI o1, OpenAI o3-mini delivers faster performance and improved efficiency. Beyond the STEM evaluations highlighted above, o3-mini demonstrates superior results in additional math and factuality evaluations with medium reasoning effort. In A/B testing, o3-mini delivered responses 24% faster than o1-mini, with an average response time of 7.7 seconds compared to 10.16 seconds.
ãšããããã«ã
A/B ãã¹ããšãããã®ã§ão3-mini 㯠o1-mini ããã 24% éãå¿çããå¹³åå¿çæé㯠o1-mini ã® 10.16 ç§ã«å¯Ÿã㊠7.7 ç§ã ã£ããšããããšã§ãã

é床ãäžãã£ãŠããŸãïŒ
å©ç𿹿³ãšçæç¹
å©ç𿹿³
ChatGPT ã®ã¢ãã«éžæã¡ãã¥ãŒããão3-miniããéžæãããã

ç¡æãã©ã³ã®å Žåã¯ãããã³ããã®ãšããã«ãããçç±ïŒReasonïŒãã¢ãŒãã§å©çšå¯èœã§ãã

ãããã«
OpenAI o3-mini ã¯ãæ°åŠãç§åŠãããã°ã©ãã³ã°ã®ææ¥ãããã«ã¯èª²é¡äœæãªã©ãããŸããŸãªã·ãŒã³ã§æŽ»çšã§ããŸãã
ç¡æãŠãŒã¶ãŒããçç±ïŒReasonïŒãã¢ãŒããéžã¶ããšã§å©çšã§ãããããå€ãã®æ¹ã§è©Šãããšãã§ããŸãã
ãã²ããã®æ©äŒã« o3-mini ã詊ããŠã¿ãŠãæ¥ã ã®ææ¥æºåãææäœæã«ã©ã®ãããªå€åããããããããããã®å¯èœæ§ãäœæããŠãã ããïŒ