Ant's Lingbo open-sources LingBot-Touch tactile foundation model
11/10/2026 — 11/10, 19:21·6 sources (all)·6 reports (all)·In the 2026-11-10 briefing
Key takeaways
- On November 10, Ant Group's Lingbo team open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real-worl…
- across 26 real-robot tasks on two dexterous hands, four action models gained roughly 14 percentage points in average success rate. The techn…
Background & analysis
On November 10, 2026, Ant Group's Robbyant opened-sourced LingBot-Touch, a tactile foundation model for robotic manipulation. According to IT Home and 36Kr, the model was pretrained on nearly 5,000 hours of real tactile data spanning three mainstream sensing principles: magneto-tactile, vision-tactile, and piezoresistive. It converts different tactile signals into a single, unified representation that can be reused across sensors, dexterous hands, models, and tasks. 36Kr reported that the technical report, model weights, and code were released and made downloadable via Hugging Face, GitHub, and ModelScope, a Chinese model hub.
Leiphone noted the same day saw Claude Fable 5.2's release, adding that Robbyant CEO朱兴 described large general-purpose models and embodied foundation models as having a synergistic 1+1＞2 relationship. LingBot-Touch was reported to be integrated into LingBot-VLA. Zhidongxi provided a detailed result: on the Sharpa Wave dexterous hand, LingBot-VLA's average task success rate rose from 45.60% to 72.51%. InfoQ China added that across 26 real-robot tasks performed with two dexterous hands, Sharpa Wave and XHand1, the average success rate of four policy models—LingBot-VLA, π0.5, Diffusion Policy, and T-Rex—improved by roughly 14 to 26 percentage points once the tactile representation was enabled. QbitAI reported that the release targets robotic tactile perception but disclosed no parameter count or training details. As of now, LingBot-Touch is publicly accessible through its project page, Hugging Face, GitHub, and ModelScope.
AI-generated from 6 reports · updated 3 days ago
Latest turnAnt Group's robotics unit Robbyant has open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real tactile data that maps magnetic, vision-based and piezoresistive sensor signals into a single representation. Across 26 real-robot tasks on two dexterous hands, plugging in the representation lifted average success rates by roughly 14 to 26 percentage points for LingBot-VLA, π0.5, Diffusion Policy and T-Rex. The model, code and technical report are available on Hugging Face, GitHub and ModelScope.
- Heat index
- 25
- All sources
- 6
- All reports
- 6
- First seen
- 3 days ago
Reports on this story headlines open the original
Ant Group's robotics unit Robbyant has open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real tactile data that maps magnetic, vision-based and piezoresistive sensor signals into a single representation. Across 26 real-robot tasks on two dexterous hands, plugging in the representation lifted average success rates by roughly 14 to 26 percentage points for LingBot-VLA, π0.5, Diffusion Policy and T-Rex. The model, code and technical report are available on Hugging Face, GitHub and ModelScope.
InfoQ 中文AI score 64Ant Lingbo open-sourced LingBot-Touch, a tactile pretraining foundation model for dexterous manipulation, on November 10, according to ZhiDongXi. It unifies piezoresistive and vision-based tactile signals into reusable representations; in real-robot tests, LingBot-VLA's average success rate on the Sharpa Wave hand rose from 45.60% to 72.51%.
智东西AI score 56Ant Lingbo has open-sourced LingBot-Touch, a foundation model for robot touch that the company says gives machines a general sense of feel rather than relying on vision alone. The release was reported by QbitAI, which did not detail model size or training data.
量子位AI score 56On the same day Claude Fable 5.2 launched, Ant Lingbo open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real touch data spanning magnetic, vision-based and piezoresistive sensing. CEO Zhu Xing framed general LLMs and embodied foundation models as complementary; LingBot-Touch is already wired into LingBot-VLA, lifting task success by 14-26 points across 26 real-robot tests.
雷峰网AI score 64Robbyant, Ant Group's robotics unit, released LingBot-Touch on November 10, a tactile foundation model pretrained on nearly 5,000 hours of real tactile data. It covers magnetic, vision-based and piezoresistive sensing and converts their signals into one shared tactile representation usable across dexterous hands and action models. According to 36Kr, the technical report, model and code are open-sourced on Hugging Face, GitHub and ModelScope.
36氪AI score 53Ant's Lingbo team has open-sourced LingBot-Touch, a tactile foundation model pretrained on nearly 5,000 hours of real-world touch data across magnetic, visuotactile, and piezoresistive sensors. It maps different tactile signals into a unified representation reusable across sensors, dexterous hands, models, and tasks, and the team reports a 14-26 percentage point average task-success improvement over no-touch baselines in 26 real-robot tests.
IT之家AI score 58
Other stories people are talking about
- Heat index 2236Nvidia stock plunges after Jensen Huang resigns, with retail investors buying the dip35 sources
- Heat index 1463RisingGoogle Opens Gemini 4 Hydrogen to All Users as High-Speed Model21 sources
- Heat index 1429SurgeJensen Huang exits as NVIDIA CEO, leadership vacuum lingers16 sources
- Heat index 785SurgeAnthropic hires gas supply lead to power data center expansion after IPO10 sources
- Heat index 673SurgeMicrosoft races to fix global outage hitting Outlook, Teams and Copilot8 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion