https://blog.youkuaiyun.com/PlagueJudgezh-cnhttps://blog.youkuaiyun.com/https://blog.youkuaiyun.com/PlagueJudge/article/details/146462017https://blog.youkuaiyun.com/PlagueJudge/article/details/146462017PlagueJudgeSun, 23 Mar 2025 21:07:59 +0800https://blog.youkuaiyun.com/PlagueJudge/article/details/146404659https://blog.youkuaiyun.com/PlagueJudge/article/details/146404659PlagueJudgeThu, 20 Mar 2025 19:42:39 +0800https://blog.youkuaiyun.com/PlagueJudge/article/details/146325104https://blog.youkuaiyun.com/PlagueJudge/article/details/146325104PlagueJudgeMon, 17 Mar 2025 20:36:44 +0800 冷启动SFT、推理RL（用于合成微调数据）-> RFT&SFT（Base模型微调）-> 全场景RL（SFT模型RL）-> DeepSeek R1）大模型：GPT-3（模型规模达到了1750亿参数、涌现出上下文学习能力）、CodeX（代码数据训练、推理与代码合成能力）、GPT-3.5、GPT-4（推理能力显著提升，建立可预测的训练框架、可支持多模态信息的大语言模型）推理大模型：o-Series（类似人类的“慢思考”过程）]]>https://blog.youkuaiyun.com/PlagueJudge/article/details/146282024https://blog.youkuaiyun.com/PlagueJudge/article/details/146282024PlagueJudgeSat, 15 Mar 2025 17:02:50 +0800https://blog.youkuaiyun.com/PlagueJudge/article/details/146213896https://blog.youkuaiyun.com/PlagueJudge/article/details/146213896PlagueJudgeWed, 12 Mar 2025 20:22:12 +0800