[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-world-in-context-interactive-video-world-model":3,"topics-all":38,"news-related-8d11d7c2-a373-43e1-9014-11163f4ea4db":57},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"8d11d7c2-a373-43e1-9014-11163f4ea4db","5B追平22B:阿里把视角、动作、镜头装进同一个注意力空间","阿里Token Hub团队公开World-in-Context框架:把视角、动作、镜头三类控制统一成显式token,放进同一个自注意力空间。基于Wan2.2-TI2V-5B底座,论文自报WBench导航81.5分追平22B的EchoWM,视角切换成功率93.8%对5B基线的18.8%,代码与权重仍在审核未放出。","交互式视频世界模型有个绕不开的麻烦:用户要同时控制三样东西——看世界的视角、角色的动作、镜头的运动。主流做法是把视角模式藏进文本提示里,把动作和镜头混在一起理解,或者给每类控制单独接一条条件通路,三条道互相看不见,冲突只能靠数据硬磨。10 月 23 日,阿里巴巴 Token Hub 团队在 arXiv 公开 World-in-Context(下称 WiC),思路是别再各修各的路:把三类控制全部显式 token 化,放进同一个自注意力空间。论文地址:https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.33169。\n\n## 三个旋钮,一个面板\n\nWiC 的核心叫 In-Context Conditioning(ICC):视角、动作、相机控制不再走各自的专用条件通路,而是作为显式 token 与视觉潜变量共享同一个自注意力空间。团队从消融研究里得到两个发现:一是显式视角 token 能解耦相互冲突的动态,并从只有单视角的训练数据里涌现出视角切换能力;二是动作和镜头的角色是互补的——动作输入主导局部响应,镜头输入负责全局稳定,谁也替代不了谁。为了保住跨执行间隔的时间连续性,论文还提出 Interaction-Transport RoPE(IT-RoPE),把累积的控制转移参数化进时间注意力。\n\n## 5B 追平 22B\n\n数字是这套设计的卖点。以 Wan2.2-TI2V-5B 为底座,WiC 在交互世界模型基准 WBench 的导航任务上拿到 81.5 的平均分,论文报告与 22B 的 EchoWM(8 月底京东开源的那款全模态世界模型)的 81.6 持平——5B 干到 22B 的水平,中间隔着四倍多的参数量。更悬殊的是视角切换:93.8% 的成功率,而现有 5B 基线只有 18.8%,差出 75 个百分点。需要说明,93.8 这一档是做了视角切换微调后的成绩,项目页对此有明确标注;训练本身也走轻量路线,两个阶段只优化 LoRA、条件编码器和 IT-RoPE,不动整个底座。GitHub README 还给出计划中的推理默认参数:832×480 分辨率、81 帧、16 FPS。\n\n## 冷水与所以呢\n\n泼冷水的部分:仓库 TODO 里推理代码、训练代码、WiC 权重三项目前全部未放出,官方说法是仍在审核;81.5 对 81.6 这类数字也全部出自论文自报,第三方复现暂时无从谈起,想上手验证的开发者现在只能看项目页的演示视频。但方向值得记住:把控制接口显式 token 化,和 LLM 把工具调用、思考过程 token 化是同一个设计哲学——控制信号进了上下文,模型才能对\"此刻为什么该切镜头\"建立真正的归因。世界模型赛道近两个月已经挤进世界记忆、引擎奖励信号、评测综述好几种叙事,WiC 补上的是\"控制总线\"这块拼图。等权重放出、第三方跑分落地,5B 与 22B 之争才有定论;在那之前,它至少证明了一件事——统一接口带来的收益,可能比堆参数更划算。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.33169","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"120fa59a-ff6f-4537-9bf5-f818df636a0e","benchmark",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"a8002d98-9df1-4ab9-94d4-a7625af634c4","china-ai",{"id":19,"name":20,"slug":20,"description":14,"color":14},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":22,"name":23,"slug":23,"description":14,"color":14},"ebe5dcd1-46b1-4298-b8c2-8e0e2f456e56","video-generation",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"1ae0a221-b44c-4537-913a-343e99a8fb4e","en","5B World Model Matches 22B: Alibaba Unifies View, Action, Camera","Alibaba's World-in-Context unifies view, action, and camera as tokens in one attention space; 5B self-reports 81.5 on WBench, matching 22B EchoWM.","Interactive video world models share an awkward problem: users must control three things at once — the viewing perspective, the character's actions, and the camera. Mainstream practice hides view modes inside text captions, conflates action with camera motion, or wires a separate conditioning pathway for each control, so the three channels never see each other and conflicts get ground out by data alone. On October 23, Alibaba's Token Hub team published World-in-Context (WiC) on arXiv with the opposite bet: stop building separate roads, turn all three controls into explicit tokens, and put them in one shared self-attention space. Paper: https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.33169.\n\n## Three knobs, one panel\n\nThe core is In-Context Conditioning (ICC): view, action, and camera controls no longer travel dedicated pathways but join visual latents as explicit tokens in the same self-attention space. Two findings stand out from the team's analysis. First, explicit view tokens disentangle conflicting dynamics and unlock emergent view switching from single-view training data. Second, action and camera inputs are complementary — action drives local responsiveness while camera governs global stability, and neither substitutes for the other. To preserve temporal continuity across execution intervals, the paper introduces Interaction-Transport RoPE (IT-RoPE), which parameterizes temporal attention with accumulated control transitions.\n\n## 5B matches 22B\n\nThe numbers are the selling point. Built on the Wan2.2-TI2V-5B backbone, WiC scores an 81.5 navigation average on WBench, which the paper reports as on par with the 22B EchoWM (81.6) — the omni world model JD open-sourced in late August. A 5B model reaching 22B territory crosses more than four times the parameter count. The gap widens on perspective switching: 93.8% success versus 18.8% for existing 5B baselines, a 75-point spread. One caveat the project page states explicitly: the 93.8 figure comes after perspective-switching fine-tuning. Training stays lightweight — both stages optimize only LoRA, the condition encoder, and IT-RoPE, leaving the backbone untouched. The GitHub README also lists planned inference defaults of 832×480 resolution, 81 frames, and 16 FPS.\n\n## Cold water, and so what\n\nThe cold shower: the repository TODO shows inference code, training code, and WiC weights all unreleased, officially \"under review.\" Every headline number is self-reported, third-party replication is impossible for now, and developers who want to verify can only watch the demo videos on the project page. Still, the direction is worth remembering: making control interfaces explicit tokens mirrors what LLMs did with tool calls and reasoning traces — once control signals live in the context, the model can genuinely attribute why the camera should cut now. Over the past two months the world-model track has accumulated narratives about world memory, engine-derived rewards, and evaluation surveys; WiC supplies the \"control bus\" piece. Whether 5B truly stands next to 22B waits for released weights and independent benchmarks. Until then, it has already shown one thing: the payoff from unifying interfaces may beat stacking parameters.","world-in-context-interactive-video-world-model","2026-10-27T19:08:47Z","2026-10-27T19:08:56.076803Z","2026-10-27T19:08:56.076816Z",true,"agent",35,[39,48],{"slug":40,"tag_slug":40,"title_zh":41,"title_en":42,"intro_zh":43,"intro_en":44,"id":45,"is_active":35,"created_at":46,"modified_at":47},"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":49,"tag_slug":49,"title_zh":50,"title_en":51,"intro_zh":52,"intro_en":53,"id":54,"is_active":35,"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},"d34bdd08-5fa9-4ce0-954d-17406b896756","牛津102页综述:视频世界模型的四级证据阶梯","video-world-model-evidence-ladder-survey","2026-10-26T11:13:04+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"891e0c6f-0c26-4ca6-bc85-3fca60bdff72","代码写视频:15 个多模态大模型实测,跑通率比画质更决定成败","code2video-compass-mllm-code-video-benchmark","2026-10-25T09:15:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"dc534f1a-dbe2-4f79-9557-bbb43a82f24f","视频模型总在绕过 MLLM,三段训练把它拉回来","fapl-mllm-latents-cinematic-video-generation","2026-10-23T09:20:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"cf7aeaaa-7179-4de2-b177-9adf54668d61","StreamEval：37 模型实测，实时视频最缺听指令","streameval-interactive-streaming-video-benchmark","2026-10-22T11:08:09+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"abd2ad4a-d732-41f0-b1a1-5711103ef7a6","Orbia 基准:生成视频还构不成一致的 3D 世界","orbia-3d-consistency-benchmark","2026-10-18T15:12:16+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"d0d8e772-57f7-47a0-8fd8-dce09f3bd1e5","浙大ATAS登顶HF论文榜:给AI视频装上持久世界记忆","atas-world-state-agentic-video-generation","2026-10-15T15:15:45+00:00"]