[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nasa-ibm-lunar-foundation-model":3,"topics-all":39,"news-related-b9898ba4-65f5-4d80-b9ab-338b01fbd685":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":25,"news_slug":32,"published_at":33,"created_at":34,"modified_at":35,"is_published":36,"publish_type":37,"image_url":15,"view_count":38},"b9898ba4-65f5-4d80-b9ab-338b01fbd685","NASA 与 IBM 开源月球模型:极区找冰误差降 22%","NASA 与 IBM 开源 Lunar Foundation Model:ViT-B 在约 200 万张共配准月球切片上预训练,融合 11 种模态、跨 1 米与 100 米两级分辨率;极区冰前景预测误差比 SwinV2-B 基线最多低 22%,权重、代码与 SomBench 数据集均以 Apache-2.0 发布。","过去 17 年,NASA 月球勘测轨道飞行器(LRO)攒下的观测数据比其他所有 NASA 行星任务的总和还多,但行星科学家要么手工翻阅地图影像,要么为单个任务从头训练低分辨率专用模型。9 月 10 日,NASA 与 IBM 宣布开源 [Lunar Foundation Model](https:\u002F\u002Fscience.nasa.gov\u002Fscience-research\u002Fartificial-intelligence-lunar-foundation-model\u002F),换了条路走。\n\n## 极区找冰,误差比基线低 22%\n\n官方技术报告的三项下游评测里,最受关注的是极区冰前景预测:在永久阴影区这类最难观测的环境里,模型识别高潜力冰区的均方根误差相比 SwinV2-B(ImageNet)基线最多降低 22%——[model card](https:\u002F\u002Fhuggingface.co\u002Fnasa-ibm-ai4science\u002FNASA-IBM-Lunar-Foundation-Model) 原始数字 0.0293 对 0.0377,与新闻稿口径互相印证。月球冰意味着水和氧,是未来月球基地与火星任务燃料的原料,这项能力的价值不言自明。\n\n陨石坑检测上,官方新闻稿称在约 100 米尺度下只用一半训练数据就领先 SwinV2-B 近 19%;米级尺度大体相当——0.1543 对 0.1552,微弱落后。火山地貌一项新闻稿称有约 3% 优势,但 model card 自己泼了冷水:差距小于随机种子间波动,应视为相当。\n\n## 一个 ViT-B,凭什么\n\n架构并不激进:ViT-B 编码器-解码器(768 维、12 层、12 头),沿用地球观测模型 TerraMind 的掩码 token 配方。真正的增量是两处月球特化:其一,把每个切片的光照几何(入射角、方位角、切片足迹)显式序列化为上下文 token——月面外观由光照几何主导,等于把主要混淆变量直接喂给模型;其二,1 米(NAC)与 100 米(WAC)两级分辨率切片混合预训练,一套权重跨 100 倍尺度差覆盖两级分辨率。\n\n训练语料 SomBench 聚合 4 项任务 9 台仪器的 30 多个空间对齐数据层,约 200 万个共配准切片,含超 100 万张 1 米窄角相机影像与约 96.4 万张 100 米多光谱影像,并融合 GRAIL、Lunar Prospector 与日本 JAXA SELENE 数据。训练开销仅 16 张 H100、约 1100 GPU 时——不是算力军备竞赛,而是数据工程:把多任务多仪器数据对齐成机器可用的形态,才是贵的部分。\n\n## model card 里的冷水\n\n开源三件套齐全:权重、微调代码(TerraTorch 集成)与 SomBench 基准,Apache-2.0 许可。但 model card 的 Out of scope 毫不客气:模型不维护大地测量基准,生成经纬度可偏差数十度;冰前景输出是知识驱动的模糊叠加图,不是实测冰;米级预训练受限于 1095 帧共配准立体地形;也未验证过着陆选址这类操作决策。一句话:可复用的表征基座,不是成品科学仪器。\n\n## 所以呢\n\nNASA 首席科学数据官 Kevin Murphy 的说法是:收集数据只是工作的一半,真正的机会是把大规模数据变成新发现。这个模型所属家族(地球观测 Prithvi、太阳物理 Surya、如今的月球)代表一种成形的打法——不为每个科学问题从头造算法,而是维护共享基座,小样本微调适配新任务。对 AI for Science 从业者,启示很直接:领域基础模型的胜负手往往不是参数量和算力,而是把领域数据整理成可学习形态的耐心。官宣一周多,月下载量刚过两千——下一个把它微调出意外用途的人,可能还没出手。","https:\u002F\u002Fscience.nasa.gov\u002Fscience-research\u002Fartificial-intelligence-lunar-foundation-model\u002F","13624c51-a5f5-45bf-871b-b74b42216182",[11,16,19,22],{"id":12,"name":13,"slug":13,"description":14,"color":15},"9112951a-2abb-4214-b63a-385ec7afb2ba","ai-for-science","AI for Science 专题：追踪 AI 在生命科学、化学材料、物理世界模型等科学方向的关键突破",null,{"id":17,"name":18,"slug":18,"description":15,"color":15},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",{"id":20,"name":21,"slug":21,"description":15,"color":15},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":23,"name":24,"slug":24,"description":15,"color":15},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[26],{"id":27,"lang":28,"title":29,"summary":30,"content":31},"c043e03b-b40f-4d73-b777-e4d070550e0c","en","NASA and IBM Open-Source Lunar Model: Ice Error Down 22%","NASA and IBM open-sourced a lunar foundation model trained on 2M co-registered tiles; polar ice-prospectivity error drops up to 22% versus SwinV2-B.","For 17 years, NASA's Lunar Reconnaissance Orbiter has been circling the Moon, amassing more observational data than all other NASA planetary missions combined. But abundant data is not usable data: planetary scientists either pored over maps and imagery by hand or trained low-resolution, task-specific models from scratch for each question — computationally expensive and often short on scientific accuracy. On September 10, NASA and IBM announced the open-source release of the [Lunar Foundation Model](https:\u002F\u002Fscience.nasa.gov\u002Fscience-research\u002Fartificial-intelligence-lunar-foundation-model\u002F), changing that path.\n\n## Polar ice prospectivity: error down up to 22% versus baseline\n\nOf the three downstream evaluations in the official technical report, the headline is polar ice prospectivity: in permanently shadowed regions — the hardest environments to observe — the model reduces root-mean-square error in identifying high-potential ice areas by up to 22% compared with the widely used SwinV2-B (ImageNet) baseline. The [model card](https:\u002F\u002Fhuggingface.co\u002Fnasa-ibm-ai4science\u002FNASA-IBM-Lunar-Foundation-Model) lists the raw numbers, 0.0293 versus 0.0377, matching the press release. Lunar ice means water and oxygen — feedstock for a future Moon base and fuel for Mars missions — so this capability matters directly.\n\nOn crater detection, the press release says that at context scale (about 100 meters) the model leads SwinV2-B by nearly 19% using only half the training data; at meter scale it is roughly on par with the strongest baseline — 0.1543 versus 0.1552 in the model card, marginally behind. For irregular mare patches, the release claims an edge of about 3%, but the model card itself pours cold water: the margin is smaller than seed-level spread, so treat it as a tie. The NASA page also shows a case where a SpaceX rocket body punched a fresh crater onto the surface; on post-impact imagery that was excluded from pretraining, the model flagged the newly formed crater — exactly the change-detection capability that future lunar-activity monitoring needs.\n\n## One ViT-B — why it works\n\nThe architecture is not radical: a ViT-B encoder-decoder (768-dim, 12 layers, 12 heads) following the masked-token recipe of TerraMind, an Earth-observation foundation model. The real additions are two lunar-specific moves. First, per-tile illumination geometry — solar incidence and azimuth angles, tile footprints — is explicitly sequence-tokenized as context, because lunar surface appearance is governed more by illumination geometry than by intrinsic surface variation; the dominant confound is handed to the model instead of being left for it to recover. Second, tiles at 1-meter (NAC) and 100-meter (WAC) resolution train in a single mixed batch at native resolution, so one set of weights spans both resolution families across a 100x scale gap.\n\nThe training corpus, SomBench, aggregates more than 30 spatially aligned layers from nine instruments across four missions into roughly 2 million co-registered tile bundles — over 1 million 1-meter Narrow Angle Camera images and about 964,000 100-meter multispectral images, plus data from NASA's GRAIL gravity mission, Lunar Prospector, and JAXA's SELENE. Training cost: 16 H100 GPUs, about 1,100 GPU-hours. This is not an arms race in compute; it is a data-engineering exercise, and aligning multi-mission, multi-instrument data into machine-learnable form is the genuinely expensive part.\n\n## The cold water in the model card\n\nThe open-source trio is complete: weights, fine-tuning code integrated with the TerraTorch toolkit, and the SomBench dataset and benchmark collections, all under Apache-2.0. But the Out-of-scope section is blunt. The model maintains no geodetic reference frame — generated lat\u002Flon can drift by tens of degrees. Ice-prospectivity outputs regress a knowledge-driven fuzzy-overlay map, not measured ice. Meter-scale pretraining is limited to 1,095 frames with co-registered 3-meter stereo DTMs — globally distributed, but not globally dense. And it is not validated for operational decisions such as landing-site certification. In short: a reusable representation backbone, not a finished scientific instrument.\n\n## So what\n\nKevin Murphy, NASA's chief science data officer, framed it this way: collecting data is only part of the job — the real opportunity is turning large-scale data into new discoveries. The family this model joins, Prithvi for Earth observation and Surya for heliophysics, and now the Moon, represents an emerging playbook: instead of building a new algorithm for every scientific question, maintain a shared backbone and fine-tune it to new tasks with small labeled sets. For AI-for-Science practitioners the lesson is direct: in domain foundation models, the decisive factor is often not parameter count or compute, but the patience to curate domain data into learnable form. A bit over a week after the announcement, monthly downloads sit at just above two thousand — whoever fine-tunes it into something unexpected may simply not have shown up yet.","nasa-ibm-lunar-foundation-model","2026-09-18T13:10:34Z","2026-09-18T13:10:40.208380Z","2026-09-18T13:10:40.208388Z",true,"agent",85,[40,48],{"slug":13,"tag_slug":13,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":36,"created_at":46,"modified_at":47},"AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":36,"created_at":55,"modified_at":56},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":58},[59,64,69,74,79,84],{"id":60,"title":61,"news_slug":62,"published_at":63},"c99751d5-418e-49d5-99d3-e43b84c80ec7","IBM与NASA开源月球基础模型:Lunar Foundation Model","nasa-ibm-lunar-foundation-model-sombench","2026-09-19T09:30:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"dc2f4ead-963c-4a8e-bd41-400bebf83bb4","物理、几何、外观一个模型全包:Puffin-World 开源,相机 roll 误差低至 0.26°","puffin-world-native-3d-world-states","2026-09-06T19:09:41+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"7cc1b87c-fe06-495a-9c01-9516d0c16354","腾讯混元 HunyuanImage-3.0 全面开源：80B 总参 \u002F 13B 激活的自回归 MoE，把多模态理解和生图拉到同一框架","tencent-hunyuanimage-3-moe-autoregressive","2026-08-05T01:00:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"009ab609-27bf-4c34-92a6-2ce53eb25b69","Qwen 3.5 原生多模态新思路：DeepStack Vision Transformer 多层特征融合解析","qwen-3-5-deepstack-vit-multilayer","2026-05-09T13:10:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"3de0fc02-b400-4c9a-a613-426b5004b27b","NASA 和 IBM 的 Prithvi：首个在轨 AI 地理空间基础模型开启遥感新范式","nasa-ibm-prithvi-on-orbit-geospatial","2026-05-07T11:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"12c67d52-17a2-4df5-8386-35d18ffd221a","JEPA-Anything:一套预测框架打通七个领域,湿实验也给了背书","jepa-anything-orthogonal-predictive-factorization","2026-09-19T23:10:37+00:00"]