具身智能赛道正陷入数据可信度危机。2026年上半年该赛道融资规模超过935亿元，融资事件达322笔，百亿估值独角兽企业从2025年同期的3家增至22家[1]。然而繁荣表象下，行业涌现出超过20个评测榜单，企业通过榜单排名造势融资已成行业惯例，导致评测体系严重失序。
RoboArena榜单事件暴露了分布式评测的系统性漏洞。数据显示，该榜单上72%的分数仅来自两个注册账户，其中ECUST Robot Lab对千寻智能Spirit v1.6进行了179次评测且胜率高达97%，而英伟达官方同款模型仅测试21次胜率为0%[1]。这两个高频评测账号均在5月底集中注册[1]。智源研究院院长王仲远直言"现阶段榜单并不完全可信"[1]，指向行业内部分账户刷分、专项针对榜单优化现象普遍存在的问题。
行业当前正经历从泡沫期向产业化成熟期的过渡。技术路线分散、缺乏统一的价值标尺，使得真实场景落地能力与榜单排名严重脱节[1]。为规范发展方向，《全球工业具身智能导则》编制已正式启动[1]，WAIC 2026也对机器人提出了零遥操、纯自主运行的新要求[1]。
The embodied artificial intelligence sector has raised over 93.5 billion yuan in the first half of 2026, attracting 322 financing events [1]. However, this rapid growth has given rise to more than 20 competing evaluation rankings across the industry [1]. Using high rankings to generate momentum for fundraising has become standard practice, with at least 22 unicorn enterprises valued at over one billion yuan emerging in this period, a dramatic increase from just three companies in the same period of 2025 [1].
The rapid proliferation of rankings has exposed significant vulnerabilities in the sector's evaluation systems. The RoboArena ranking scandal revealed that 72% of scores came from just two registered accounts, with ECUST Robot Lab conducting 179 evaluations of Qianxun Intelligence's Spirit v1.6 model at a 97% win rate, while Nvidia's official model of the same architecture achieved only a 0% win rate across 21 tests [1]. Both high-frequency evaluation accounts were registered in late May, suggesting coordinated activity [1]. According to Wang Zhongyuan, director of Zhiyuan Research Institute, "current rankings are not entirely trustworthy" [1]. The disconnect between benchmark performance and practical application success reflects a broader industry challenge, as technical pathways remain fragmented and real-world deployment capabilities diverge significantly from ranking positions [1].
As the sector transitions from a speculative bubble phase toward industrial maturity, efforts are underway to establish industry standards. The World Artificial Intelligence Conference 2026 is requiring robots to demonstrate zero teleoperation and fully autonomous operation [1], while the formal initiation of the Global Industrial Embodied Intelligence Guidelines marks an attempt to create unified evaluation criteria [1].
