{
    "componentChunkName": "component---src-templates-post-tsx",
    "path": "/bert_fp/",
    "result": {"data":{"logo":null,"markdownRemark":{"html":"<h1>Fine-grained Post-training for Improving Retrieval-based Dialogue Systems Paper Review</h1>\n<img src=\"https://d3i71xaburhd42.cloudfront.net/9f359007e9af7e49e95b3bba3c8621c6fa2f8cca/4-Figure1-1.png\">\n<ul>\n<li>Paper: <a href=\"https://aclanthology.org/2021.naacl-main.122/\">https://aclanthology.org/2021.naacl-main.122/</a></li>\n<li>Code: <a href=\"https://github.com/hanjanghoon/BERT_FP\">https://github.com/hanjanghoon/BERT_FP</a></li>\n</ul>\n<p>NAACL 2021에 억셉된 논문이다. 서강대 NLP 연구실, LG AI Research 팀에서 작성했다. Dialogue Retrieval 관련 논문인데,\n방법이 심플하면서도 성능 좋다. (Ubuntu Dialogue 등 몇 데이터셋에서 SOTA를 찍었다)</p>\n<p><strong>Utterance Relevance Classification (URC)</strong> 이 핵심인데, pre-training 된 모델을 가져와서 추가적으로 post-training을 시키는데,\nMLM (Masked Language Modeling) + NSP (Next Sentence Prediction) 중 NSP를 dialogue 태스크에 맞게 조금 변형한다.</p>\n<p>[Context + &#x3C;|sep|> +  Response] 형식의 입력을 주고, 이 Response가 1. Correct Response 2. Random Utterance 3. Random Utterance in the same dialogue 중\n어느 클래스에 속하는지 3개로 분류하는 식으로 학습한다. 이렇게 하면 기존의 Context - Response를 두고 적합한지 (Positive),\n부적합한지 (Negative) 를 판단하는 binary classification 태스크로 문제를 풀 때보다\n주어진 대화 데이터를 더 효과적으 활용한다는 면에서 데이터 오그멘테이션의 효과도 있게 된다.</p>\n<p>이 post-training은 MLM Loss (dynamic masking) + URC Loss로 학습이 되며,\n이후 기존 response selection 태스크로 파인튜닝해서 사용 가능하다.</p>","htmlAst":{"type":"root","children":[{"type":"element","tagName":"h1","properties":{},"children":[{"type":"text","value":"Fine-grained Post-training for Improving Retrieval-based Dialogue Systems Paper Review"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"img","properties":{"src":"https://d3i71xaburhd42.cloudfront.net/9f359007e9af7e49e95b3bba3c8621c6fa2f8cca/4-Figure1-1.png"},"children":[]},{"type":"text","value":"\n"},{"type":"element","tagName":"ul","properties":{},"children":[{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"text","value":"Paper: "},{"type":"element","tagName":"a","properties":{"href":"https://aclanthology.org/2021.naacl-main.122/"},"children":[{"type":"text","value":"https://aclanthology.org/2021.naacl-main.122/"}]}]},{"type":"text","value":"\n"},{"type":"element","tagName":"li","properties":{},"children":[{"type":"text","value":"Code: "},{"type":"element","tagName":"a","properties":{"href":"https://github.com/hanjanghoon/BERT_FP"},"children":[{"type":"text","value":"https://github.com/hanjanghoon/BERT_FP"}]}]},{"type":"text","value":"\n"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"NAACL 2021에 억셉된 논문이다. 서강대 NLP 연구실, LG AI Research 팀에서 작성했다. Dialogue Retrieval 관련 논문인데,\n방법이 심플하면서도 성능 좋다. (Ubuntu Dialogue 등 몇 데이터셋에서 SOTA를 찍었다)"}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"element","tagName":"strong","properties":{},"children":[{"type":"text","value":"Utterance Relevance Classification (URC)"}]},{"type":"text","value":" 이 핵심인데, pre-training 된 모델을 가져와서 추가적으로 post-training을 시키는데,\nMLM (Masked Language Modeling) + NSP (Next Sentence Prediction) 중 NSP를 dialogue 태스크에 맞게 조금 변형한다."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"[Context + <|sep|> +  Response] 형식의 입력을 주고, 이 Response가 1. Correct Response 2. Random Utterance 3. Random Utterance in the same dialogue 중\n어느 클래스에 속하는지 3개로 분류하는 식으로 학습한다. 이렇게 하면 기존의 Context - Response를 두고 적합한지 (Positive),\n부적합한지 (Negative) 를 판단하는 binary classification 태스크로 문제를 풀 때보다\n주어진 대화 데이터를 더 효과적으 활용한다는 면에서 데이터 오그멘테이션의 효과도 있게 된다."}]},{"type":"text","value":"\n"},{"type":"element","tagName":"p","properties":{},"children":[{"type":"text","value":"이 post-training은 MLM Loss (dynamic masking) + URC Loss로 학습이 되며,\n이후 기존 response selection 태스크로 파인튜닝해서 사용 가능하다."}]}],"data":{"quirksMode":false}},"excerpt":"Fine-grained Post-training for Improving Retrieval-based Dialogue Systems Paper Review Paper: https://aclanthology.org/2021.naacl-main.12…","fields":{"readingTime":{"text":"2 min read"}},"frontmatter":{"title":"Sooftware NLP - Fine-grained Post-training for Improving Retrieval-based Dialogue Systems Paper Review","userDate":"18 December 2021","date":"2021-12-18T10:00:00.000Z","tags":["nlp","paper"],"excerpt":null,"image":{"childImageSharp":{"gatsbyImageData":{"layout":"fullWidth","backgroundColor":"#f8f8f8","images":{"fallback":{"src":"/static/ad2df79ee3a7cc5e1d4fe4abd604473d/3a1cc/bert_fp.png","srcSet":"/static/ad2df79ee3a7cc5e1d4fe4abd604473d/69607/bert_fp.png 750w,\n/static/ad2df79ee3a7cc5e1d4fe4abd604473d/27438/bert_fp.png 1080w,\n/static/ad2df79ee3a7cc5e1d4fe4abd604473d/3a1cc/bert_fp.png 1234w","sizes":"100vw"},"sources":[{"srcSet":"/static/ad2df79ee3a7cc5e1d4fe4abd604473d/8c326/bert_fp.webp 750w,\n/static/ad2df79ee3a7cc5e1d4fe4abd604473d/32767/bert_fp.webp 1080w,\n/static/ad2df79ee3a7cc5e1d4fe4abd604473d/d399b/bert_fp.webp 1234w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.526742301458671}}},"author":[{"id":"Soohwan Kim","bio":"Co-founder/A.I. engineer at TUNiB.","avatar":{"children":[{"gatsbyImageData":{"layout":"fullWidth","backgroundColor":"#282838","images":{"fallback":{"src":"/static/a9e6b445142b247ee4cfa66155398bb2/0d6f4/soohwan.png","srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/248f9/soohwan.png 40w,\n/static/a9e6b445142b247ee4cfa66155398bb2/fd435/soohwan.png 80w,\n/static/a9e6b445142b247ee4cfa66155398bb2/0d6f4/soohwan.png 120w","sizes":"100vw"},"sources":[{"srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/e7f45/soohwan.webp 40w,\n/static/a9e6b445142b247ee4cfa66155398bb2/589ec/soohwan.webp 80w,\n/static/a9e6b445142b247ee4cfa66155398bb2/71a38/soohwan.webp 120w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.6833333333333333}}]}}]}},"relatedPosts":{"totalCount":50,"edges":[{"node":{"id":"6529d72e-80e7-5d61-a672-d66276a3f641","excerpt":"MoE(Mixture of Experts) 기초부터 DeepSeek 혁신까지 작년 이맘때쯤 DeepSeek-V3가 저비용으로 엄청난 성능을 보이면서 화제가 되었습니다. 그 핵심 기술인 MoE(Mixture of Experts…","frontmatter":{"title":"MoE(Mixture of Experts) 기초부터 DeepSeek 혁신까지","date":"2026-01-15T01:11:55.000Z"},"fields":{"readingTime":{"text":"8 min read"},"slug":"/developed-moe/"}}},{"node":{"id":"8b49d3ef-4ce3-568c-9d71-240ff17fc3e0","excerpt":"BERT는 사실 Diffusion 모델이였다?! 최근 굉장히 흥미로운 글을 읽게되어 공유합니다. 원문 : link BERT와 Diffusion이 같은 방식이다?! NLP 연구자들에게 BERT는 너무 익숙한 모델입니다. 201…","frontmatter":{"title":"BERT는 사실 Diffusion 모델이였다?!","date":"2025-10-21T12:00:00.000Z"},"fields":{"readingTime":{"text":"8 min read"},"slug":"/bert_diffusion/"}}},{"node":{"id":"7c7e1676-ea84-58de-970c-ddec9aa10660","excerpt":"RLHF는 수다쟁이를 만든다?! (Does RLHF Breed Verbose Chatterboxes?!) RLHF(Reinforcement Learning from Human Feedback)는 OpenAI의 ChatGPT…","frontmatter":{"title":"RLHF는 수다쟁이를 만든다?! (Does RLHF Breed Verbose Chatterboxes?!)","date":"2024-03-13T01:11:55.000Z"},"fields":{"readingTime":{"text":"7 min read"},"slug":"/rlhf-vervosity/"}}},{"node":{"id":"2b9d3e22-1796-5fab-878d-5941d2e76e9d","excerpt":"LLM Paper Abstract - 2023.12 LLM…","frontmatter":{"title":"LLM Paper Abstract - 2023.12","date":"2024-01-05T10:00:00.000Z"},"fields":{"readingTime":{"text":"5 min read"},"slug":"/llm-abs-202312/"}}},{"node":{"id":"f43fa33b-917b-5c00-8462-937d99592ad7","excerpt":"What it MoE? (Mixture of Experts) 현존 최강 LLM인 GPT-4에서 “MoE (Mixture of Experts)” 방식을 채택하여 사용하고 있다고 알려졌는데요, 최근 AI계의 뜨거운 감자 Mistral AI…","frontmatter":{"title":"What is MoE? (Mixture of Experts)","date":"2023-12-22T01:11:55.000Z"},"fields":{"readingTime":{"text":"9 min read"},"slug":"/moe/"}}}]}},"pageContext":{"slug":"/bert_fp/","prev":{"excerpt":"K(orea) Startups Intro 한국에 있는 테크 스타트업들 리스트와 간단한 설명을 기록합니다. ‘내가 아는 스타트업들이 얼마나 될까?’ 라는 간단한 생각을 시작으로 기록하게 됐습니다. 허훈님의 nlp-startups, 염기웅님의 Korea…","frontmatter":{"title":"K-Startups (Korea AI Startups)","tags":["record"],"date":"2021-12-05T10:00:00.000Z","draft":false,"excerpt":null,"image":{"childImageSharp":{"gatsbyImageData":{"layout":"fullWidth","placeholder":{"fallback":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAMCAYAAABiDJ37AAAACXBIWXMAAAsTAAALEwEAmpwYAAAChklEQVQoz3WT3UuTURzH9y90WQleSHRRISmx4evzTMWLJIKIEoxuuqk/oUjoIgiDSDJ7W5YvY5K5ufdpmIg6tWdOzYxS5151e/ZMSzC6CPrE82zTSXVxON9zzu98z+93ft+vTnDLiB4ZwZNG1HAa0aug7gu59V9YnV0pLV4o3POk0RUGqrNRPXQkMKo4f1F9JIeNvgyieu6II9ii1LlTCB4lm4xbzhLmyaqcKUTrOnrzZyocSeo9MrX2ODXOJNVvVqnuW8LQu0SlNcJRc5gjfWGKTEvU2BN7SelUVjUTvT3Fzakkn9K7NNrDFD+RtEwb7FFO94co7ZrnTMd7xG6Jokd+LvbP0dI1yimThL5/BaNrc7/keq9CmS3JjeE1/IF5zrljXPBt0DaTwGCNITo3afZEaHrtp/GVn8OdEtU9QRofD1NsWqTClfuebMlprbTSgSgt1o+0j8xxL7jFUynGg9kEbYvf6Qn9xPY1w4uJZbqDcW5Np2gdW8WyLHM3oFA7uI7gSmpl67RGOBLoB8LckTKs7/7m/vgKrWYvJvc4/ojCbPIHY9Ed2he2eLaoMBLZYWwtzcL2L24HtikzSRidBSWrMql2bHLeE+f62Ablli+cfTlBU/c0zb2TXLZI1FmWaOwNcHUgyCVXiCujKa5NfcNgjVM1FN9vyl6X3TI1LhnBmaT0bYxKa5QK8zLlD99R2bNASecMxzsmOdbp51DbKAbzMkZbJCspTTpZHevEHBByWmvwKdT7FIxeJRs8FEUYXKfOuYFxMMRJyxolpnlOPP9AnSOBkDNBnkMnFKr9H0P9Ds05moMy1Lhl9E75AIlYYA7dngvyttPG/x8Rcw4Sc1YTD9xL8wcjuLOuXOFpNAAAAABJRU5ErkJggg=="},"images":{"fallback":{"src":"/static/bb766eb0837f5b7957ea04f792e7cab4/d2fbe/startup.png","srcSet":"/static/bb766eb0837f5b7957ea04f792e7cab4/2512b/startup.png 750w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/45604/startup.png 1080w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/8b6b1/startup.png 1366w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/d2fbe/startup.png 1600w","sizes":"100vw"},"sources":[{"srcSet":"/static/bb766eb0837f5b7957ea04f792e7cab4/80f13/startup.webp 750w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/4f9a6/startup.webp 1080w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/af4ab/startup.webp 1366w,\n/static/bb766eb0837f5b7957ea04f792e7cab4/d0dfe/startup.webp 1600w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.6}}},"author":[{"id":"Soohwan Kim","bio":"Co-founder/A.I. engineer at TUNiB.","avatar":{"childImageSharp":{"gatsbyImageData":{"layout":"fullWidth","placeholder":{"fallback":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAOCAYAAAAvxDzwAAAACXBIWXMAABYlAAAWJQFJUiTwAAADjElEQVQ4y23Oy48TdQDA8TExEaTttvSBfdHddru7bWc6bWd+8+i0u23n0SdlBRQXZGHDgiiLDwgqXjCIJEaC8cLBmHgwUU9EYzRqQuLJi/4NJB6NF+8evgbOHr7XT76SM5ix7s4QnRHnnm9yZzvBTIswN8JcP5bC67ZoWi7CdtFsn5bloZo+dSNAMYbIIqCmD6mJIVXdR3J6I7r9IQ3T59pZlW+uxzndCbPdC/HZXoq3TikowsfqBOjOmFZ7SrM9p2kfRbXn1O0jKNbsSbJ5BMnu9Gl3+qiiz4OPSvz0QRLHFGwOZH6+neLHT5YRbRerM0J0JrTas/8FZXNKzZgg2es+ZtvDdbs8+iHJl+9lWVP7OG2br26k+PvXJCfmFqoYIpwhTXtMwxpTN0coxhjFmCKbE2rGmKoYPQY9qi2fyzsq//4e5sNXi5TlLqbd5t6lPH/9kuDTG2usqgGi7dO0hzSsIXUzQDaGKGKMLEZUxZCKCJB02+PUpMH9vQwP7ybYnSxyfNrjwumA3dEiD94v8cfnK7xzWsbpujQtH9X0qBs+iuEjC5+a7lPVPSqai9SyelyYFbjkh5hqC3TkKOutLON2nkYpgtuMcT6IszdPslo8SLlcpmUH1A0XRQyQhUtNuFQ1l4o2QFI1h3OTVXbcg+grIfSVBRqlEGopgrYSRi0e4KiIse1mScSeopA/iLADmoZL3Rgg631q+oCq1qeq95FOzhQefpzn7eMpBmqcoBVj04pzwkky0eNMRYIdL8F3twusLe2nslrE6g5pGX1U8/FhD1nvUW1tUJK7SH9+n+DR1zF2vQSvH0lz88wiV45V2J1WuHqsxLXNAnuzNJdfWmK5/BylUgFhb6DZAxR9g7rpIos+Wttja2uE9M9vYb69k+DqfJErWzLOuoVqdKlpDo4jODms4zlFckuHObx4iHwuRi6fYqmYY201x/LqMrnFDNl8gRvvukhNdT9394rcOlNFMxrUhYFmW9imTEsp0mhWKC2nWSokyWSiFA7HSKcXiEb34fayyLVniUQk8rmn2drKIqlqiNcuVrh2ymQS1Nh8sc/GoMEbl7u8eb5D1yqTzkRJJsOkkiGymQVShyLE4yHWKnEq1RjpTIhmI8r2y3mkyloI38uha1n6bpVXrkw5+4LD/S/2uHnrIopaJhZ9hmQqwqHkAXJP8BDhyD46ZoWdMxOGcx/PNfECk/8AbxXdRjRliPoAAAAASUVORK5CYII="},"images":{"fallback":{"src":"/static/a9e6b445142b247ee4cfa66155398bb2/7cf1f/soohwan.png","srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/34f77/soohwan.png 750w,\n/static/a9e6b445142b247ee4cfa66155398bb2/a94f6/soohwan.png 1080w,\n/static/a9e6b445142b247ee4cfa66155398bb2/7cf1f/soohwan.png 1148w","sizes":"100vw"},"sources":[{"srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/38420/soohwan.webp 750w,\n/static/a9e6b445142b247ee4cfa66155398bb2/7470d/soohwan.webp 1080w,\n/static/a9e6b445142b247ee4cfa66155398bb2/b5ef6/soohwan.webp 1148w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.6829268292682927}}}}]},"fields":{"readingTime":{"text":"26 min read"},"layout":"","slug":"/k-startups/"}},"next":{"excerpt":"Slack Bot Python과 Slack API를 사용하여, 특정 채널에 자동으로 글을 올리거나 댓글을 달아주는 슬랙봇을 만들어보겠습니다. 두 개의 과정으로 진행되는데, 첫 번째는 Slack API에 bot을 등록하는 것이고 두 번째는 등록된 bot…","frontmatter":{"title":"Slack Bot","tags":["toolkit"],"date":"2021-12-23T10:00:00.000Z","draft":false,"excerpt":null,"image":{"childImageSharp":{"gatsbyImageData":{"layout":"fullWidth","placeholder":{"fallback":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABQAAAAKCAYAAAC0VX7mAAAACXBIWXMAAAsTAAALEwEAmpwYAAABH0lEQVQoz61SPU/DMBDNTwR1RZVQmiAY+Qc0v4EOpBJLWzohBjYGxqx8SUikggGJSDSobp0ogtj3UOzGJE2BpSedz/dsv3tn28KGzSonRFSMKIIgglS5xqWUKlZ9FWsQKjJDrE2SVAdLE1JCCIG6CKwn1AL1wuWM4Za/m03hZALGmMmTJEEYhrrosmBJav2o0S2fv35i6/4Zhw+neJm9YTQY4cjrout5OPF9+P0+BsMhdm0bQRDUSOsKl+D04gbTfQ/sroP5xxNs9wDungvbcdBxHOy02zju9bDdauFsPNZXIURTYdGugr4AuroG+KOCF4s50jRVHscxsixDFEXI8xycc9Nug5CqpJX8P6OVPda6FyPVgjSv/pf/qnBTH/sbyKT7kUtH55oAAAAASUVORK5CYII="},"images":{"fallback":{"src":"/static/6ae60f09d8f0de6ab000dc2fe6bee740/00aaa/slack.png","srcSet":"/static/6ae60f09d8f0de6ab000dc2fe6bee740/00aaa/slack.png 412w","sizes":"100vw"},"sources":[{"srcSet":"/static/6ae60f09d8f0de6ab000dc2fe6bee740/1e77b/slack.webp 412w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.4878640776699029}}},"author":[{"id":"Soohwan Kim","bio":"Co-founder/A.I. engineer at TUNiB.","avatar":{"childImageSharp":{"gatsbyImageData":{"layout":"fullWidth","placeholder":{"fallback":"data:image/png;base64,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"},"images":{"fallback":{"src":"/static/a9e6b445142b247ee4cfa66155398bb2/7cf1f/soohwan.png","srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/34f77/soohwan.png 750w,\n/static/a9e6b445142b247ee4cfa66155398bb2/a94f6/soohwan.png 1080w,\n/static/a9e6b445142b247ee4cfa66155398bb2/7cf1f/soohwan.png 1148w","sizes":"100vw"},"sources":[{"srcSet":"/static/a9e6b445142b247ee4cfa66155398bb2/38420/soohwan.webp 750w,\n/static/a9e6b445142b247ee4cfa66155398bb2/7470d/soohwan.webp 1080w,\n/static/a9e6b445142b247ee4cfa66155398bb2/b5ef6/soohwan.webp 1148w","type":"image/webp","sizes":"100vw"}]},"width":1,"height":0.6829268292682927}}}}]},"fields":{"readingTime":{"text":"7 min read"},"layout":"","slug":"/slack_bot/"}},"primaryTag":"nlp"}},
    "staticQueryHashes": ["3170763342","3229353822"]}