Add research blog pages and refresh blog index

Author: xpqiu

assets/highlights.md

@@ -22,7 +22,7 @@ desc: A “Thinking with Video” paradigm—video-generation models like Sora-2
 标签: 多模态, 视频
 tags: Multimodal, Video
 图: assets/img/highlights/thinking-with-video.png
-链接: https://openaccess.thecvf.com/content/CVPR2026/html/Tong_Thinking_with_Video_Video_Generation_as_a_Promising_Multimodal_Reasoning_CVPR_2026_paper.html
+链接: /blog/cn/thinking-with-video/
 
 ---
 
@@ -34,7 +34,7 @@ desc: A scalable speech-generation foundation model supporting zero-shot voice c
 标签: 语音
 tags: Speech
 图: assets/img/highlights/moss-tts.png
-链接: https://arxiv.org/abs/2603.18090
+链接: /blog/cn/moss-tts/
 
 ---
 
@@ -46,7 +46,7 @@ desc: Introduces Reinforcement Learning from Community Feedback (RLCF): a "Scien
 标签: AI4AI
 tags: AI4AI
 图: assets/img/highlights/scientific-taste.png
-链接: https://arxiv.org/abs/2603.14473
+链接: /blog/cn/scientific-taste/
 
 ---
 
@@ -58,7 +58,7 @@ desc: An open-source joint video-audio generation model producing high-quality,
 标签: 视频, 语音
 tags: Video, Speech
 图: assets/img/highlights/mova.jpg
-链接: https://arxiv.org/abs/2602.08794
+链接: /blog/cn/mova/
 
 ---
 

assets/js/content-data.js

@@ -5,6 +5,34 @@
       tagline: '追求极致的开放 AI 研究'
     },
     highlights: [
+      {
+        title: { zh: 'Thinking with Video：用视频生成做多模态推理', en: 'Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm' },
+        desc: { zh: '提出“用视频思考”范式：让 Sora-2 等视频生成模型以视频帧为统一媒介进行多模态推理，弥补文字/图像难以刻画动态过程的不足。', en: 'A “Thinking with Video” paradigm where video-generation models like Sora-2 reason over generated video frames as a unified medium for dynamic multimodal reasoning.' },
+        date: '2026.6',
+        image: 'assets/img/highlights/thinking-with-video.png',
+        url: '/blog/cn/thinking-with-video/'
+      },
+      {
+        title: { zh: 'MOSS-TTS 技术报告', en: 'MOSS-TTS Technical Report' },
+        desc: { zh: '可扩展的语音生成基座模型，支持零样本音色克隆、时长与发音控制、流畅中英混说与长语音生成。', en: 'A scalable speech-generation foundation model supporting zero-shot voice cloning, duration and pronunciation control, smooth code-switching, and stable long-form generation.' },
+        date: '2026.3',
+        image: 'assets/img/highlights/moss-tts.png',
+        url: '/blog/cn/moss-tts/'
+      },
+      {
+        title: { zh: 'AI 也能学会“科学品味”', en: 'AI Can Learn Scientific Taste' },
+        desc: { zh: '提出“社区反馈强化学习”(RLCF)：用大规模引用信号训练 Scientific Judge，并对齐 Scientific Thinker 提出高影响力研究构想。', en: 'Introduces Reinforcement Learning from Community Feedback (RLCF), training a Scientific Judge and Scientific Thinker from large-scale scientific community signals.' },
+        date: '2026.3',
+        image: 'assets/img/highlights/scientific-taste.png',
+        url: '/blog/cn/scientific-taste/'
+      },
+      {
+        title: { zh: 'MOVA：可扩展的同步视频-音频生成', en: 'MOVA: Towards Scalable and Synchronized Video-Audio Generation' },
+        desc: { zh: '开源视频-音频联合生成模型，可同步生成高质量画面与声音，覆盖唇形同步语音、环境音效与内容匹配的音乐。', en: 'An open-source joint video-audio generation model producing high-quality, synchronized visuals and audio, including lip-synced speech, environmental sound effects, and content-aligned music.' },
+        date: '2026.2',
+        image: 'assets/img/highlights/mova.jpg',
+        url: '/blog/cn/mova/'
+      },
       {
         title: { zh: 'MOSS-Speech: 真语音到语音生成', en: 'MOSS-Speech: True Speech-to-Speech Generation' },
         desc: { zh: '原生端到端语音交互，无需任何中间文本引导', en: 'Native end-to-end speech interaction without any intermediate text guidance' },

assets/js/spa.js

@@ -389,12 +389,19 @@
     var items = [];
     (SPA_DATA.highlights || []).forEach(function (h) {
       var htags = (h.tags && (h.tags[currentLang] || h.tags.zh)) || [];
+      var url = h.url || 'javascript:void(0)';
+      var isBlogUrl = /^\/?blog\//.test(url);
+      if (isBlogUrl) {
+        url = currentLang === 'en'
+          ? url.replace(/^\/?blog\/cn\//, '/blog/en/')
+          : url.replace(/^\/?blog\/en\//, '/blog/cn/');
+      }
       items.push({
-        tags: htags.concat([t('blog.tag.external')]),
+        tags: htags.concat([t(isBlogUrl ? 'blog.tag.blog' : 'blog.tag.external')]),
         title: pick(h.title),
         desc: pick(h.desc),
         date: h.date, sort: dateKey(h.date), image: h.image,
-        url: h.url || 'javascript:void(0)', external: !!h.url
+        url: url, external: !!h.url && !isBlogUrl
       });
     });
     (SPA_DATA.blogPosts || []).forEach(function (p) {