[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-orbia-3d-consistency-benchmark":3,"topics-all":38,"news-related-abd2ad4a-d732-41f0-b1a1-5711103ef7a6":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},"abd2ad4a-d732-41f0-b1a1-5711103ef7a6","Orbia 基准:生成视频还构不成一致的 3D 世界","Shuicheng Yan 等作者发布 Orbia 基准:用 720 个探索-回访案例审计 27 个交互视频世界模型,世界模型控制分均值 57.29、领先文生视频组 24.80 分,但生成内容持久性与 3D 自一致性普遍薄弱,Genie3 还会把侧移指令执行成转身。基准、代码、数据全部开源。","视频生成模型的下一步叙事叫「世界模型」:不止生成画面,还让用户走进去探索、交互,当一个可以进入的小宇宙。但一个真世界有条底线——你转身再回头看,刚才那个房间还得待在原地。Shuicheng Yan、Mike Zheng Shou 等作者在 arXiv 发布的 Orbia 基准专门审计这件事:模型生成的视频,究竟构不构成一个一致的 3D 世界。\n\n## 720 个「探索-回访」案例\n\nOrbia 的思路朴素:让模型生成一段可交互探索的视频,再回到起点看第二眼。基准共 720 个实例,按六种「探索-回访」协议执行;短轨迹约 22 秒、含一次回访,分钟级扩展则包含三次回访,专门压测长时程记忆。数据管道混合公开数据集、自采场景与 UE 仿真:短案例中 120 个室内、360 个室外,300 个来自真实观测、180 个来自仿真,输出统一 720p、16 帧。评估沿三个维度:输入保持(生成是否忠实保留输入图像)、生成内容持久性(模型自己「创造」的区域,离开视野再回来还在不在)、3D 自一致性(跨视角深度、特征对应与留出高斯渲染,是否指向同一个 3D 场景)。\n\n## 27 个模型,一组残酷的分数\n\n团队用原生接口适配器评测 27 个模型,覆盖文本、动作、相机三类条件接口。控制能力上,世界模型组均值 57.29,文图生视频组仅 32.48,差距 24.80 分;文图生视频组里控制最好的 Seedance2.5 拿 43.23,仍远低于 Lyra2 的 81.64。分项冠军各有归属:Genie3 领跑时序控制分,Lyra2 平均控制水平最高,EVOKE 居世界模型组时序质量首位,SolarWM-H3 拿下几何稳定性与总榜,HY-WorldPlay 平均几何水平最高;文图生视频组内,Seedance2.5 靠控制领先,Cosmos3-Super 在质量与几何占优。\n\n## 连 Genie3 都会「侧移变转身」\n\n逐案检查暴露的失败模式比总分更有信息量:Genie3 在多个受检案例中把侧移请求执行成原地转向,平移与转向的混淆也体现在其偏低的 Look Accuracy(转向执行准确率)上。这说明「看起来像世界」与「能当地图用」之间隔着结构可靠性的沟:视觉说服力不等于跨视角一致。对把世界模型当交互训练场的路线而言,回访一致性是硬前提——一个会悄悄改写自己历史的模拟器,训出来的 agent 也会学会容忍矛盾。\n\n基准、代码与数据集全部开源:项目页 orbia-bench.github.io,代码在 GitHub(worldbench\u002FOrbia),数据集上架 Hugging Face。这份审计给行业的不只是「谁最强」的榜单,而是一把统一的尺子:下次再有人宣传「可交互世界」,先问一句——回头再看,它还在吗?\n\n参考:arxiv.org\u002Fabs\u002F2610.20693","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.20693","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},"499f4b56-819d-49a3-9609-33e775143b86","multimodal",{"id":19,"name":20,"slug":20,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",{"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},"3bf746dd-d867-475f-ad12-60b265c7dfab","en","Orbia benchmark: generated video is not yet a consistent 3D world","Orbia audits 27 video world models on 720 revisit cases: control 57.29 vs 32.48 for text-to-video, yet 3D consistency stays weak. Code and data open.","The next narrative in video generation is the \"world model\": not just rendering frames, but letting users step inside, explore, and interact. A real world has a bottom line, though—turn around and look back, and the room you just left should still be there. Orbia, a benchmark released on arXiv by a team including Shuicheng Yan and Mike Zheng Shou, audits exactly that: do generated videos form a consistent 3D world?\n\n## 720 explore-and-revisit cases\n\nOrbia's idea is simple: have a model generate an interactive exploration, then return to the start and look again. The benchmark ships 720 instances following six exploration-and-revisit protocols; short trajectories run about 22 seconds with one revisit, while minute-long extensions include three revisits to stress long-horizon memory. The data pipeline mixes public datasets, self-captured scenes, and UE simulation: among short cases, 120 indoor and 360 outdoor, 300 from real-world observations and 180 from simulation, standardized at 720p and 16 FPS. Evaluation spans three dimensions: input preservation, generated-content persistence, and 3D self-consistency (cross-view depth, feature correspondence, and held-out Gaussian rendering).\n\n## 27 models, one harsh scoreboard\n\nVia native-interface adapters, the team evaluated 27 models spanning text-, action-, and camera-conditioned interfaces. On control, world models average 57.29 versus 32.48 for text\u002Fimage-to-video systems—a 24.80-point gap; the best TI2V model, Seedance2.5, reaches 43.23, still far below Lyra2's 81.64. Category champions: Genie3 leads the temporal control score, Lyra2 holds the highest mean control level, EVOKE tops world-model temporal quality, SolarWM-H3 takes geometric stability and the overall tier, and HY-WorldPlay has the highest mean geometry level; within the TI2V group, Seedance2.5 leads overall through stronger control while Cosmos3-Super leads quality and geometry.\n\n## Even Genie3 turns side-steps into rotations\n\nPer-case inspection is more informative than totals: in several inspected Genie3 cases, lateral requests produce turns instead of translation—confusion between moving and turning that shows up in its lower Look Accuracy. Looking like a world and working like a map are separated by a structural-reliability gap: visual persuasiveness is not cross-view consistency. For anyone using world models as interactive training grounds, revisit consistency is a hard prerequisite—a simulator that quietly rewrites its own history teaches agents to tolerate contradiction.\n\nThe benchmark, code, and dataset are all open: project page at orbia-bench.github.io, code on GitHub (worldbench\u002FOrbia), dataset on Hugging Face. What this audit gives the field is less a leaderboard than a shared ruler. Next time someone pitches an \"interactive world,\" ask one question: look back—does it still hold?\n\nReference: arxiv.org\u002Fabs\u002F2610.20693","orbia-3d-consistency-benchmark","2026-10-18T15:12:16Z","2026-10-18T15:12:46.340884Z","2026-10-18T15:12:46.340895Z",true,"agent",144,[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},"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":65,"title":66,"news_slug":67,"published_at":68},"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",{"id":70,"title":71,"news_slug":72,"published_at":73},"ba0ed7bf-3de3-4f92-98fe-a50d6ac274d0","WROP 开源:用 150 个物体恒存任务给世界模型补认知课","wrop-object-permanence-world-models","2026-09-25T17:08:02+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"b95b93e8-294a-4c5b-b53d-ce6ea07c1519","SemComp-Bench 登顶 Hugging Face 日榜:视频生成开始考「任务做没做成」","semcomp-bench-video-task-completion","2026-08-20T13:30:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"9fa15063-58b1-4b87-a0cb-0e19ffc4dc6c","4B 参数横扫四大具身基准：开悟世界模型让小模型重新定义 SOTA","kaiwu-4b-world-model-sensetime-72x","2026-06-12T04:00:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"bef7a773-36fb-406b-949d-8d728b97b242","Matrix-Game 4.0 开源:单卡 60FPS 的实时可玩游戏世界","matrix-game-4-0-open-source-60fps","2026-10-21T13:09:32+00:00"]