Ge Yan

@GeYan_21

CS PhD Student @UW | Previously @UCSanDiego | Intern @ToyotaResearch @allen_ai

Seattle, WA

geyan21.github.io

Replies Reposts Media

Pinned
Ge Yan
@GeYan_21
Aug 12
Are VLAs dead? No, and you don't have to choose between VLAs and WAMs. Introducing Flex-π: a multi-stream world-action model (WAM) that jointly predicts future RGB, 3D pointmaps and DINO semantics with actions in training, then deploys as a VLA, a full WAM, or anything in
00:00
8
Ge Yan
@GeYan_21
Aug 23
Thanks for the thoughtful comment. We fully agree that predicting future sensory states is indeed a world-model/dynamics objective, which is exactly why we call Flex-π in the paper a world-action model (WAM). In the post, we wanted to emphasize that our model can behave like a
Jitendra MALIK
@JitendraMalikCV
Aug 23
Nice idea to train to predict future sensory inputs along with action sequences. @ir413 Ilija Radosavovic et al (NeurIPS 2024) showed this for humanoid locomotion but the idea is perfectly general proceedings.neurips.cc/paper_files/pa… PS: Calling these models VLAs is a stretch.
1
Ge Yan
@GeYan_21
Aug 20
Flex-π is now open source. 💻 github.com/geyan21/flex-pi 🤗 huggingface.co/flex-pi Code, checkpoints, and all our real-robot data from the five bimanual YAM tasks. One checkpoint runs every mode, action-only to full joint. Large-scale pre-training checkpoints coming next. Give it
Ge Yan
@GeYan_21
Aug 12
Are VLAs dead? No, and you don't have to choose between VLAs and WAMs. Introducing Flex-π: a multi-stream world-action model (WAM) that jointly predicts future RGB, 3D pointmaps and DINO semantics with actions in training, then deploys as a VLA, a full WAM, or anything in
00:00
1
Ge Yan
@GeYan_21
Aug 12
Thank you Yixin! We were also really impressed when we first saw Flex-π successfully handle these high-precision and dexterous manipulation tasks. It was exciting to see how capable it is.
Yixin Zhang
@yzhang4179
Aug 12
Impressive to see world-action modeling help on tasks this hard — long-horizon, contact-rich bimanual manipulation with sub-mm precision. Flex-π makes a strong case that predicting the future can improve robot behavior when the task gets hard. Amazing work @GeYan_21 and team! 🔥
Ge Yan
@GeYan_21
Mar 26
Try it out! Great hardware for anyone who wants to do tactile sensing.
Binghao Huang
@binghao_huang
Mar 25
🤲Tactile sensing is powerful for robot manipulation, but hardware is still difficult to access, reproduce, and scale. 🎯That’s why we built FlexiTac: an open-source, low-cost, and scalable tactile sensing solution designed for real robotic systems. • Project page:
00:00
1