[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fEhkquGPmbnfIsCGyatM_Vk5dHeT38muGg1kFQIGzKXo":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1530," 腾讯混元世界模型1.5发布：重新定义实时3D世界建模的未来 ","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">在数字化时代的浪潮中，技术的飞速发展不断为我们的生活带来新的可能性。近日，腾讯宣布推出混元世界模型1.5（Tencent HYWorldPlay），这不仅是一项技术更新，更是一场世界建模领域的革命。作为业界最系统、最全面的实时世界模型框架，混元世界模型1.5的发布标志着3D建模技术的新高度。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">一、混元世界模型1.5的核心创新\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">腾讯混元团队通过这一版本的发布，向我们展示了在实时交互和空间记忆能力上的重大突破。混元世界模型1.5采用了一系列创新的算法模块，包括重构记忆力、长上下文蒸馏和基于3D的自回归扩散模型强化学习，提升了模型在复杂环境下的表现。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">实时交互生成\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：混元世界模型1.5通过原创的Context Forcing蒸馏方案和流式推理优化，能够以24FPS的速度生成720P的高清视频。这一速度对于用户体验至关重要，让用户在生成的世界中可以实现流畅的探索。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">长范围的3D一致性\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：通过重构记忆机制，模型支持分钟级内容的几何一致性生成。这意味着用户在离开某个区域后，再次返回时，模型能够“记住”该区域的三维结构，呈现出前后一致的场景。这一特性在构建高质量的3D空间模拟器时尤为重要。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">多样化的交互体验\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：混元世界模型的应用场景广泛，能够适用于不同风格的游戏或现实场景。它不仅支持第一和第三人称视角，还能够实时触发事件和视频续写，为用户提供了丰富的互动体验。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">二、开放源代码的意义\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">此次发布的混元世界模型1.5首次开源了实时世界模型训练体系，涵盖了数据、训练和推理部署的全链路、全环节。混元团队在技术报告中详细介绍了模型的预训练、持续训练、自回归视频模型强化学习以及带记忆力的模型蒸馏等训练细节。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这种开放源代码的方式，不仅能够推动技术的进步，也为研究人员和开发者提供了一个理想的实验平台，促进更广泛的技术应用和创新。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">三、世界模型的未来发展\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">混元世界模型1.5的发布不仅是技术上的突破，更是对“AI生成内容”未来的展望。随着这一技术的不断迭代，未来的AI将不仅仅局限于生成文本、图像或视频，而是能够创造出实时互动、可探索的完整“世界”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这种全新的内容生成模式可以在多个领域大放异彩：\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">游戏开发\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：在AI游戏开发领域，混元世界模型可以作为智能关卡生成器，根据玩家的文本描述实时创建可探索的游戏世界。\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">影视制作\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：在影视制作和虚拟现实（VR）领域，创作者可以通过简单的文本指令，快速预览和迭代场景设计，大幅提升创作效率。\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">具身智能研究\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">：在具身智能（Embodied AI）研究领域，混元世界模型1.5提供了一个理想的训练和测试平台，研究人员可以更高效地探索智能体的感知、决策、规划及长期交互能力。\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">四、总结\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">腾讯混元世界模型1.5的发布为3D建模领域带来了新的希望和可能性。通过不断的技术创新和开源分享，腾讯不仅推动了自身的技术进步，也为整个行业的发展贡献了力量。未来，随着这一技术的普及和应用，我们可以期待一个更加丰富多彩的数字世界。无论是在游戏、影视还是智能技术领域，混元世界模型1.5都将成为未来发展的重要基石。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】搜狐新闻 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.sohu.com\u002Fa\u002F966153461_121956424\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002F9388449477d3431d95db7bc9d809a393\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F12\u002Fthumbs\u002F9388449477d3431d95db7bc9d809a393\u002FAI领域.jpg",0,1,54,"2025-12-22 08:50",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Ada2210ef-5dce-401d-9ba6-4bf1d2dc63de%3A0.wav?Expires=1766384556&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=Q19xwLSXxRaCVMUoMxNuUw6qtWQ%3D",6740962,"da2210ef-5dce-401d-9ba6-4bf1d2dc63de","2025-12-22 08:48"," Tencent HYWorldPlay 1.5 Launch: Redefining the Future of Real-Time 3D World Modeling","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">In the tide of the digital age, the rapid development of technology continues to bring new possibilities to our lives. Recently, Tencent announced the release of HYWorldPlay, which is not only a technical update but also a revolution in the field of world modeling. As the most systematic and comprehensive real-time world model framework in the industry, the release of HYWorldPlay 1.5 marks a new height in 3D modeling technology.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">I. Core Innovations of HYWorldPlay 1.5\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Through this version, the Tencent HYWorld team has shown us significant breakthroughs in real-time interaction and spatial memory capabilities. HYWorldPlay 1.5 employs a series of innovative algorithm modules, including reconstructive memory, long context distillation, and 3D-based autoregressive diffusion model reinforcement learning, improving the model's performance in complex environments.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Real-Time Interaction Generation\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: HYWorldPlay 1.5 can generate high-definition videos at 720P with a speed of 24FPS through its original Context Forcing distillation scheme and streaming inference optimization. This speed is crucial for user experience, allowing users to explore the generated world smoothly.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Long-Range 3D Consistency\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: Through the reconstructed memory mechanism, the model supports geometric consistency generation for content lasting minutes. This means that when users leave a certain area and return later, the model can \"remember\" the 3D structure of that area, presenting a consistent scene before and after. This feature is particularly important in building high-quality 3D space simulators.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Diverse Interactive Experiences\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: The application scenarios of HYWorldPlay are extensive and can be applied to different styles of games or real-world scenarios. It not only supports first-person and third-person perspectives but can also trigger events and video continuation in real time, providing users with rich interactive experiences.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">II. Significance of Open Source\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The release of HYWorldPlay 1.5 marks the first open source of the real-time world model training system, covering the entire process from data, training to inference deployment. In the technical report, the HYWorld team detailed the training details of the model, including pre-training, continuous training, autoregressive video model reinforcement learning, and model distillation with memory.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This open-source approach not only promotes technological advancement but also provides researchers and developers with an ideal experimental platform, promoting broader technological applications and innovations.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">III. Future Development of World Models\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The release of HYWorldPlay 1.5 is not only a technological breakthrough but also a vision for the future of \"AI-generated content.\" As this technology continues to iterate, future AI will no longer be limited to generating text, images, or videos but will be able to create real-time interactive, exploratory complete \"worlds.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This new mode of content generation can shine brightly in multiple fields:\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">Game Development\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: In the field of AI game development, HYWorldPlay can serve as an intelligent level generator, creating exploratory game worlds in real time based on players' textual descriptions.\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">Film Production\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: In the fields of film production and virtual reality (VR), creators can quickly preview and iterate scene designs through simple text instructions, greatly improving creative efficiency.\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cul>\u003Cli>\u003Cstrong style=\"font-size: 18px;\">Embodied Intelligence Research\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">: In the field of embodied intelligence (Embodied AI) research, HYWorldPlay 1.5 provides an ideal training and testing platform. Researchers can more efficiently explore the perception, decision-making, planning, and long-term interaction capabilities of agents.\u003C\u002Fspan>\u003C\u002Fli>\u003C\u002Ful>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">IV. Summary\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The release of Tencent HYWorldPlay 1.5 brings new hope and possibilities to the field of 3D modeling. Through continuous technological innovation and open-source sharing, Tencent not only promotes its own technological progress but also contributes to the development of the entire industry. In the future, as this technology becomes more widespread and applied, we can expect a more colorful digital world. Whether in gaming, film, or intelligent technology, HYWorldPlay 1.5 will become an important cornerstone for future development.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【News Source】 Sohu News \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.sohu.com\u002Fa\u002F966153461_121956424\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（This article is reprinted by this site to provide readers with more information. The content does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant parties. The views of the article are not the views of this site and are for reference only.）\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A41892c2b-7370-4705-9e59-87ccb4147099%3A0.wav?Expires=1774838439&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=iiJjqcBCSX20V3%2F7bWHDI9XNEcE%3D","41892c2b-7370-4705-9e59-87ccb4147099",8777046]