Ant Group Unveils LingBot-Touch Tactile Foundation Model for Robotic Manipulation
Fable 5.2吃掉机器人？「手感」这块，还得靠具身原生基模来补全
Ant Group has released LingBot-Touch, a tactile foundation model designed to enhance robotic manipulation through unified processing of resistive, magnetic, and vision-based sensor data. Trained on approximately 4,983 hours of real-world interaction data, the model aims to enable cross-sensor generalization for fine motor tasks such as grasping fragile objects and rotating items, demonstrating improved success rates in both physical and simulated environments.
Source: Machine Heart — Robotics on WeChat · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported duration
5,000 hours
View original evidence
过程中，团队收集近 5000 小时的真实触觉数据， 覆盖了磁触觉、压阻式和视触觉三种不同的传感技术，让模型学习不同传感器背后共同的接触规律。
Open source S14
Dataset details reported by the source
- Modalities
- tactile Open source S14
- Tasks
- manipulation Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Dataset licensing and commercial-use rights are not established here.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
过程中，团队收集近 5000 小时的真实触觉数据， 覆盖了磁触觉、压阻式和视触觉三种不同的传感技术，让模型学习不同传感器背后共同的接触规律。
Open source S14
要完成这类精细操作，机器人需要视觉、空间和触觉等多种物理模态协同工作。
Open source S11
以拧瓶盖为例，机器人手指要准确找到抓握位置，施加合适的力度，并根据瓶盖转动、指尖打滑的程度及时调整动作。
Open source S10
官方演示中，由该模型控制的机器人可以轻松地完成解魔方、拧瓶盖、画机器人等一系列操作任务。
Open source S4
这些触觉表征还能够接入不同的机器人动作模型，在不同灵巧手和真实操作任务中复用。
Open source S15
另外，在公开的 RoboDojo 基准测试中，Fable 5.2 也取得了 64.9% 的成绩。
Open source S5
从两个多前的 GPT-6 Astra 到现在的 Claude Fable 5.2，它们秀出的花式操作能力，让行业看到了通用大模型驱动机器人的更多可能。
Open source S7
但演示中的成功和仿真测试中的高分，是否就意味着机器人已经掌握了真正通用的物理操作能力？
Open source S8
当机器人「伸出手」，面对一个需要精细操作的任务，考验才刚刚开始。
Open source S9
Implications for data suppliers
RoboSignal interpretation and collection questions, not statements of buyer demand.
- Confirm the required data type and collection setting with the buyer; this source does not establish a complete collection specification.
- Compare the reported units and scope before using these quantities in a budget. Recording hours, sensor-hours and trajectories are different measures.
- Check the exact dataset version, license, modalities and task coverage against the intended use.
- Validate demand and acceptance criteria with a buyer before scaling. Publication, popularity and a research result do not establish a purchase commitment.
Source：Machine Heart — Robotics on WeChat · mp.weixin.qq.com