TacRefineNet
Search documents
小米发布具身智能触觉抓取模型TacRefineNet,机器人ETF(562500)高位震荡,固高科技领涨
Mei Ri Jing Ji Xin Wen· 2026-02-09 06:47
截至今日14点30分,机器人ETF(562500)午后维持早盘态势,持续高位震荡,最新价报1.068元,较 开盘价上涨1.521%。持仓层面,该ETF跟踪的66只成分股55只个股上涨,固高科技大涨超8%,瑞松科 技、云天励飞、大族激光涨幅居前,形成强劲的核心跟涨梯队,成为盘面拉升的主要动力,板块内做多 情绪浓厚,无明显下跌标的形成拖累。流动性方面,该ETF成交额达7.44亿元,换手率2.99%,资金交 投节奏平稳,成交活跃度维持适中水平,场内做多资金逐步释放。 机器人ETF(562500)是全市场唯一规模超两百亿的机器人主题ETF,成分股覆盖人形机器人、工业机 器人、服务机器人等多个细分领域,帮助投资者一键布局机器人上中下游产业链。 随着成分股调整落地,机器人ETF(562500)跟踪的中证机器人指数人形机器人含量提升至近70%;此 次调仓精准剔除近期走势疲软的个股,纳入表现尚可的优质标的,成功实现"去弱留强"。 (文章来源:每日经济新闻) 消息方面,小米机器人团队发布具身智能领域研究成果——触觉驱动的机器人精细抓取微调模型 TacRefineNet。该模型为通用框架,仅依靠触觉即可实现毫米级位姿微调,无需视 ...
追觅CEO称每天研发投入4000万;泡泡玛特王宁:LABUBU去年全球销量超1亿只;OpenAI首款硬件命名曝光丨邦早报
创业邦· 2026-02-08 01:08
Group 1 - OpenAI's first hardware product is named Dime, resembling Apple's AirPods, and is expected to be released this year. The company has shifted its strategy from developing a high-end device to focusing on a simpler AI audio headset due to high hardware costs and component price increases caused by a shortage of storage chips [2] - Bubble Mart's founder Wang Ning announced that LABUBU's global sales exceeded 100 million units last year, with plans for the company to have over 10,000 employees and 100 million registered members by 2025 [3] - Tesla plans to increase investments in AI hardware and software, as well as energy sectors in China, with a projected capital expenditure exceeding $20 billion globally in 2026 [8] Group 2 - Amazon will change its policy on sharing reviews among product variants starting February 12, 2026, to improve customer trust and reduce return rates [8] - The German industrial output declined by 1.1% in 2025, marking the fourth consecutive year of decline, primarily due to weak automotive production [14] - The second-hand car market in China is projected to see a cumulative transaction volume of 20.11 million units in 2025, with a year-on-year increase of 2.5% [14]
小米推出触觉驱动的机器人精细抓取微调模型TacRefineNet
Xin Lang Cai Jing· 2026-02-07 11:28
Core Insights - The core focus of the article is on the advancements made by Xiaomi's robotics team in the field of embodied intelligence, specifically through the introduction of TacRefineNet, a framework that enables precise adjustments in object positioning without relying on visual input or 3D models [1] Group 1: Technology Development - Xiaomi's robotics team has announced a significant research achievement in embodied intelligence with the development of TacRefineNet [1] - TacRefineNet can achieve millimeter-level adjustments in positioning, which is crucial for tasks in various automotive manufacturing environments [1] - The technology allows for rapid iteration in reducing average positional error to millimeter levels after adjustments are made for imprecise grabs [1] Group 2: Future Prospects - The team has made technical details and experimental videos available to the public, indicating a commitment to transparency and ongoing development [1] - Further advancements and updates related to TacRefineNet are expected to be released soon, suggesting a proactive approach to innovation in robotics [1]