Skip to content

Commit b3dbd54

Browse files
committed
deploy: e36dc59
1 parent ba27c81 commit b3dbd54

3 files changed

Lines changed: 8 additions & 8 deletions

File tree

zh/blog/index.html

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,10 +1,10 @@
11
<!doctype html><html lang=zh dir=auto><head><meta charset=utf-8><meta http-equiv=x-ua-compatible content="IE=edge"><meta name=viewport content="width=device-width,initial-scale=1,shrink-to-fit=no"><meta name=robots content="index, follow"><title>Blog | INCLUSION AI</title><meta name=keywords content><meta name=description content="Blog - INCLUSION AI"><meta name=author content="inclusionAI, Ant Group"><link rel=canonical href=https://inclusionai.github.io/zh/blog/><link crossorigin=anonymous href=/assets/css/stylesheet.fa92d8da4b04eca0dd3a888f432d2e4fabc3dec8bebb65e348f9ce26d866c08e.css integrity="sha256-+pLY2ksE7KDdOoiPQy0uT6vD3si+u2XjSPnOJthmwI4=" rel="preload stylesheet" as=style><link rel=icon href=https://inclusionai.github.io/favicon.png><link rel=apple-touch-icon href=https://inclusionai.github.io/favicon.png><link rel=manifest href=https://inclusionai.github.io/site.webmanifest><meta name=theme-color content="#615CED"><link rel=alternate type=application/rss+xml href=https://inclusionai.github.io/zh/blog/index.xml><link rel=alternate hreflang=en href=https://inclusionai.github.io/blog/><link rel=alternate hreflang=zh href=https://inclusionai.github.io/zh/blog/><noscript><style>#theme-toggle,.top-link{display:none}</style></noscript><script defer crossorigin=anonymous src=/js/custom.e9080c0a180dc80cf80d5fef7c857effb14f65c998e22134feb9896034b1b81a.js integrity="sha256-6QgMChgNyAz4DV/vfIV+/7FPZcmY4iE0/rmJYDSxuBo="></script>
22
<script async src="https://www.googletagmanager.com/gtag/js?id=G-NMEMBZ8R90"></script>
3-
<script>var doNotTrack=!1;if(!doNotTrack){window.dataLayer=window.dataLayer||[];function gtag(){dataLayer.push(arguments)}gtag("js",new Date),gtag("config","G-NMEMBZ8R90",{anonymize_ip:!1})}</script><meta property="og:title" content="Blog"><meta property="og:description" content="inclusionAI"><meta property="og:type" content="website"><meta property="og:url" content="https://inclusionai.github.io/zh/blog/"><meta property="og:image" content="https://inclusionai.github.io/%3Clink%20or%20path%20of%20image%20for%20opengraph,%20twitter-cards%3E"><meta property="og:site_name" content="inclusionAI"><meta name=twitter:card content="summary_large_image"><meta name=twitter:image content="https://inclusionai.github.io/%3Clink%20or%20path%20of%20image%20for%20opengraph,%20twitter-cards%3E"><meta name=twitter:title content="Blog"><meta name=twitter:description content="inclusionAI"><script type=application/ld+json>{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Blog","item":"https://inclusionai.github.io/zh/blog/"}]}</script></head><body class=list id=top><script>const hasHeaderBg=!0</script><header class=header><div class="nav-container nav-background"><nav class=nav><div class=logo><a href=/ accesskey=h title="inclusionAI (Alt + H)"><img src=https://inclusionai.github.io/img/logo_head.png alt aria-label=logo height=50></a></div><ul id=menu><li><a href=/blog/ title=Blog><span>Blog</span></a></li><li><a href=/publication title=Publication><span>Publication</span></a></li><li><a href=/about title=About><span>About</span></a></li><li><a href=https://github.com/inclusionAI title="Try Ling & Ming"><span>Try Ling & Ming</span>&nbsp;<svg fill="none" shape-rendering="geometricPrecision" stroke="currentcolor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.5" viewBox="0 0 24 24" height="12" width="12"><path d="M18 13v6a2 2 0 01-2 2H5a2 2 0 01-2-2V8a2 2 0 012-2h6"/><path d="M15 3h6v6"/><path d="M10 14 21 3"/></svg></a></li></ul></nav></div></header><div class=hero-container><div class=hero-background style="background:linear-gradient(-225deg,#2CD8D5 0%,#6B8DD6 48%,#8E37D7 100%)"></div><div class="hero text-light"><h1 class=post-title>Blog<sup class=title-translation>&nbsp;&nbsp;[<ul class=i18n_list><li><a href=https://inclusionai.github.io/blog/>English</a></li></ul>]</sup></h1></div></div><main class=main><article class=post-entry><header class=entry-header><h2>Ming-Lite-Omni V1.5:全能模型再升级,性能与体验双重突破!</h2></header><div class=entry-content><p>GITHUB 🤗 Hugging Face| 🤖 ModelScope
3+
<script>var doNotTrack=!1;if(!doNotTrack){window.dataLayer=window.dataLayer||[];function gtag(){dataLayer.push(arguments)}gtag("js",new Date),gtag("config","G-NMEMBZ8R90",{anonymize_ip:!1})}</script><meta property="og:title" content="Blog"><meta property="og:description" content="inclusionAI"><meta property="og:type" content="website"><meta property="og:url" content="https://inclusionai.github.io/zh/blog/"><meta property="og:image" content="https://inclusionai.github.io/%3Clink%20or%20path%20of%20image%20for%20opengraph,%20twitter-cards%3E"><meta property="og:site_name" content="inclusionAI"><meta name=twitter:card content="summary_large_image"><meta name=twitter:image content="https://inclusionai.github.io/%3Clink%20or%20path%20of%20image%20for%20opengraph,%20twitter-cards%3E"><meta name=twitter:title content="Blog"><meta name=twitter:description content="inclusionAI"><script type=application/ld+json>{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Blog","item":"https://inclusionai.github.io/zh/blog/"}]}</script></head><body class=list id=top><script>const hasHeaderBg=!0</script><header class=header><div class="nav-container nav-background"><nav class=nav><div class=logo><a href=/ accesskey=h title="inclusionAI (Alt + H)"><img src=https://inclusionai.github.io/img/logo_head.png alt aria-label=logo height=50></a></div><ul id=menu><li><a href=/blog/ title=Blog><span>Blog</span></a></li><li><a href=/publication title=Publication><span>Publication</span></a></li><li><a href=/about title=About><span>About</span></a></li><li><a href=https://github.com/inclusionAI title="Try Ling & Ming"><span>Try Ling & Ming</span>&nbsp;<svg fill="none" shape-rendering="geometricPrecision" stroke="currentcolor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2.5" viewBox="0 0 24 24" height="12" width="12"><path d="M18 13v6a2 2 0 01-2 2H5a2 2 0 01-2-2V8a2 2 0 012-2h6"/><path d="M15 3h6v6"/><path d="M10 14 21 3"/></svg></a></li></ul></nav></div></header><div class=hero-container><div class=hero-background style="background:linear-gradient(-225deg,#2CD8D5 0%,#6B8DD6 48%,#8E37D7 100%)"></div><div class="hero text-light"><h1 class=post-title>Blog<sup class=title-translation>&nbsp;&nbsp;[<ul class=i18n_list><li><a href=https://inclusionai.github.io/blog/>English</a></li></ul>]</sup></h1></div></div><main class=main><article class=post-entry><header class=entry-header><h2>Ming-lite-omni v1.5:全能模型再升级,效果与体验双优化</h2></header><div class=entry-content><p>GITHUB 🤗 Hugging Face| 🤖 ModelScope
44
概述 本次发布的 Ming-lite-omni V1.5 是对 Ming-lite-omni(Github) 全模态能力的一次全面升级, 在包括图文理解、文档理解、视频理解、语音理解和合成、图像生成和编辑等任务上均有明显提升。Ming-lite-omni V1.5 基于Ling-lite-1.5 构建,总参数20.3B, MoE部分激活参数为3B。与各领域同等规模的业界领先模型相比,在各模态基准测试中展现出极具竞争力的结果:
55
性能对比图
66
Ming-lite-omni v1.5能力介绍:三大维度全面优化,效果与体验双提升! 可控图像生成:像素级掌控,创意无限 Ming-lite-omni v1.5 重点优化了图像编辑的 场景一致性(Scene Consistency)、ID 一致性(Character / Style Consistency),在人物图像编辑时,在场景和人物ID 保持上展现出明显的优势,同时拓展了对生成式分割、深度预测、目标检测 以及 边缘轮廓生成 等感知任务的支持。
7-
生成式图像分割 Next User Given the following instructions: little girl, pink, your monitors colors off friend p pink shirt girl; please perform referring segmentation on this image. Ming-Lite-Omni V1.5 User Please segment different classes in this image....</p></div><footer class=entry-footer><span title='2025-07-21 00:00:03 +0800 +0800'>2025年7月21日</span>&nbsp;·&nbsp;3 分钟&nbsp;·&nbsp;603 字&nbsp;·&nbsp;inclusionAI, Ant Group</footer><a class=entry-link aria-label="post link to Ming-Lite-Omni V1.5:全能模型再升级,性能与体验双重突破!" href=https://inclusionai.github.io/zh/blog/ming-lite-omni-1_5/></a></article><article class=post-entry><header class=entry-header><h2>M2-Reasoning: 赋予多模态大语言模型统一的通用与空间推理能力</h2></header><div class=entry-content><p>📖 Technical Report | 🤗 Hugging Face| 🤖 ModelScope
7+
生成式图像分割 Next User Given the following instructions: little girl, pink, your monitors colors off friend p pink shirt girl; please perform referring segmentation on this image. Ming-Lite-Omni V1.5 User Please segment different classes in this image....</p></div><footer class=entry-footer><span title='2025-07-21 00:00:03 +0800 +0800'>2025年7月21日</span>&nbsp;·&nbsp;3 分钟&nbsp;·&nbsp;603 字&nbsp;·&nbsp;inclusionAI, Ant Group</footer><a class=entry-link aria-label="post link to Ming-lite-omni v1.5:全能模型再升级,效果与体验双优化" href=https://inclusionai.github.io/zh/blog/ming-lite-omni-1_5/></a></article><article class=post-entry><header class=entry-header><h2>M2-Reasoning: 赋予多模态大语言模型统一的通用与空间推理能力</h2></header><div class=entry-content><p>📖 Technical Report | 🤗 Hugging Face| 🤖 ModelScope
88
介绍 我们推出了 M2-Reasoning-7B,一个在通用与空间推理方面都表现卓越的模型。我们的方法融合了两项关键创新:(1) 一个全新的数据管道,生成了29.42万个高质量数据样本(其中16.8万用于冷启动微调,12.62万用于RLVR)。这些数据具有逻辑连贯的推理轨迹,并经过了全面评估。(2) 一种动态多任务训练策略,通过逐步优化来缓解数据间的冲突,并利用针对特定任务的奖励机制来提供定制化的激励信号。通过这种精心筛选的数据与先进训练方法的结合,M2-Reasoning-7B 在8个基准测试中创造了新的业界最佳水平(SOTA),在通用和空间推理领域均展现出卓越的性能。 📌 更新 [2025.07.14] 🔥 我们的技术报告已公开发布于 arxiv。 [2025.07.11] 🔥 M2-Reasoning模型开源: 🤗 Hugging Face、🤖 ModelScope。 主要特性 高质量的数据构建流程:我们设计并实现了一个多阶段的数据合成与筛选流程,能够生成大量的推理数据。 动态多任务训练策略:我们提出了一种高效的训练策略,能够有效应对数据异构性问题。该策略包括逐步动态优化,以缓解不同数据源之间的冲突,以及任务特定的奖励机制,提供定制化的激励信号。 统一的通用与空间推理模型:我们提出了 M2-Reasoning-7B,这是一款专为通用推理与空间推理任务而设计的多模态大语言模型(MLLM)。在8个不同的基准测试中进行的广泛评估表明,借助我们定制的数据和训练流程,M2-Reasoning在通用推理和空间推理领域均取得了新的SOTA成果。 评测 我们在通用推理和空间推理对模型进行了全面评估。我们的评估使用了一组多样化的公开基准测试,这些测试根据它们主要衡量的能力进行分类:
99
通用推理(数学与逻辑):为了评估这一能力,我们采用了六项基准测试:MathVista、MathVision、MathVerse、DynaMath、WeMath 和 LogicVista。 Models MathVista MathVision MathVerse DynaMath WeMath LogicVista Avg. (Δ) 基础规模通用模型 InternVL3-8B 70.5 30.0 38.5 25.7 39.5 44.5 41.4 InternVL3-9B 69.0 29.3 37.9 25.1 34.8 49.0 40.8 Qwen2.5-VL-7B 68.1 25.4 41.1 21.8 36.2 47....</p></div><footer class=entry-footer><span title='2025-07-11 00:00:03 +0800 +0800'>2025年7月11日</span>&nbsp;·&nbsp;4 分钟&nbsp;·&nbsp;736 字&nbsp;·&nbsp;inclusionAI, Ant Group</footer><a class=entry-link aria-label="post link to M2-Reasoning: 赋予多模态大语言模型统一的通用与空间推理能力" href=https://inclusionai.github.io/zh/blog/m2-reasoning/></a></article><article class=post-entry><header class=entry-header><h2>AWorld:为智能体自我演进提供运行环境</h2></header><div class=entry-content><p>“自我认知:最难的问题不是在有局限的情况下解决问题,而是发现自己的局限性” 目录 最新动态 — 项目最新更新与公告。 简介 — 项目概述与目标。 安装 — 步骤详尽的安装指南。 快速开始 — 使用示例,快速上手。 架构 — 多智能体系统设计解析。 演示 — 项目实际运行演示。 贡献 — 如何参与和贡献代码。 许可证 — 项目授权信息。 最新动态 🦤 [2025/07/07] AWorld 作为运行时现已准备好进行智能体训练。详情请参见自我改进部分。我们在 GAIA 测试中的得分已更新至 77.08。在演示部分了解如何构建 GAIA 运行时。 🦩 [2025/06/19] GAIA 测试分数提升至 72.43,新增本地运行模式,详见 ./README-local.md。 🐳 [2025/05/22] GAIA 评测、MCP 工具、AWorld 及模型现已集成于单一 Docker 镜像,详见 ./README-docker.md,演示视频。 🥳 [2025/05/13] 浏览器场景状态管理升级,视频处理 MCP server 增强,GAIA 验证分数 77.58(Pass@1 = 61.8),继续保持开源框架第一。详见 GAIA 排行榜。 ✨ [2025/04/23] GAIA 基准测试排名第三(69.7 分),Pass@1 = 58.8,开源框架第一。可用 python examples/gaia/run....</p></div><footer class=entry-footer><span title='2025-07-07 00:00:03 +0800 +0800'>2025年7月7日</span>&nbsp;·&nbsp;2 分钟&nbsp;·&nbsp;363 字&nbsp;·&nbsp;inclusionAI, Ant Group</footer><a class=entry-link aria-label="post link to AWorld:为智能体自我演进提供运行环境" href=https://inclusionai.github.io/zh/blog/aworld/></a></article><article class=post-entry><header class=entry-header><h2>Ming-Omni:一个用于感知与生成的统一多模态模型</h2></header><div class=entry-content><p>GITHUB 📑 Technical Report|📖Project Page |🤗 Hugging Face| 🤖 ModelScope
1010
介绍 Ming-lite-omni 是 Ming-omni 的轻量版,源自 Ling-lite,拥有 28 亿激活参数。Ming-lite-omni 是一个统一的多模态模型,能够处理图像、文本、音频和视频,并在语音和图像生成方面表现出较强能力。Ming-lite-omni 使用专用编码器从不同模态提取 token,然后由 Ling 处理,Ling 是一个 MoE 架构,配备了新提出的模态专用路由器。该设计使单一模型能在统一框架内高效处理和融合多模态输入,从而支持多样化任务,无需使用多个模型、任务专用微调或结构改动。重要的是,Ming-lite-omni 超越传统多模态模型,支持音频和图像生成。通过集成先进的音频解码器实现自然语音,以及利用 Ming-Lite-Uni 实现高质量图像生成,模型还能进行上下文感知聊天、文本转语音及多功能图像编辑。我们的实验结果表明,Ming-lite-omni 在所有模态上的统一感知与生成方面提供了强大解决方案。值得注意的是,Ming-lite-omni 是我们所知首个模态支持与 GPT-4o 匹配的开源模型,且我们发布了全部代码和模型权重,以促进社区进一步研究和发展。

zh/blog/index.xml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -11,7 +11,7 @@
1111
<generator>Hugo -- gohugo.io</generator>
1212
<lastBuildDate>Mon, 21 Jul 2025 00:00:03 +0800</lastBuildDate><atom:link href="https://inclusionai.github.io/zh/blog/index.xml" rel="self" type="application/rss+xml" />
1313
<item>
14-
<title>Ming-Lite-Omni V1.5:全能模型再升级,性能与体验双重突破!</title>
14+
<title>Ming-lite-omni v1.5:全能模型再升级,效果与体验双优化</title>
1515
<link>https://inclusionai.github.io/zh/blog/ming-lite-omni-1_5/</link>
1616
<pubDate>Mon, 21 Jul 2025 00:00:03 +0800</pubDate>
1717

0 commit comments

Comments
 (0)