蚂蚁灵波 / Ant Lingbo

Overview

Ant Lingbo / 蚂蚁灵波 is the 蚂蚁集团-incubated embodied-intelligence company represented by 沈宇军 / Shen Yujun in the wiki. Its route is robot-brain-first: build Embodied Native Foundation Models that can run across bodies while learning from physical sensors, real-world execution, and cross-embodiment data.

Current Profile

The new four-company debate strengthens Lingbo’s real-world-data and continuous-input stance. Shen does not reject simulation, but argues that true robot deployment must handle sensor noise, tactile-signal variation, physical imperfection, and real-time strategy changes while the robot is acting.

Lingbo also now carries a corporate-strategy role inside Ant’s AI portfolio. It is treated as an independently developed asset linked to the world-model and embodied-intelligence financing window, so the company is part of Ant’s broader market-pricing, talent-incentive, and multi-business valuation story as well as a technical robot-brain bet.

Key Characteristics

• Builds the robot brain before committing to a single universal robot body.
• Treats embodied-native models as distinct from simply attaching a language or video model to a robot.
• Prioritizes real sensors, spatial perception, video time, action, and execution feedback.
• Values simulation for some task-specific and evaluation uses while emphasizing real physical imperfections.
• Frames embodied models as possible tools for higher-level models and as future data sources for broader AI systems.
• Functions as one of Ant Group’s independently valued AI assets in the Bund Conference source.

Evidence

• Origin evidence: 147. 和蚂蚁灵波沈宇军聊：机器人原生基础模型、大脑和本体的关系、预训练与数据scale up、老师汤晓鸥 says Ant Lingbo was formed inside 蚂蚁集团 / Ant Group to explore physical-world AI and build a robot brain first.
• Model evidence: 147. 和蚂蚁灵波沈宇军聊：机器人原生基础模型、大脑和本体的关系、预训练与数据scale up、老师汤晓鸥 links Lingbo to Embodied Native Foundation Models, Robot Data Scale Up, and cross-body data cleaning.
• Sensor evidence: 当具身智能走到十字路口｜对谈苏度、蚂蚁灵波、自变量、破壳：四种一线判断 records Shen saying real robot sensors contain errors and tactile signals that differ from simulation in frequency, amplitude, and consistency.
• Continuous-input evidence: 当具身智能走到十字路口｜对谈苏度、蚂蚁灵波、自变量、破壳：四种一线判断 records Shen saying robots must keep receiving sensor input during inference and adjust strategy continuously.
• Corporate-strategy evidence: 273.逛完外滩大会，发现蚂蚁找到了新位置 groups Lingbo with Ant International, OceanBase, and Digital Ant as technology businesses being independently governed, incentivized, financed, or valued.

Qualifications

The wiki has not yet seen independent benchmarks showing how Lingbo’s robot brain compares with other VLA, world-model, or full-stack manipulation routes. The company’s emphasis on real data is also in productive tension with simulation-heavy routes such as Sudu’s. The new financing and valuation discussion remains source-scoped; the page does not independently verify Lingbo’s funding amount, investor terms, or commercial traction.

What Changed

• Converted the page to the synthesis-v1 entity schema.
• Added the Shizilukou Crossing debate as evidence for Lingbo’s sensor-noise and continuous-input stance.
• Added the Bund Conference source as evidence that Lingbo is also part of Ant Group’s independently valued AI-asset portfolio.

Relationships

• 沈宇军 / Shen Yujun - chief scientist and source speaker.
• 蚂蚁集团 / Ant Group - parent company and AI-first context.
• Embodied Native Foundation Models - core model route.
• Robot Data Scale Up - bottleneck Lingbo continues to emphasize.
• Embodied Robot Data Tradeoff - debate where Lingbo supplies the real-data side.
• 蚂蚁集团 / Ant Group - strategic and organizational portfolio context.