{"html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Ftechblog.vpoint.co.jp%2Fentry%2F2025%2F01%2F07%2F144458\" title=\"DPO(Direct Preference Optimization)\u3092\u4f7f\u3063\u3066LLM\u306e\u56de\u7b54\u3092\u8abf\u6574\u3059\u308b\u65b9\u6cd5\u3092\u8a66\u3057\u3066\u307f\u307e\u3057\u305f\u3002 - V\u30dd\u30a4\u30f3\u30c8\u30de\u30fc\u30b1\u30c6\u30a3\u30f3\u30b0\uff5cTECH LAB\u306e Tech Blog\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","provider_url":"https://hatena.blog","author_name":"miu4930","categories":["Hugging Face","AI","LLM","\u751f\u6210AI","DPO"],"url":"https://techblog.vpoint.co.jp/entry/2025/01/07/144458","published":"2025-01-07 14:44:58","version":"1.0","title":"DPO(Direct Preference Optimization)\u3092\u4f7f\u3063\u3066LLM\u306e\u56de\u7b54\u3092\u8abf\u6574\u3059\u308b\u65b9\u6cd5\u3092\u8a66\u3057\u3066\u307f\u307e\u3057\u305f\u3002","type":"rich","description":"\u3053\u3093\u306b\u3061\u306f\u3001CCCMK\u30db\u30fc\u30eb\u30c7\u30a3\u30f3\u30b0\u30b9 TECH LAB\u306e\u4e09\u6d66\u3067\u3059\u3002 \u3042\u3051\u307e\u3057\u3066\u304a\u3081\u3067\u3068\u3046\u3054\u3056\u3044\u307e\u3059\u30022025\u5e74\u304c\u306f\u3058\u307e\u308a\u307e\u3057\u305f\u3002\u4eca\u5e74\u3082\u307e\u305f\u3001\u8272\u3005\u306a\u3053\u3068\u3092\u8a66\u3057\u3066\u3044\u304d\u305f\u3044\u306a\u3068\u601d\u3044\u307e\u3059\uff01 \u6628\u5e74\u672b\u306bNeurIPS 2024\u306b\u53c2\u52a0\u3057\u3066\u304b\u3089\u3001LLM\u306e\"Post Training\"\u3068\u3044\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u306b\u8208\u5473\u3092\u6301\u3063\u3066\u3044\u307e\u3059\u3002Post Training\u306f\u3001\u65e5\u672c\u8a9e\u3067\u306f\"\u4e8b\u524d\u5b66\u7fd2\"\u3068\u547c\u3070\u308c\u3066\u3044\u308b\"Pre Training\"\u306e\u5f8c\u306b\u884c\u308f\u308c\u308bLLM\u306e\u5b66\u7fd2\u5de5\u7a0b\u3067\u3059\u3002\u4eca\u56de\u306fPost Training\u3067\u884c\u308f\u308c\u308b\u3001LLM\u306e\u51fa\u529b\u3092\u3088\u308a\u597d\u307e\u3057\u3044\u3082\u306e\u306b\u8abf\u6574\u3059\u308b\"Preference Learning\"\u3067\u4f7f\u7528\u3055\u308c\u308bDPO(Direct P\u2026","blog_title":"V\u30dd\u30a4\u30f3\u30c8\u30de\u30fc\u30b1\u30c6\u30a3\u30f3\u30b0\uff5cTECH LAB\u306e Tech Blog","blog_url":"https://techblog.vpoint.co.jp/","height":"190","provider_name":"Hatena Blog","image_url":"https://cdn-ak.f.st-hatena.com/images/fotolife/m/miu4930/20250107/20250107103132.jpg","author_url":"https://blog.hatena.ne.jp/miu4930/","width":"100%"}