Ant Lingbo Embodied AI Challenge crowns TARS team after 48-hour real-robot hackathon
新智元 · wechat · 2026-11-16
The first Ant Lingbo Embodied AI Challenge ended in Shanghai, drawing 1,855 teams from 13 countries. Fifteen finalists had 48 hours to run the open-source LingBot-VLA2.0 on dual-arm robots across three physical tasks: fruit sorting, layered cosmetic storage, and dual-arm sweeping.
- Winners TARS (82.50): built a "gravity droop compensation table" from 100 real-machine tests with a 500g water bottle as counterweight; discarded the official 216 sweeping demos for 334 self-collected ones; invented two-person teleoperation and cut training step time from 15s to 3.4s.
- Runners-up EVA (74.77): started at 10% success, then ran an RSI data flywheel with DAgger human corrections — 7 iterations, 705 extra demos in 48 hours.
- Third SST (65.87): added only 13 demos, using "reversed video" frames plus DeepSeek parsing and SAM3 masks to isolate target pixels.
On-site failures showed why leaderboard worship doesn't survive physics: a 5cm table offset made a robot consistently place items one shelf too low; a bumped table destroyed 7,000 training steps; identical robot arms couldn't replay the same trajectory. Ant Lingbo also launched its LingBot developer ecosystem program; its open embodied stack is already used by 39 overseas projects.
More from Embodied
- YC-backed Ornadyne's flapping-wing drone was classified as a bird by Air Force detectors — import_jmr · 2026-11-16
- Uber starts offering Amazon's Zoox robotaxi rides in Las Vegas — Polymarket · 2026-11-16
- Tesla FSD fleet tops 7 billion city miles, adding ~500M in under four weeks — XFreeze · 2026-11-16
- Ropedia Releases Xperience 2: 1,500+ Scanned Real-World Scenes Paired With Egocentric Demos — liuziwei7 · 2026-11-16
- Hangzhou's Manycore Tech launches Aholo MS07 3D laser scanner with CHCNAV — teortaxesTex · 2026-11-16
- LeRobot 0.7 adds YAM, reBot B601 and Unitree G1 with NVIDIA SONIC support — mishig25 · 2026-11-16