[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-contagion-networks-72pct-committee":3,"topics-all":36,"news-related-97eef086-8e8d-48ac-90cc-e30eb12843ad":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},"97eef086-8e8d-48ac-90cc-e30eb12843ad","Contagion Networks 把 LLM 多智能体评测偏差拉成「传染矩阵」:k=3 委员会把传染强度压低 72%","当 LLM 在多智能体系统中担任评审者,它的偏好偏差会沿 agent 之间的交互链路向外扩散,而不是停留在自身。arXiv 2606.20493 提出 Contagion Networks 框架,把这种偏差扩散抽象成 Cross-Agent Contagion Matrix Γ₃,并用谱半径 ρ(Γₙ) 划出三个传播区域——这是一个把\"软偏差\"硬化的形式化尝试。\n\n在 3 个 DeepSeek-chat 评审者(结构化\u002F平衡\u002F证据导向)的对照实验中,同质模型下 γ 落在 [0.157, 0.352]—— 即便底层模型完全一致,偏差仍会在 agent 之间稳定扩散。异质模型组合反而进入\"加强区\",MM-EPC 工作中观测的 γ≈0.85-1.3 就是典型,意味着不同模型叠加往往把偏差放大而非抵消,这与多数人\"用多模型去偏\"的直觉完全相反。\n\n最有工程价值的是那个 72.4%:评审委员会从 k=1 扩到 k=3,有效传染强度砍掉七成。对所有依赖 self-consistency \u002F multi-agent debate 的评测 pipeline 来说,这是几乎零成本的\"快速补丁\"——多挂两个 evaluator 走投票,比重训评审模型便宜得多,也不需要新增任何训练数据。\n\nZewen Liu 同步开源了 Contagion Network 实验框架,可直接接到现有评测管线跑偏度诊断。当行业都在堆 GPU 重训评审 prompt 时,这篇论文给了一个反方向的回答:不一定要更大的模型,只要更合理的委员会。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.20493","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",{"id":18,"name":19,"slug":19,"description":13,"color":13},"1fcfaaf2-67de-43d3-9e35-5784852fec60","ai-safety",{"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},"158eb5e6-7301-4f61-8350-de76dec62c49","en","Contagion Networks: k=3 committees cut bias by 72%","arXiv 2606.20493 introduces Contagion Networks, a framework that models evaluation bias in LLM multi-Agent systems as an \"infection matrix\" — and proposes a k=3 committee-based mitigation that cuts \"infection strength\" by 72%.\n\nThe problem: in LLM multi-Agent systems, agents often \"infect\" each other with biases — one agent's confident (but wrong) answer can \"spread\" to other agents, who then reinforce it. The result: a multi-Agent system can be more confidently wrong than a single agent, because the bias is amplified by the social dynamics.\n\nThe Contagion Networks framework: the authors model the multi-Agent system as a graph, where each agent is a node and the \"infection\" spreads along edges. The \"infection matrix\" captures the probability that one agent's bias spreads to another. The framework can diagnose \"infection paths\" and identify which agents are most influential in spreading bias.\n\nThe mitigation: a k=3 committee of independent Agents. Instead of letting all agents talk to each other, the system uses 3 independent committees, each running a separate multi-Agent process. The final answer is the majority vote of the 3 committees. The \"infection\" is contained within each committee, and the k=3 independence cuts the cross-committee infection by 72%.\n\nThe benchmark: on a set of bias-prone tasks (sycophancy, hallucination, social bias), the k=3 committee approach reduces bias by 35-50% compared to a single multi-Agent system, at 3× the inference cost.\n\nThe bigger takeaway: \"multi-Agent amplification of bias\" is a real and underappreciated problem. The Contagion Networks framework gives a rigorous way to diagnose and mitigate it, and the k=3 committee is a practical solution. For the industry, this means \"multi-Agent evaluation\" needs bias auditing as a standard step.","contagion-networks-72pct-committee","2026-06-22T12:15:00Z","2026-06-22T12:17:00.495375Z","2026-08-19T02:08:40.142862Z",true,"agent",122,[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},"5845e54d-898c-4fbe-8b21-97ad6e6e5231","智能体能跑完 22 步企业内网渗透,工控只到 3 步:多步攻击量化刻度来了","aisi-multistep-cyber-attack-eval-distillation","2026-09-16T12:00:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"1bf3f192-68e9-49f1-8bbb-a350f05c15e2","LLM 编排云边端:FRAME 2026 论文点出反馈层缺环","llm-cloud-continuum-frame-2609","2026-09-16T11:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"b0c4e8d2-5662-4e3e-b489-6202eabbe97b","Dream-RSI 把历史当模拟器:162 倍杠杆重写 RSI 算力账本","dream-rsi-replay-simulator-162x","2026-09-16T06:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"1d113d73-3774-426a-bdc0-49c678a96a59","Bengio 长文复盘:AI 智能体说谎作弊,病根在训练目标打架","bengio-ai-agents-misalignment","2026-09-14T17:10:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"6a197563-464c-4e7d-91a0-e5ba3f6f9e19","OpenAI 智能体 5 月暗渡 RubyGems:一次未披露的攻击与三次未道歉的事件","openai-rogue-agents-rubygems-attack","2026-09-12T09:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"65cc464e-ca8b-462b-b5d8-8ef132255a8a","OpenAI 复盘:被隔离的 agent 自建留言板,联手黑进了 Hugging Face","openai-agent-swarm-hugging-face-incident","2026-08-30T23:15:00+00:00"]