[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-chenguang-1-satellite-llm":3,"news-related-59a18390-a856-4251-8407-96e641cf74bc":36},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"59a18390-a856-4251-8407-96e641cf74bc","\"辰光一号\"把大模型搬上天:国内首次航天垂直大模型在轨训练开启","7月24日,中科天塔自研的航天垂直大模型\"华山\"卫星健康管理模块随\"辰光一号\"技术试验卫星在酒泉发射升空,开展在轨大模型训练与航天器智能管控全流程验证。这是国内首次把\"训练-推理\"闭环搬到卫星上。\n\n**技术看点:**\"华山\"是垂直领域模型,针对卫星遥测、姿态、轨道数据做了深度微调,在辐射、真空、热循环环境下运行,对鲁棒性提出全新要求。关键创新是\"训练在轨\"——模型权重可根据在轨数据继续微调,形成\"飞行中学习-决策-验证\"闭环,而非传统\"地面训好、星上推理\"。\n\n**行业影响:**这并非孤例。NASA与IBM今年5月联合发布Prithvi,中美不约而同把大模型搬上天,本质是抢\"轨道算力\"这条新赛道。对国内大模型生态,航天场景提供了三个稀缺价值:极端环境训练数据、边缘部署工程范式、军工\u002F通信\u002F气象高价值入口。地面\"百模大战\"还在卷参数,天上的模型必须卷\"小而稳\",这可能催生新的模型范式。如果试验跑通,大模型\"云-端-星\"三级部署的拼图就完整了。","https:\u002F\u002F36kr.com\u002Fnewsflashes\u002F3909268957582471","5e4fd3d1-9cb4-44a6-bae5-9ffb449c05c1",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",{"id":18,"name":19,"slug":19,"description":13,"color":13},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":28},"d6ffabbc-aeb9-49b9-a144-2366b800c6b0","en","Chenguang-1 starts training LLMs in orbit","On July 24, the \"Huashan\" satellite health-management module — an aerospace vertical large model developed by CAS Tianlian — was launched aboard the \"Chenguang-1\" technology test satellite from Jiuquan, beginning in-orbit large-model training and end-to-end verification of spacecraft intelligent control. This is the first time China has put a \"training-inference\" closed loop onto a satellite. **Technical highlight:** \"Huashan\" is a vertical-domain model, deep-fine-tuned on satellite telemetry, attitude, and orbit data, and running in radiation, vacuum, and thermal-cycling environments — placing entirely new demands on robustness. The key innovation is \"in-orbit training\" — model weights can continue to be fine-tuned on in-orbit data, forming a \"learn in flight — decide — verify\" closed loop, rather than the traditional \"train on the ground, infer in orbit\". **Industry impact:** This isn't an isolated case. NASA and IBM jointly released Prithvi in May, and the US and China have independently taken large models to space — essentially racing for the new track of \"orbital compute\". For China's large-model ecosystem, the aerospace scenario offers three scarce values: extreme-environment training data, edge-deployment engineering paradigms, and high-value entry points into military \u002F communications \u002F meteorology. While the \"hundred-model war\" on the ground is still racing on parameters, the models in space must race on \"small and stable\" — which may give birth to a new model paradigm. If the experiment succeeds, the \"cloud–edge–satellite\" three-tier deployment of large models will be complete.","chenguang-1-satellite-llm","2026-07-25T00:00:00Z","2026-07-24T20:05:12.035137Z","2026-08-19T02:08:40.142862Z",true,"agent",111,{"items":37},[38,43,48,53,58,63],{"id":39,"title":40,"news_slug":41,"published_at":42},"49d19ba1-8f45-475c-bed1-a69dc353523e","字节跳动用 10 万亿参数下注：规模赛跑与张一鸣的「不蒸馏」表态","bytedance-10t-mythos-zhangyiming-no-distill-2026-08","2026-08-08T00:00:00+00:00",{"id":44,"title":45,"news_slug":46,"published_at":47},"5f5bd5f2-9a02-470b-aa25-3f27fb9bb093","字节跳动正训练 10 万亿参数模型，规模对标 Anthropic Mythos 5","bytedance-10t-parameter-model-ft","2026-08-07T09:30:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"86c258f4-5fd5-45fa-9fe5-60dbb585bfff","DeepSeek 梁文锋路线图:持续学习才是 Agent 之后的真瓶颈","deepseek-liang-wenfeng-roadmap","2026-07-24T08:30:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"adbe213f-d27c-41a7-803c-c5823e1a63fd","字节跳动 Seed STEM 科学家计划启动:把豆包算力+模型搬到 STEM 学科的最前线","bytedance-seed-stem-scientist","2026-07-23T08:00:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"8492350c-74ff-4ce3-bda5-c17b95b9e385","清华 CausalMix 把 LLM 数据混合从回归问题改成因果推断：换数据池不再重跑 proxy","tsinghua-causalmix-data-mix","2026-07-07T18:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"0d8fdf45-4585-47c0-9e78-3652e318b156","Apple Intelligence 中国版落地:通义千问接管语言 AI,百度负责视觉搜索","apple-intelligence-china-qwen-baidu-2026","2026-08-25T12:00:00+00:00"]