<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>智身居 · embodied</title><description>智身居里标了「embodied」的稿。</description><link>https://notesofone.com/</link><item><title>9 月 18 日：Apple 用可逆性标尺测没有外部重置的强化学习</title><link>https://notesofone.com/notes/daily-2026-09-18/</link><guid isPermaLink="true">https://notesofone.com/notes/daily-2026-09-18/</guid><description>Apple 的 REVERSAL-BENCH 用连续参数 ρ 控制环境可逆性，并给 reset oracle 当恢复真值。同批论文里还有 Agent 的选择性记忆、列描述和价值诱导。成绩表没有写进摘要的部分，这里不补。</description><pubDate>Fri, 18 Sep 2026 00:00:00 GMT</pubDate><category>Apple</category><category>REVERSAL-BENCH</category><category>强化学习</category><category>具身智能</category></item><item><title>TWIML：2026 年的智能机器人，差的不是又能多走几步</title><link>https://notesofone.com/notes/twiml-robots-2026-flexion/</link><guid isPermaLink="true">https://notesofone.com/notes/twiml-robots-2026-flexion/</guid><description>Flexion 的 Nikita Rudin 在 TWIML #760 里反复回到：运动在仿真里好看，不代表能部署；视觉一进环，sim2real 就变难；今天更务实的是分层，而不是一把梭端到端。</description><pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate><category>具身智能</category><category>播客</category><category>Sim2Real</category><category>人形</category><category>Flexion</category></item><item><title>Asimov：把开源腿、整机套件和 Sim2Real 写进同一条控制链</title><link>https://notesofone.com/notes/asimov-open-humanoid/</link><guid isPermaLink="true">https://notesofone.com/notes/asimov-open-humanoid/</guid><description>Menlo Research 的 Asimov 不是又一个演示视频。v0 开源双足、Asimov 1 DIY 套件、Processor-in-the-Loop 仿真，三件事合在一起，才对得上「模型接到身体」这一轨。</description><pubDate>Mon, 14 Sep 2026 00:00:00 GMT</pubDate><category>Asimov</category><category>开源人形</category><category>Sim2Real</category><category>MuJoCo</category><category>DIY</category></item><item><title>Open X-Embodiment：具身智能的公开数据集，不是又一个演示视频</title><link>https://notesofone.com/notes/open-x-embodiment/</link><guid isPermaLink="true">https://notesofone.com/notes/open-x-embodiment/</guid><description>多个实验室把机器人操作轨迹放到同一套格式里。真正要核对的是任务定义、失败样本和许可，而不是剪辑后的成功镜头。</description><pubDate>Thu, 03 Sep 2026 00:00:00 GMT</pubDate><category>机器人</category><category>数据集</category><category>RT-X</category><category>具身智能</category></item></channel></rss>