{"id":65102,"date":"2024-11-18T19:14:49","date_gmt":"2024-11-18T10:14:49","guid":{"rendered":"https:\/\/smilegate.ai\/?p=65102"},"modified":"2024-11-18T19:18:32","modified_gmt":"2024-11-18T10:18:32","slug":"ragas-rag-%ec%8b%9c%ec%8a%a4%ed%85%9c-%ec%84%b1%eb%8a%a5-%ed%8f%89%ea%b0%80-%ec%9d%b4%ec%a0%9c%eb%8a%94-%ec%a0%9c%eb%8c%80%eb%a1%9c-%ed%95%b4%eb%b3%b4%ec%9e%90","status":"publish","type":"post","link":"https:\/\/smilegate.ai\/cn\/2024\/11\/18\/ragas-rag-%ec%8b%9c%ec%8a%a4%ed%85%9c-%ec%84%b1%eb%8a%a5-%ed%8f%89%ea%b0%80-%ec%9d%b4%ec%a0%9c%eb%8a%94-%ec%a0%9c%eb%8c%80%eb%a1%9c-%ed%95%b4%eb%b3%b4%ec%9e%90\/","title":{"rendered":"RAGAS: RAG \uc2dc\uc2a4\ud15c \uc131\ub2a5 \ud3c9\uac00, \uc774\uc81c\ub294 \uc81c\ub300\ub85c \ud574\ubcf4\uc790!"},"content":{"rendered":"

[Interactive AI \uae30\uc220\ud300 \uae40\uc724\ud61c]<\/p>\n\n\n\n

RAG(Retrieval-Augmented Generation, \uac80\uc0c9 \uc99d\uac15 \uc0dd\uc131) \ud30c\uc774\ud504\ub77c\uc778\uc740 \ub300\uaddc\ubaa8 \uc5b8\uc5b4 \ubaa8\ub378(LLM)\uc758 \ud55c\uacc4\ub97c \ubcf4\uc644\ud558\uae30 \uc704\ud574 \uc678\ubd80 \uc9c0\uc2dd\uc744 \ud65c\uc6a9\ud558\uc5ec \ubcf4\ub2e4 \uc815\ud655\ud558\uace0 \uad00\ub828\uc131 \ub192\uc740 \uc751\ub2f5\uc744 \uc0dd\uc131\ud558\ub294 \uae30\uc220\uc785\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \uc774\ub7ec\ud55c \ud30c\uc774\ud504\ub77c\uc778\uc758 \uc131\ub2a5\uc744 \uac1d\uad00\uc801\uc73c\ub85c \ud3c9\uac00\ud558\uae30\ub780 \uc27d\uc9c0 \uc54a\uc740 \uc77c\uc785\ub2c8\ub2e4. \uc774\ubc88 \uae00\uc5d0\uc11c\ub294 \uc774\uc5d0 \ubd80\uc751\ud558\uc5ec \ub4f1\uc7a5\ud55c RAG \ud30c\uc774\ud504\ub77c\uc778 \ud3c9\uac00 \ud504\ub808\uc784\uc6cc\ud06c \uc911 \ud558\ub098\uc778 RAGAS(Retrieval-Augmented Generation Assessment)\uc5d0 \ub300\ud574 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.


<\/p>\n\n\n\n


\n\n\n\n

RAGAS\ub780?<\/h2>\n\n\n\n

RAGAS\ub294 RAG \ud30c\uc774\ud504\ub77c\uc778\uc758 \uc815\ub7c9\uc801 \ud3c9\uac00\ub97c \uc9c0\uc6d0\ud558\ub294 \uc624\ud508 \uc18c\uc2a4 \ud504\ub808\uc784\uc6cc\ud06c<\/strong>\uc785\ub2c8\ub2e4. \uc774 \ud504\ub808\uc784\uc6cc\ud06c\ub294 \ub2e8\uc21c\ud55c \uc131\ub2a5 \ud3c9\uac00\ub97c \ub118\uc5b4, \uc2dc\uc2a4\ud15c\uc758 \uac01 \uad6c\uc131 \uc694\uc18c\ub97c \uc138\ubc00\ud788 \ubd84\uc11d\ud560 \uc218 \uc788\ub3c4\ub85d \uc124\uacc4\ub418\uc5c8\uc2b5\ub2c8\ub2e4. \uc774\ub97c \ud1b5\ud574 \uac1c\ubc1c\uc790\ub294 \ubb38\uc81c\uc758 \uc6d0\uc778\uc744 \ube60\ub974\uac8c \ud30c\uc545\ud558\uace0, \uc2dc\uc2a4\ud15c\uc744 \ucd5c\uc801\ud654\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n


<\/p>\n\n\n\n


\n\n\n\n


<\/p>\n\n\n\n

RAGAS Metrics<\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

RAGAS\ub294 RAG \ud30c\uc774\ud504\ub77c\uc778\uc758 \uc131\ub2a5\uc744 \ub2e4\uac01\ub3c4\ub85c \ud3c9\uac00\ud558\uae30 \uc704\ud574 \uc5ec\ub7ec \ud575\uc2ec \uc9c0\ud45c\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \uc8fc\uc694 \uc9c0\ud45c\ub294 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n\n\n\n

    \n
  • Faithfulness(\ucda9\uc2e4\uc131):<\/strong> \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc774 \uc8fc\uc5b4\uc9c4 \ucee8\ud14d\uc2a4\ud2b8\uc5d0 \uc5bc\ub9c8\ub098 \ucda9\uc2e4\ud55c\uc9c0\ub97c \ud3c9\uac00\ud569\ub2c8\ub2e4. \uc989, \ub2f5\ubcc0\uc774 \ucee8\ud14d\uc2a4\ud2b8\uc5d0 \uae30\ubc18\ud558\uc5ec \uc815\ud655\ud55c \uc815\ubcf4\ub97c \uc81c\uacf5\ud558\ub294\uc9c0\ub97c \uce21\uc815\ud569\ub2c8\ub2e4.<\/li>\n\n\n\n
  • Answer Relevancy(\ub2f5\ubcc0 \uad00\ub828\uc131):<\/strong> \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc774 \uc8fc\uc5b4\uc9c4 \uc9c8\ubb38\uacfc \uc5bc\ub9c8\ub098 \uad00\ub828\uc131\uc774 \uc788\ub294\uc9c0\ub97c \ud3c9\uac00\ud569\ub2c8\ub2e4. \uc9c8\ubb38\uc5d0 \ub300\ud55c \uba85\ud655\ud558\uace0 \uc801\uc808\ud55c \ub2f5\ubcc0\uc744 \uc81c\uacf5\ud558\ub294\uc9c0\ub97c \uce21\uc815\ud569\ub2c8\ub2e4.<\/li>\n\n\n\n
  • Context Precision(\ucee8\ud14d\uc2a4\ud2b8 \uc815\ubc00\ub3c4):<\/strong> \uc9c8\ubb38\uc5d0 \ub300\ud55c \ucee8\ud14d\uc2a4\ud2b8 \uc815\ubcf4\ub97c \uac80\uc0c9\ud558\uc5ec \uac00\uc838\uc628 \ucee8\ud14d\uc2a4\ud2b8 \uc911, \uc9c8\ubb38\uc5d0 \ub300\ud55c \ub2f5\ubcc0\uacfc \uad00\ub828 \uc788\ub294 \ubb38\uc11c\uac00 \uc5bc\ub9c8\ub098 \uc0c1\uc704\uc5d0 \ub7ad\ud06c\ub418\uc5b4 \uc788\ub294\uc9c0\ub97c \ud3c9\uac00\ud569\ub2c8\ub2e4. \uc989, \uac80\uc0c9\ub41c \uc815\ubcf4\uc758 \uc815\ud655\ub3c4\ub97c \uce21\uc815\ud569\ub2c8\ub2e4.<\/li>\n\n\n\n
  • Context Recall(\ucee8\ud14d\uc2a4\ud2b8 \uc7ac\ud604\uc728):<\/strong> \uc9c8\ubb38\uc5d0 \ub300\ud55c \ub2f5\ubcc0\uc744 \uc0dd\uc131\ud558\ub294 \ub370 \ud544\uc694\ud55c \ucee8\ud14d\uc2a4\ud2b8 \uc815\ubcf4\ub97c \uac80\uc0c9\ud560 \uc218 \uc788\ub294\uc9c0\ub97c \ud3c9\uac00\ud569\ub2c8\ub2e4. \uc989, \ud544\uc694\ud55c \uc815\ubcf4\ub97c \uc5bc\ub9c8\ub098 \uc798 \ucc3e\uc544\ub0b4\ub294\uc9c0\ub97c \uce21\uc815\ud569\ub2c8\ub2e4.<\/li>\n<\/ul>\n\n\n\n


    <\/p>\n\n\n\n


    \n\n\n\n


    <\/p>\n\n\n\n

    \uac01 \uc9c0\ud45c\uc758 \uac1c\ub150\uacfc \uacc4\uc0b0 \ubc29\ubc95\uc744 \ub354 \uc790\uc138\ud788 \uc54c\uc544\ubd05\uc2dc\ub2e4.<\/p>\n\n\n\n

    Faithfulness<\/h3>\n\n\n\n
      \n
    • \uc8fc\uc5b4\uc9c4 \ubb38\ub9e5\uc744 \uc5bc\ub9c8\ub098 \uc798 \ubc18\uc601\ud558\uc5ec \ub2f5\ubcc0\uc744 \uc0dd\uc131\ud558\uc600\ub294\uc9c0 \ud3c9\uac00.<\/li>\n\n\n\n
    • \uc810\uc218 \ubc94\uc704: 0~1 (1\uc5d0 \uac00\uae4c\uc6b8\uc218\ub85d \uc88b\uc74c.)<\/li>\n\n\n\n
    • context\uc5d0\uc11c \ub2f5\ubcc0\uc5d0 \ub300\ud55c claims\ub97c \uc720\ucd94 \uac00\ub2a5\ud560 \uc218\ub85d \ub192\uc740 \uc810\uc218\ub97c \ubd80\uc5ec.<\/li>\n<\/ul>\n\n\n\n

      Calculation<\/h4>\n\n\n\n

      Step1. \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc73c\ub85c\ubd80\ud130 statements\ub97c \ucd94\ucd9c\ud568. (llm \ud65c\uc6a9)<\/p>\n\n\n\n

      Step2. \ucd94\ucd9c\ud55c statements\uac00 \uc8fc\uc5b4\uc9c4 context\uc5d0\uc11c \ucd94\ub860\ub420 \uc218 \uc788\ub294\uc9c0 \uc5ec\ubd80\ub97c \ud310\ubcc4\ud568. (llm \ud65c\uc6a9)<\/p>\n\n\n\n

      \"\"<\/figure>\n\n\n\n


      <\/p>\n\n\n\n

      Answer Relevancy<\/h3>\n\n\n\n
        \n
      • \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc774 \uc9c8\ubb38\uc5d0 \uc5bc\ub9c8\ub098 \uc798 \ubd80\ud569\ud558\ub294\uc9c0 \ud3c9\uac00<\/li>\n\n\n\n
      • \uc810\uc218 \ubc94\uc704: -1~1 (1\uc5d0 \uac00\uae4c\uc6b8\uc218\ub85d \uc88b\uc74c)<\/li>\n\n\n\n
      • \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc744 \uae30\ubc18\uc73c\ub85c \uc0dd\uc131\ud55c \uc9c8\ubb38\uacfc \uc6d0\ub798 \uc9c8\ubb38\uc758 cosine \uc720\uc0ac\ub3c4\ub97c \ud65c\uc6a9\ud568.\n
          \n
        • answer\uacfc context\ub9cc\uc73c\ub85c question\uc744 \uc7ac\uad6c\uc131\ud560 \uc218 \uc788\ub2e4\ub294 \uac1c\ub150\uc744 \ud65c\uc6a9\ud55c \ubc29\uc2dd.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n

          Calculation<\/h4>\n\n\n\n

          Step1. \ucee8\ud14d\uc2a4\ud2b8\ub97c \uae30\ubc18\uc73c\ub85c \uc0dd\uc131\ub41c \ub2f5\ubcc0\uc5d0 \ub300\ud55c \uc9c8\ubb38\uc744 \uc0dd\uc131\ud568. (Reverse-engineer, llm \ud65c\uc6a9)<\/p>\n\n\n\n

          Step2. \uc2e4\uc81c \uc9c8\ubb38\uacfc \uc0dd\uc131 \uc9c8\ubb38 \uc0ac\uc774\uc758 cosine similarity\ub97c \uc5f0\uc0b0\ud568.<\/p>\n\n\n\n

          \"\"<\/figure>\n\n\n\n


          <\/p>\n\n\n\n

          Context Precision<\/h3>\n\n\n\n
            \n
          • \uac80\uc0c9\ub41c \uac01 k\uac1c\uc758 context\uac00 question, ground truth\uc640 \uc5f0\uad00\uc774 \uc788\ub294\uc9c0 \ud3c9\uac00.<\/li>\n\n\n\n
          • \uc810\uc218 \ubc94\uc704: 0~1 (1\uc5d0 \uac00\uae4c\uc6b8\uc218\ub85d \uc88b\uc74c)<\/li>\n\n\n\n
          • \uac80\uc0c9 \uacb0\uacfc\uc640 \uc720\ud6a8\uc131\uc744 \uae30\ubc18\uc73c\ub85c precision\uc744 \uacc4\uc0b0\ud568.<\/li>\n<\/ul>\n\n\n\n

            Calculation<\/h4>\n\n\n\n

            Step1. \uac80\uc0c9\ub41c k\uac1c\uc758 context(chunk)\uac00 question, ground truth\uc640 \uc5f0\uad00\uc774 \uc788\ub294\uc9c0 \ud310\ubcc4. (llm \ud65c\uc6a9)<\/p>\n\n\n\n

            Step2. \uac01 context\uc5d0 \ub300\ud55c precision \uacc4\uc0b0<\/p>\n\n\n\n

            Step3. precision\uc758 \ud3c9\uade0\uc744 \uad6c\ud568<\/p>\n\n\n\n

            \"\"<\/figure>\n\n\n\n


            <\/p>\n\n\n\n

            Context Recall<\/h3>\n\n\n\n
              \n
            • \uac80\uc0c9\ub41c context\uac00 ground truth\uc640 \uc5bc\ub9c8\ub098 \uc798 \uc77c\uce58\ud558\ub294\uc9c0 \ud3c9\uac00.<\/li>\n\n\n\n
            • \uc810\uc218 \ubc94\uc704: 0~1 (1\uc5d0 \uac00\uae4c\uc6b8\uc218\ub85d \uc88b\uc74c)<\/li>\n\n\n\n
            • ground truth\uc758 \uac01 claim\uc774 \uac80\uc0c9\ub41c context\uc640 \uc5bc\ub9c8\ub098 \uc77c\uce58\ud558\ub294\uc9c0 \uacc4\uc0b0\ud568.<\/li>\n<\/ul>\n\n\n\n

              Calculation<\/h4>\n\n\n\n

              Step1. ground truth\ub97c \ub3c5\ub9bd\ub41c statements\ub85c \ubd84\ub9ac\ud568.<\/p>\n\n\n\n

              Step2. \uac01 statements\uc758 \uc815\ubcf4\ub97c \uac80\uc0c9\ub41c contexts\uc5d0\uc11c \ucc3e\uc744 \uc218 \uc788\ub294\uc9c0 \ud310\ubcc4\ud568.<\/p>\n\n\n\n

              Step3. step2\uc758 \uacb0\uacfc\ub97c \uae30\ubc18\uc73c\ub85c recall\uc744 \uacc4\uc0b0\ud568.<\/p>\n\n\n\n

              \"\"<\/figure>\n\n\n\n


              <\/p>\n\n\n\n


              \n\n\n\n


              <\/p>\n\n\n\n

              \uc2e4\uc2b5: RAGAS\ub85c RAG \ud3c9\uac00\ud574\ubcf4\uae30<\/h2>\n\n\n\n

              \uc2e4\uc81c\ub85c Streamlit\uc744 \ud65c\uc6a9\ud55c \uac04\ub2e8\ud55c Demo\ub97c \ub9cc\ub4e4\uc5b4 RAGAS \ud3c9\uac00\ub97c \ud574\ubcf4\uc558\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n


              <\/p>\n\n\n\n

              1. \ud3c9\uac00 \ub370\uc774\ud130\uc14b \uc785\ub825<\/h3>\n\n\n\n

              \uc6b0\uc120 \ud3c9\uac00 \ub370\uc774\ud130\uc14b\uc744 \uc785\ub825\ud569\ub2c8\ub2e4. \ubcf8\uc778\uc740 \uc9c1\uc811 gpt-4o\ub97c \ud1b5\ud574 \uc0dd\uc131\ud558\uace0 manual \uc815\uc81c\ud55c \ub370\uc774\ud130\uc14b\uc744 \ud65c\uc6a9\ud588\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

              \ud3c9\uac00 \ub370\uc774\ud130\uc14b\uc5d0\ub294 ‘question’, ‘ground_truth’ \uceec\ub7fc\uc774 \uc874\uc7ac\ud574\uc57c \ud569\ub2c8\ub2e4.<\/p>\n\n\n\n

                \n
              • question: \uc0ac\uc6a9\uc790 \uc785\ub825\uc5d0 \ud574\ub2f9\ud558\ub294 \uc9c8\ubb38<\/li>\n\n\n\n
              • ground_truth: \uc9c8\ubb38\uc5d0 \ub300\uc751\ub418\ub294 \uc815\ub2f5<\/li>\n<\/ul>\n\n\n\n


                <\/p>\n\n\n\n

                2. \ubaa8\ub378 \uc120\ud0dd<\/h3>\n\n\n\n

                \ud3c9\uac00\ud558\uace0 \uc2f6\uc740 RAG \ud30c\uc774\ud504\ub77c\uc778\uc758 LLM(\uc0dd\uc131)\uacfc embedding model(\uac80\uc0c9)\uc744 \uc120\ud0dd\ud569\ub2c8\ub2e4.<\/p>\n\n\n\n

                \"\"<\/figure>\n\n\n\n


                <\/p>\n\n\n\n

                3. RAGAS \ud3c9\uac00 \uc2e4\ud589<\/h3>\n\n\n\n

                ragas \ud3c9\uac00 \uacb0\uacfc \ud655\uc778<\/p>\n\n\n\n

                \"\"<\/figure>\n\n\n\n

                <\/p>\n\n\n\n


                <\/p>\n\n\n\n

                +) LangSmith\ub97c \ud65c\uc6a9\ud55c \ud3c9\uac00 \ubaa8\ub2c8\ud130\ub9c1<\/h3>\n\n\n\n

                \uc704\uc640 \uac19\uc740 \uc815\ub7c9 \uc810\uc218\uac00 \uc5b4\ub5bb\uac8c \uc0b0\uc815\ub410\ub294\uc9c0 \uc138\uc138\ud558\uac8c \ud655\uc778\ud560 \uc218 \uc788\ub3c4\ub85d RAGAS\ub294 LangSmith \uc5f0\ub3d9 \uae30\ub2a5\ub3c4 \uc81c\uacf5\ud558\uace0 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

                \uac01 metrics\uc744 \uacc4\uc0b0\ud560 \ub54c \uc0ac\uc6a9\ub41c \ubaa8\ub378, \ud504\ub86c\ud504\ud2b8, output, \ud1a0\ud070 \uc218 \ubc0f \ube44\uc6a9, latency \ub4f1\uc744 \uc544\uc8fc \uc790\uc138\ud788 \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n

                \"\"<\/figure>\n\n\n\n


                <\/p>\n\n\n\n

                <\/h1>\n\n\n\n
                \n\n\n\n


                <\/p>\n\n\n\n

                \ub9c8\ubb34\ub9ac<\/h2>\n\n\n\n

                RAG \ud30c\uc774\ud504\ub77c\uc778\uc740 \uac80\uc0c9\uacfc \uc0dd\uc131\uc774\ub77c\ub294 \ub450 \ucd95\uc774 \uacb0\ud569\ub41c \ubcf5\uc7a1\ud55c \uc2dc\uc2a4\ud15c\uc785\ub2c8\ub2e4. \uc774\ucc98\ub7fc \ub2e4\uce35\uc801\uc778 \uad6c\uc870\uc5d0\uc11c\ub294 \uac01 \ub2e8\uacc4\uc758 \uc131\ub2a5\uc744 \uc138\ubc00\ud788 \ud3c9\uac00\ud558\uace0 \uac1c\uc120\ud558\uae30 \uc704\ud55c \uc815\ub7c9\uc801 \ud3c9\uac00 \ud504\ub808\uc784\uc6cc\ud06c<\/strong>\uac00 \ud544\uc218\uc801\uc785\ub2c8\ub2e4.<\/p>\n\n\n\n

                \uac1d\uad00\uc801\uc778 \uc9c0\ud45c\uc640 \ub370\uc774\ud130\ub97c \ubc14\ud0d5\uc73c\ub85c \ud30c\uc774\ud504\ub77c\uc778\uc758 \uac15\uc810\uacfc \uc57d\uc810\uc744 \ud30c\uc545\ud558\uba74, \ub2e8\uc21c\ud788 \ubb38\uc81c\ub97c \uc9c4\ub2e8\ud558\ub294 \ub370 \uadf8\uce58\uc9c0 \uc54a\uace0 \uc2dc\uc2a4\ud15c\uc758 \uc804\ubc18\uc801\uc778 \ud488\uc9c8\uc744 \uc9c0\uc18d\uc801\uc73c\ub85c \ud5a5\uc0c1\uc2dc\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ub294 \ub2e8\uc21c\ud55c \uacb0\uacfc\uc758 \uc815\ud655\uc131\uc744 \ub118\uc5b4, \uc2e0\ub8b0\uc131\uacfc \uc0ac\uc6a9\uc790 \uacbd\ud5d8\uc744 \ubaa8\ub450 \ud5a5\uc0c1\uc2dc\ud0a4\ub294 \ub370 \ud575\uc2ec\uc801\uc778 \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4.<\/p>\n\n\n\n

                \uacb0\uad6d, \uc815\ub7c9\uc801 \ud3c9\uac00 \ud504\ub808\uc784\uc6cc\ud06c\ub294 AI \uc2dc\uc2a4\ud15c\uc758 \uc131\ub2a5\uc744 \ud3c9\uac00\ud558\uace0 \ucd5c\uc801\ud654\ud558\ub294 \ub370 \uc788\uc5b4 \uc5c6\uc5b4\uc11c\ub294 \uc548 \ub420 \ub3c4\uad6c\uc785\ub2c8\ub2e4. \uc774\ub97c \ud1b5\ud574 \ubcf5\uc7a1\ud55c RAG \uc2dc\uc2a4\ud15c\uc5d0\uc11c\ub3c4 \ub370\uc774\ud130 \uae30\ubc18\uc758 \ubc1c\uc804\uc774 \uac00\ub2a5\ud558\uba70, \ubcf4\ub2e4 \uacac\uace0\ud558\uace0 \ud6a8\uc728\uc801\uc778 AI \uc194\ub8e8\uc158\uc744 \uad6c\ucd95\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n\n\n\n


                <\/p>\n\n\n\n

                \u53c2\u8003<\/h4>\n\n\n\n
                  \n
                • https:\/\/github.com\/explodinggradients\/ragas<\/li>\n<\/ul>\n
                  <\/span><\/div>","protected":false},"excerpt":{"rendered":"

                  [Interactive AI \uae30\uc220\ud300 \uae40\uc724\ud61c] RAG(Retrieval-Augmented Generation, \uac80\uc0c9 \uc99d\uac15 \uc0dd\uc131) \ud30c\uc774\ud504\ub77c\uc778\uc740 \ub300\uaddc\ubaa8 \uc5b8\uc5b4 \ubaa8\ub378(LLM)\uc758 \ud55c\uacc4\ub97c \ubcf4\uc644\ud558\uae30 \uc704\ud574 \uc678\ubd80 \uc9c0\uc2dd\uc744 \ud65c\uc6a9\ud558\uc5ec \ubcf4\ub2e4 \uc815\ud655\ud558\uace0 \uad00\ub828\uc131 \ub192\uc740 \uc751\ub2f5\uc744 \uc0dd\uc131\ud558\ub294 \uae30\uc220\uc785\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \uc774\ub7ec\ud55c \ud30c\uc774\ud504\ub77c\uc778\uc758 \uc131\ub2a5\uc744 \uac1d\uad00\uc801\uc73c\ub85c \ud3c9\uac00\ud558\uae30\ub780 \uc27d\uc9c0 \uc54a\uc740 \uc77c\uc785\ub2c8\ub2e4. \uc774\ubc88 \uae00\uc5d0\uc11c\ub294 \uc774\uc5d0 \ubd80\uc751\ud558\uc5ec \ub4f1\uc7a5\ud55c RAG \ud30c\uc774\ud504\ub77c\uc778 \ud3c9\uac00 \ud504\ub808\uc784\uc6cc\ud06c \uc911 \ud558\ub098\uc778 RAGAS(Retrieval-Augmented Generation Assessment)\uc5d0 \ub300\ud574 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4. RAGAS\ub780? RAGAS\ub294 RAG \ud30c\uc774\ud504\ub77c\uc778\uc758…<\/p>\n

                  <\/span><\/div>","protected":false},"author":1,"featured_media":65132,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"footnotes":""},"categories":[532,19],"tags":[667,762],"class_list":["post-65102","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-nlp","category-tech04","tag-llm","tag-rag","category-532","category-19","description-off"],"_links":{"self":[{"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/posts\/65102","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/comments?post=65102"}],"version-history":[{"count":15,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/posts\/65102\/revisions"}],"predecessor-version":[{"id":65134,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/posts\/65102\/revisions\/65134"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/media\/65132"}],"wp:attachment":[{"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/media?parent=65102"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/categories?post=65102"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/smilegate.ai\/cn\/wp-json\/wp\/v2\/tags?post=65102"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}