SJTU's LIFT adds force sensing to VLAs with zero force-labeled pretraining data
jiqizhixin · x · 2026-10-05
Shanghai Jiao Tong University (Lu Cewu and Wen Chuan's team) and Qiangche present LIFT (Late Reactive Injection of Force for VLA Post-Training), accepted with high scores at CoRL 2026.
- The hard part of manipulation is contact: at the moment of touching, vision barely changes, so the model must infer task state from force alone.
- Force data is tightly bound to real robot hardware, expensive to collect, and hard to scale via UMI/ego-style methods; adding a new modality also risks retraining and catastrophic forgetting.
- LIFT keeps pretrained knowledge intact and teaches force perception purely through an online post-training phase — zero force-labeled pretraining examples, one post-training stage — enabling a VLA that feels force and reacts at high frequency.
More from Embodied
- Physical AI field session in Bengaluru to tackle post-deployment evals and retraining loops — carrycooldude · 2026-10-05
- Meta's NAVA-WAM pretrains robot action policies directly from action-free videos — meta · 2026-10-05
- Robotics' most-hyped model completed just 7 of 100 real manipulation tasks, demo reel hides the rest — carrycooldude · 2026-10-05
- Sergey Levine on robotics: hardware is good enough, the real gap is decision-making and data — 机器之心 · 2026-10-05
- Rebuilding the Muse AI toy on M5Stack with Muse Gadgets SDK, customized to Chinese UI — op7418 · 2026-10-05
- Guangzhou hotel deploys wheeled humanoid robots to greet guests and brew coffee — CyberRobooo · 2026-10-05