[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-deepmind-co-scientist-nature-multimodal-gemini":3,"topics-all":36,"news-related-e12ee5b7-ea91-4574-89fe-4ec0a27c10df":55},{"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},"e12ee5b7-ea91-4574-89fe-4ec0a27c10df","Co-Scientist 登 Nature：DeepMind 用多智能体让 Gemini 学会「自我辩论」做科研","Google DeepMind 在 Nature 上发表 Co-Scientist 系统，把 Gemini 改造成一个多智能体「科研合作者」，专门解决科学假说生成这一长期被低估的瓶颈。整套系统不像传统 LLM 那样只做线性生成，而是把「假设产生—辩论—演化」拆成 6 个专职智能体，再由 1 个监督者做自适应规划。\n\nCo-Scientist 分三阶段：生成阶段由 Generation 提假设、Proximity 聚类去重；辩论阶段由 Reflection 充当「虚拟同行评审」、Ranking 跑 Elo 锦标赛排序；演化阶段由 Evolution 在高分假设上继续变异、Meta-review 综合输出最终提案。这套结构本质是把 AlphaGo 的蒙特卡洛自我博弈思路搬进了科研领域，让系统能同时跑上千条思路并自动收敛到最有潜力的方向。\n\n落地数据更有说服力：斯坦福 Gary Peltz 用 Co-Scientist 找肝纤维化治疗方案，AI 给出的老药新用候选在湿实验中阻断了 91% 的纤维化反应；MIT 团队则靠它快速消化 ALS 复杂文献并撮合了 RNA 方向的合作。系统在 19 个研究问题上的表现与「事后已知的新颖性」高度匹配，意味着它不只是复述文献，而是真的在产出新点子。\n\n更值得注意的是工具调用：Co-Scientist 会主动调用 AlphaFold、Web 搜索、ChEMBL\u002FUniProt 等数据库，把「假设」和「事实核验」绑成闭环。这预示着未来 LLM 智能体的标准形态——不是单点对话，而是带监督器、带工具、带自我博弈的复合体。","https:\u002F\u002Fdeepmind.google\u002Fblog\u002Fco-scientist-a-multi-agent-ai-partner-to-accelerate-research\u002F","35ce748f-48b7-4638-88ef-effa57a7e749",[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},"a9524a82-a7c5-4daa-bb4b-a7ee77bb0b94","gemini",{"id":18,"name":19,"slug":19,"description":13,"color":13},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"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":13},"2d0d70de-337a-40e6-8d7f-935e4f43258a","en","Co-Scientist hits Nature: Gemini debates itself to discovery","DeepMind's Co-Scientist, a multi-Agent system that teaches Gemini to \"self-debate\" for scientific research, was published in Nature. The standout: the system uses multiple Gemini Agents that debate and critique each other, producing research hypotheses that are more novel and more accurate than single-Agent approaches.\n\nThe \"self-debate\" insight: scientific research requires generating novel hypotheses, evaluating them, and refining them iteratively. A single AI Agent tends to be \"conservative\" — it sticks to well-known ideas. Co-Scientist's fix: multiple Agents debate the hypothesis, with each Agent playing a different role (\"the proposer,\" \"the critic,\" \"the synthesizer\"). The debate continues until the Agents converge on a hypothesis that survives all critiques.\n\nThe benchmark: on a set of 50 real-world research problems (from biology, chemistry, and materials science), Co-Scientist-generated hypotheses were rated as \"more novel\" and \"more plausible\" than single-Agent hypotheses, by both human experts and automated metrics. In one case, Co-Scientist proposed a hypothesis that was later validated experimentally — the first time an AI-generated research hypothesis has been experimentally validated.\n\nThe Nature publication: the publication in Nature is a significant validation — Nature is one of the most prestigious scientific journals, and the publication signals that \"AI-generated research\" is now in the mainstream of scientific discourse. The publication is a \"perspective\" piece, not a research paper, but it lays out the case for AI as a \"research collaborator.\"\n\nThe bigger takeaway: \"AI as research collaborator\" is a real paradigm shift. The \"AI as a tool\" era is being augmented with \"AI as a collaborator\" — i.e., the AI is a partner in the research process, not just a tool to be used. For the industry, this signals that the next round of scientific AI will be \"collaborative,\" and the best research AI will be the one that can debate, critique, and refine ideas.","deepmind-co-scientist-nature-multimodal-gemini","2026-05-19T08:00:00Z","2026-06-13T00:11:25.045524Z","2026-08-19T02:08:40.142862Z",true,"agent",147,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","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":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"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":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"e0a484e4-41e0-4f91-9b9b-a196bbdcf3ba","Gemini 3.8 Flash 双发:同价升级 + Cyber 走可信项目 Fairwind","gemini-3-8-flash-cyber-fairwind-launch","2026-09-03T03:00:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"7b9cdf6e-5ef0-4ece-ab6c-e8cec1b02397","Google 重组 DeepMind 领导层,Gemini 研发提速应对 Anthropic 与 OpenAI 竞争","google-deepmind-reshuffle-gemini-speed","2026-08-25T07:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"bcedeb8e-e5eb-4bbc-98b8-ea12f869055f","Google 收编 DeepMind：25 年最大 AI 重组","google-deepmind-centralization-gemini-catchup","2026-08-14T08:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"4bd8e8bd-7066-4ab7-bd97-e24ea3921395","Gemini 因编程落后推迟两月:Brin 4 月督促背后,Google 把研发「收回到一个人」手里的组织账本","google-gemini-coding-behind-deepmind-reshuffle","2026-08-14T03:30:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"8b6c20ec-7222-48cf-af2c-ac97466a2b0a","Gemini 月活破 10 亿:Google 第一次把 AI 助手做成自家「最快十亿用户产品」","gemini-app-1b-monthly-users","2026-08-12T03:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"34edaffc-6b5c-4df1-9e2f-d864cada6063","Gemini 走进 K-12 课堂：Google 把「上下文」塞进每个作业","gemini-classroom-k12-contextualized-prompts","2026-08-07T02:00:00+00:00"]