[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-wevisone-vl-unified-visual-perception":3,"topics-all":38,"news-related-a744ef64-ee58-41ea-892f-1c2e5ba50bb1":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},"a744ef64-ee58-41ea-892f-1c2e5ba50bb1","微信视觉团队开源WeVisOne-VL:检测到抠图全包","腾讯微信视觉团队开源 WeVisOne-VL,放出 2B\u002F4B 两档权重。模型以结构化 next-token 预测统一检测、计数、OCR、分割与抠图,像素解码器仅 9.6M 参数、无需 SAM。官方表格称 4B 版在 30 余项基准领先最强开源基线,自建基准上拒答率 96.4% 对 73.0,数字为官方自报。","视觉语言模型在\"看图说话\"上进步飞快,细粒度感知却始终露馅:数清画面里有几只鸟、给出某个目标的精确轮廓、或者把前景人物抠成带透明通道的图,多数模型要么含糊其辞,要么对不存在的目标硬编结果。10 月 22 日,腾讯微信视觉团队在 arXiv 公开技术报告并开源 WeVisOne-VL,思路是把对象级和像素级感知装进同一个模型,用一套接口覆盖\"从框到像素\"的全部定位需求。\n\n## 一个模型,七种活\n\nWeVisOne-VL 的全称是 WeChat Vision All-in-One。按官方 README,它支持物体检测、指代定位、计数、指认、OCR、指代分割和 alpha 抠图七类任务,输出粒度从点、框、二值 mask 一路覆盖到连续 alpha 通道。这次同步放出 2B 与 4B 两档权重、推理包、Cookbooks 和评测管线,还开源了数据引擎里把二值 mask 转成高质量 alpha 的标注器 Mask2Alpha。GitHub 仓库 10 月 22 日创建,截至发稿约 47 个 star。\n\n## 9.6M 参数的像素解码器\n\n架构的看点不是大,而是轻:模型通过结构化 next-token 预测统一各类任务,再由一个仅 9.6M 参数的像素解码器经特殊 token 接入,直接输出实例级像素结果——不需要外挂 SAM 这类分割专用大件。训练侧采用三阶段课程 SFT 加混合任务 RL,官方称在 30 余项基准上同时保持对象级与像素级的领先表现。\n\n## 官方口径的跑分\n\nREADME 的对比表里,WeVisOne-VL-4B 的对手是每个基准上最强的开源基线,包括至 9B 的 Qwen3-VL\u002FQwen3.5、Rex-Omni、LocateAnything 与各任务专用模型。数字上:LVIS 检测 F1 55.3 对 50.7;密集检测 Dense200 71.3 对 58.7;RefCOCOg 指代定位 81.8 对 77.6;计数 PixMo-Count 89.6 对 82.8;OCR TotalText 52.0 对 43.3;指代分割 ReasonSeg 71.2 对 65.0;框抠图 P3M-500-NP 的 MAD 从 15.4 降到 7.3;文本抠图 MAD 从 22.4 降到 9.8。最扎眼的是拒答:WeRef-Bench 上对不存在目标的拒绝率 96.4%,基线只有 73.0,对应 README 的\"可靠拒绝、不幻觉\"设计。\n\n## 该泼的冷水\n\n上述数字全部为官方自报,WeRef-Bench、WeRefMat 带着\"We\"前缀、属团队自建基准,第三方复现有待社区验证。仓库 TODO 里 Gradio 网页 demo 和微调代码尚未放出,许可证标注为自定义(Other),商用前需要细读条款。热度倒是实打实:发布次日模型冲上 Hugging Face Daily Papers 当日榜第一,220 个 upvote。\n\n## 所以呢\n\n比起\"更大的模型、更长的上下文\",工业界真正缺的是\"该说没有时敢说没有\"的感知可靠性——96.4% 的拒答率比任何单项跑分提升都更能决定这类模型敢不敢上产线。2B\u002F4B 主干配 9.6M 解码器这条轻量路线,也给\"感知专用小模型 + 通用大模型分工\"的架构提供了新样本。下一个问题是:当感知被做成廉价的专用件,通用 VLM 还需要自己长出像素级的手吗?\n\n论文:arxiv.org\u002Fabs\u002F2610.31329;代码:github.com\u002FTencent\u002FWeVisOne-VL","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.31329","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},"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},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"14d9199c-7297-4fcd-8dfa-e16ec43660a4","en","WeChat Open-Sources WeVisOne-VL: From Detection to Matting","WeChat open-sources WeVisOne-VL: one VLM for detection, OCR, segmentation and matting via a 9.6M decoder; 4B tops open baselines; rejection 96.4%.","Vision-language models have gotten fluent at describing images, but fine-grained perception still trips them up: counting every bird in a frame, tracing an exact contour, or cutting a foreground subject into a transparent alpha matte. Most models either hedge or confidently invent results for targets that are not even there. On October 22, the WeChat Vision Team at Tencent published a technical report on arXiv and open-sourced WeVisOne-VL, an attempt to pack object-level and pixel-level perception into one model behind a single interface.\n\n## One model, seven jobs\n\nWeVisOne-VL stands for WeChat Vision All-in-One. Per the official README, it covers object detection, referring localization, counting, pointing, OCR, referring segmentation, and alpha matting, with output granularity spanning points, boxes, binary masks, and continuous alpha channels. The release ships 2B and 4B weight checkpoints, an inference package, cookbooks, and a full evaluation pipeline; the team also open-sourced Mask2Alpha, the annotator its data engine used to turn binary masks into high-quality alpha mattes. The GitHub repo was created on October 22 and sat at roughly 47 stars at writing time.\n\n## A 9.6M-parameter pixel decoder\n\nThe architectural story is lightness rather than scale: the model unifies tasks through structured next-token prediction, then a 9.6M-parameter pixel decoder plugs in via a special token to emit instance-level pixel outputs — no external SAM required. Training uses a three-stage curriculum SFT followed by mixed-task RL, and the team reports leading performance across more than 30 benchmarks at both object and pixel level.\n\n## The vendor-reported numbers\n\nIn the README's comparison table, WeVisOne-VL-4B faces the strongest open baseline per benchmark, including Qwen3-VL \u002F Qwen3.5 models up to 9B, Rex-Omni, LocateAnything, and task-specific segmentation and matting models. The numbers: LVIS detection F1 55.3 vs 50.7; Dense200 dense detection 71.3 vs 58.7; RefCOCOg referring 81.8 vs 77.6; PixMo-Count accuracy 89.6 vs 82.8; TotalText OCR 52.0 vs 43.3; ReasonSeg referring segmentation 71.2 vs 65.0; P3M-500-NP box matting MAD 7.3 vs 15.4; WeRefMat text matting MAD 9.8 vs 22.4. The standout is rejection: on WeRef-Bench, the model refuses absent targets 96.4% of the time versus 73.0 for the baseline — the direct payoff of the \"reliable rejection, no hallucination\" design.\n\n## The cold water\n\nEvery number above is vendor-reported, and WeRef-Bench and WeRefMat carry the team's own \"We\" prefix — self-built benchmarks awaiting independent replication. The repo's TODO still lists the Gradio web demo and fine-tuning code as unreleased, and the license reads as custom (\"Other\"), so commercial use warrants a careful read. The traction is real, though: the paper topped the Hugging Face Daily Papers board the day after release with 220 upvotes.\n\n## So what\n\nMore than bigger models or longer contexts, industrial deployment needs perception that dares to say \"nothing here\" — a 96.4% rejection rate matters more than any single benchmark gain. The 2B\u002F4B-backbone-plus-9.6M-decoder recipe is another data point for splitting perception into cheap, dedicated components. The next question: when perception becomes a cheap special-purpose part, do general VLMs still need to grow pixel-level hands of their own?\n\nPaper: arxiv.org\u002Fabs\u002F2610.31329 — Code: github.com\u002FTencent\u002FWeVisOne-VL","wevisone-vl-unified-visual-perception","2026-10-23T17:07:19Z","2026-10-23T17:07:41.593035Z","2026-10-23T17:07:41.593044Z",true,"agent",44,[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},"4c7f5330-3aff-458a-9ef5-f04cc5585703","微信视觉团队开源 WeMM 嵌入模型:2B 反超 8B 前基线,9B 达 MMEB-v2 80.6","wemm-embedding-wechat-multimodal","2026-08-26T21:07:30+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"652b1484-31eb-4a83-a932-21fcf97b3a50","Boson AI Higgs Audio v3 TTS：4B 参数原生可控百语种语音生成","higgs-audio-v3-boson-4b-100-language","2026-06-04T18:00:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"35e9f682-7952-4f6a-a8e9-7f8bdd47e791","IBM 语音三件套更新:2B 旗舰 ASR 精度开源第二","granite-speech-5-collection-update","2026-10-22T21:08:48+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"d83c4910-6dce-4171-8a44-9f68c9f841af","Qwen-Audio-3.1-ASR开源:转写自带说话人","qwen-audio-3-1-asr-speaker-diarization","2026-10-22T17:13:29+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"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":85,"title":86,"news_slug":87,"published_at":88},"bbbda7ff-b1c7-4142-b37f-2b3fd0547e02","Fibo 2.0 开源:生成编辑合一,登顶开源编辑榜","bria-fibo-2-0-unified-editing-generation","2026-10-21T21:08:11+00:00"]