@physical_int profile picture
Physical Intelligence @physical_int

Physical Intelligence (Pi), bringing AI into the physical world.

Joined March 2024
  • Posts 122
  • Following 30
  • Followers 45,666
83 Photos and videos
New work with @nvidia: evaluating robot policies entirely inside a world model. The policy acts, the model imagines the consequences, and the imagined evals predict real-world results. 🧵 real vs world-model rollout side by side📷
21
102
717
101,645
The model also knows when not to trust itself: its uncertainty estimates track how far generations deviate from reality. The uncertainty accurately predicts when model generation degrades
1
25
6,272
Joint work with the NVIDIA Cosmos team. Paper: arxiv.org/abs/2606.18610 Website: weichengtseng.github.io/sc3-…

SC3-Eval: Evaluating Robot Foundation Models via Self-Consistent...

Evaluating generalist robot manipulation policies in the real world is expensive, slow, and difficult to scale. Action-conditioned video world models offer a scalable alternative by simulating...

arxiv.org
3
46
5,543
π0.5 driving autonomous anastomosis!
Ryan McGuire
@r_mcguire
Apr 23
Replying to @r_mcguire
alephsurgical.com/blogs/rese…
2
14
172
74,332
Our newest model, π0.7, has some interesting emergent capabilities: it can control a new robot to fold shirts for which we had no shirt folding data, figure out how to use an appliance with language-based coaching, and perform a wide range of dexterous tasks all in one model!
4:02
62
317
2,534
458,214
We are still discovering what π0.7 can do. It's fun to play with and the results so far have been quite surprising!
1:43
2
4
91
18,393
To find out more, check out our blog post, videos, and full-length research paper: pi.website/pi07
2
4
63
12,609
Physical Intelligence reposted
Stanford MSL@StanfordMSL
Mar 28
π, But Make It Fly ✈️ We fine-tuned π0, a VLA model pretrained entirely on manipulators, to fly a drone that picks up objects, navigates through gates, and composes both skills from language commands.
1:00
14
43
365
102,305
We developed an RL method for fine-tuning our models for precise tasks in just a few hours or even minutes. Instead of training the whole model, we add an “RL token” output to π-0.6, our latest model, which is used by a tiny actor and critic to learn quickly with RL.
2:15
38
291
2,189
431,778
With RL, the robot can learn very precise tasks, like fastening a zip tie, and can actually do it more consistently and more quickly than even human teleoperation.
0:58
3
3
108
17,112
To learn more about RLT, check out our blog post: pi.website/research/rlt

Precise Manipulation with Efficient Online RL

We extract an RL Token from VLA models to enable fast online RL and improve throughput on precise tasks with few hours of data.

pi.website
13
137
23,645
Load more