Lucy Shi
@lucy_x_shi
Followers
2K
Following
2K
Media
33
Statuses
244
CS PhD student @Stanford. Robotics research @physical_int. Interested in robots, rockets, and humans.
San Francisco, CA
Joined October 2021
Introducing Hi Robot – Hierarchical Interactive Robot.Our first step at @physical_int towards teaching robots to listen and think harder. A 🧵 on how we make robots more steerable 👇
12
45
277
Sadly I won’t be at ICML in person, but @chelseabfinn will be presenting Hi Robot tomorrow at 4:30pm in West Exhibition Hall B2-B3 (#W-403). Don’t miss it!.
Introducing Hi Robot – Hierarchical Interactive Robot.Our first step at @physical_int towards teaching robots to listen and think harder. A 🧵 on how we make robots more steerable 👇
1
8
97
If you are interested in solving complex long-horizon tasks, please join us at the 3rd workshop on Learning Effective Abstractions for Planning (LEAP) at @corl_conf!. 📅 Submission deadline: Sep 5.🐣 Early bird deadline: Aug 12.
We're excited to announce the third workshop on LEAP: Learning Effective Abstractions for Planning, to be held at #CoRL2025 @corl_conf!. Early submission deadline: Aug 12.Late submission deadline: Sep 5. Website link below 👇
1
2
31
SRT-H has been published in Science Robotics (and featured on the cover). :).Turns out, when we apply YAY Robot to surgical settings, the robot can perform real surgical procedures like gallbladder removal autonomously.
Introducing Hierarchical Surgical Robot Transformer (SRT-H), a language-guided policy for autonomous surgery🤖🏥. On the da Vinci robot, we perform a real surgical procedure on animal tissue. Collaboration b/w @JohnsHopkins & @Stanford
1
3
37
RT @Haoyu_Xiong_: Your bimanual manipulators might need a Robot Neck 🤖🦒. Introducing Vision in Action: Learning Active Perception from Huma….
0
86
0
RT @seohong_park: Q-learning is not yet scalable. I wrote a blog post about my thoughts on scalable RL algorithms.….
0
188
0
RT @kvablack: In LLM land, a slow model is annoying. In robotics, a slow model can be disastrous! Visible pauses at best, dangerously jerky….
0
60
0
RT @DannyDriess: How to build vision-language-action models that train fast, run fast & generalize? In our new paper, we formalize & analyz….
0
23
0
RT @marceltornev: Giving history to our robot policies is crucial to solve a variety of daily tasks. However, diffusion policies get worse….
0
40
0
RT @t_k_233: Humanoid robots should not be black boxes 🔒 or budget-busters 💸!. Meet Berkeley Humanoid Lite!.▹ 100% open source & under $5k….
0
91
0
at Pi, we’re working with partners deploying real robots in the real world — and scaling data the smart way. interested in collaborating? shoot us a message ✉️
1
0
7
another gem: verbal instruction data. imagine teleoperating a robot — but just by talking to it. easy to collect, surprisingly rich signal — if your low-level policy is steerable.
1
0
5
why? our best guess: it teaches the model to "think harder" — to plan and structure behavior, even when used as a co-training signal.
1
0
4
what surprised me most: high-level semantic prediction isn’t just useful for open-ended instruction following (like in Hi Robot) — it actually boosts policy performance directly. especially for complex, long-horizon tasks.
Introducing Hi Robot – Hierarchical Interactive Robot.Our first step at @physical_int towards teaching robots to listen and think harder. A 🧵 on how we make robots more steerable 👇
1
0
4
the recipe?.co-train on:.- diverse robot data across embodiments.- web data.- high-level semantic prediction tasks. this simple mix turns out to be surprisingly effective.
1
0
10
Tired: Collect tons of data and retrain the model for every new environment (💸).Wired: Zero-shot generalization to new environments 🏠. we've been cooking something to make that possible — and I'm excited to share it today. a short 🧵:.
We got a robot to clean up homes that were never seen in its training data! Our new model, π-0.5, aims to tackle open-world generalization. We took our robot into homes that were not in the training data and asked it to clean kitchens and bedrooms. More below⤵️
1
5
97
RT @svlevine: We made π0 “think harder”: our new Hierarchical Interactive Robot (Hi Robot) method “thinks” through complex tasks and prompt….
0
52
0
RT @chelseabfinn: Can we prompt robots, just like we prompt language models?. With hierarchy of VLA models + LLM-generated data, robots can….
0
68
0
Hi Robot isn’t just about following instructions - it’s about understanding, reasoning, and adapting to human goals in the wild. More details in the paper: 🔗
pi.website
Physical Intelligence is bringing general-purpose AI into the physical world.
0
1
6
Huge thanks to @chelseabfinn, @svlevine, @brian_ichter, @michael_equi, and @hausman_k for your guidance and support. Grateful to @KarlPertsch, @QuanVng, Jimmy Tanner, @annawalling, Haohuan Wang, @NiccoloFusai, Adrian Li-Bell, @lachygroom, @DannyDriess, and the Pi team - this.
1
0
10