[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-qwen-drive-1-vlm-autonomous-driving":3,"news-related-ea425005-49e7-477b-9f64-54361254c2d2":38},{"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},"ea425005-49e7-477b-9f64-54361254c2d2","Qwen 开进驾驶场景:Qwen-Drive-1.0 保留 VLM 主干,外挂 BEV 感知与规划专家","Qwen 团队发布 Qwen-Drive-1.0 驾驶视觉语言基座模型:保留 Qwen3.5-4B 主干,外挂 BEV 感知头与规划专家,统一 3D 感知、视觉问答与运动规划,冲上 HF 当日论文热榜第二。","自动驾驶这两年被\"端到端\"推着往前跑,但一个老问题始终横在中间:通用视觉语言模型看不懂 3D 世界,而传统驾驶栈又没有语言与推理。8 月 31 日,Qwen 团队在 arXiv 上传了 Qwen-Drive-1.0,标题直译过来就是《迈向自动驾驶视觉语言基座模型的第一步》,16 位作者署名,两天内冲上 Hugging Face 论文热榜当日第二、335 个点赞——这是 Qwen 第一次把自家 VLM 正式开进驾驶场景。\n\n## 主干不动,外挂两个模块\n\nQwen-Drive-1.0 最值得注意的设计选择,是预训练 VLM 的架构完全保留、一行主干不改,只在外面挂两个模块。\n\n一个是 BEV 感知头,联合执行 3D 目标检测、语义占据预测和 BEV 地图分割三项任务。论文特意把它定位成\"探针\"——探测共享表征里到底能榨出多少 3D 信息,同时给 3D 场景结构提供一个显式、可检查的接口。\n\n另一个是规划专家(Planning Expert),以共享 VLM 表征为条件生成自车未来轨迹。论文一作 Xin Zhou 在 HF 评论区补充了两个摘要里没写的细节:基座是原生多模态的 Qwen3.5-4B,轨迹生成走 flow matching。\n\n## 分阶段训练:学会开车,别忘聊天\n\n领域适配最怕\"学了驾驶、丢了通用\"。Qwen-Drive-1.0 用分阶段训练配方,把驾驶监督数据与通用视觉语言数据混合,在拿到驾驶专长的同时,保住宽泛的视觉理解与指令跟随能力。\n\n评测覆盖开环、伪闭环、闭环三类设定,任务面包括 3D 感知、驾驶视觉问答与运动规划。论文的结论措辞很克制:3D 感知与驾驶场景理解表现强,通用视觉语言能力\"基本保留\",运动规划在三类设定下\"高度竞争\"。\n\n## 权重还没放,代码在路上\n\n按时间线提醒一句:arXiv 页面写的是代码\"将\"在 GitHub(QwenLM\u002FQwen-Drive-1.0)开放,HF 详情页显示关联模型数为 0——此刻只能读论文,下不到权重跑推理。一作称它是\"首个在预训练阶段统一 3D 感知与视觉问答的自动驾驶视觉语言基座模型\",注意\"首个\"是作者口径,留给独立评测检验。\n\n## 所以呢\n\nQwen-Drive-1.0 的真正信号不在跑分,而在范式:与其从零训练驾驶专用模型,不如拿通用 VLM 当底座、外挂领域专家——这与\"基座模型 + 领域适配\"的行业大方向完全同频。下一步只看一件事:驾驶场景会不会成为 VLM 基座厂商的下一个必争之地?\n\n参考资料:https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.00111","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.00111","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"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},"c187600e-804c-4697-b828-1e4330e0eb10","qwen",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"d705fd58-b93d-41d0-9d9b-69ea0cf88338","en","Qwen-Drive-1.0: Frozen VLM Backbone Meets Autonomous Driving","Qwen's driving VLM: Qwen3.5-4B backbone untouched, BEV head + Planning Expert unify 3D perception, VQA and motion planning. #2 on HF Daily Papers.","Autonomous driving has spent two years being pulled forward by end-to-end pipelines, but one old problem still sits in the middle: general-purpose vision-language models cannot read the 3D world, while classical driving stacks have no language or reasoning. On August 31, the Qwen team posted Qwen-Drive-1.0 to arXiv — \"An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving\" — with 16 authors. Within two days it reached #2 paper of the day on Hugging Face with 335 upvotes: Qwen's first official move to drive its VLM line into the driving domain.\n\n## Untouched backbone, two bolt-on modules\n\nThe most notable design choice: the pretrained VLM architecture is kept entirely intact. Two external modules are attached instead.\n\nOne is a bird's-eye-view (BEV) perception head that jointly performs 3D object detection, semantic occupancy prediction, and BEV map segmentation. The paper frames it as a probe — measuring how much 3D information the shared representations actually contain, and providing an explicit, inspectable interface to 3D scene structure.\n\nThe other is a Planning Expert that conditions on shared VLM representations to generate future ego trajectories. First author Xin Zhou added two details in the HF comment section that the abstract omits: the base is the natively multimodal Qwen3.5-4B, and trajectories are generated through flow matching.\n\n## Staged training: learn to drive, don't forget to chat\n\nDomain adaptation usually fears catastrophic specialization. Qwen-Drive-1.0 uses a staged training recipe that mixes driving supervision with general-purpose vision-language data, acquiring driving-specific competence while preserving broad visual understanding and instruction-following.\n\nEvaluation spans open-loop, pseudo-closed-loop, and closed-loop settings across 3D perception, driving VQA, and motion planning. The paper's wording is careful: strong 3D perception and driving scene understanding, general capability \"largely preserved,\" and \"highly competitive\" motion planning.\n\n## No weights yet, code \"will be\" on GitHub\n\nTimeline note: the arXiv page says code will be available at github.com\u002FQwenLM\u002FQwen-Drive-1.0, and the HF page shows zero linked models — for now you can read the paper but cannot download weights. The first author calls it \"the first vision-language foundation model for autonomous driving that unifies 3D perception and visual question answering at the pretraining stage\" — an author claim, to be tested by independent evaluation.\n\n## So what\n\nThe real signal is the paradigm, not the scores: instead of training driving models from scratch, take a general VLM as the base and bolt on domain experts. That is exactly the \"foundation model + domain adaptation\" direction the industry is converging on. The only question left: will driving become the next battleground for VLM foundation vendors?\n\nReference: https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.00111","qwen-drive-1-vlm-autonomous-driving","2026-09-02T19:35:00Z","2026-09-02T19:08:12.945865Z","2026-09-02T19:08:12.945880Z",true,"agent",116,{"items":39},[40,45,50,55,60,65],{"id":41,"title":42,"news_slug":43,"published_at":44},"40095b51-97b0-4fd4-9b1d-f636c970572e","阿里 Qwen 团队发布 Qwen3.8-Max:2.4 万亿参数 MoE 模型首度开放权重","qwen3-8-max-2-4t-moe-open-weights","2026-08-07T02:00:00+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"d941056b-c2e7-42e5-965a-a982c20b1169","Qwen3.8-Flash-Next 架构细节:Gated Residual 多分支残差 + QSA micro-block 稀疏注意力","qwen3-8-flash-next-cost-efficiency-architecture","2026-09-02T02:00:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"637f84e0-e6dc-490a-bba1-879f6527bdd5","Qwen3.8-Max 2.4T 开源:Gated DeltaNet 把长上下文成本砍到 1\u002F8","qwen3-8-max-2-4t-open-weights-gated-deltanet","2026-08-30T03:00:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"cb64371f-62b6-473d-8150-b576001d3f56","Qwen3.8-27B 开源权重上线:单卡跑得动的 Qwen3.8,还塞了个视觉编码器","qwen3-8-27b-open-weights-release","2026-08-14T19:30:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"804b44fb-66c6-4355-83a4-b3a03a776d2a","Inkling-Small 开放权重：12B 激活参数换来更高 Agent 效率，也暴露事实性短板","inkling-small-multimodal-moe-efficiency","2026-08-05T16:32:13+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"d7b6d14d-7257-4794-b92f-31956bbc7eae","原生多模态 vs 后训练加压:国产头部基模两条路线的工程账","native-multimodal-vs-posttraining-2026","2026-08-05T00:00:00+00:00"]