Arsh Tangri

Robot Learning @ToyotaResearch | Prev: Robot Learning @PathRobotics, MS Robotics (CS) Student @HelpingHandsLab @Northeastern | GGMU!!

[Arsh Tangri @T_MUFC_A]

1/ 🚀Excited to share our latest work: Equivariant Offline Reinforcement Learning. Our work explores the use of equivariant neural networks for improving the performance of Offline RL algorithms for robotic-manipulation tasks, with small offline datasets.
https://t.co/TX2GbH390H

[Arsh Tangri @T_MUFC_A]

@cameron_smith_ As always, Tombradymaxxing. Legit!

[Arsh Tangri @T_MUFC_A]

@stolenpyjak @anjalie_f Congratulations @stolenpyjak

[Arsh Tangri @T_MUFC_A]

@omcamsmith @_krishna_murthy Tombradymaxxing ftw!

[Cameron Smith @cameron_smith_]

introducing claude-as-cad

i've been actually building robots and parts without knowing any CAD, using a claude-as-cad recipe -- here's how and you can too:

language-based prompting for CAD-agents is a good start, but it's imprecise -- these builds are difficult to steer exactly, among other issues

instead, you should have claude build you a claude-native CAD interface.
i've had success with this and found 3 important ingredients:

1. click to reference -- click any part of the build and the [3D coordinate, mesh face, object id] are copied to your clipboard to paste into chat
➡️ lets you be hyper-specific about what part you are referring to in chat
2. 3D primitive editor -- reposition/reorient any existing object; add a cube to the scene and scale/position for claude to use for reference

➡️ often times its hard to describe a position/transform in language and easier to illustrate it yourself to claude
3. hierarchical notes -- claude organizes separate builds into nested folders and writes skills that survive conversation compacting

➡️ saving different builds in folders with common skills, e.g. servo mounts, means once you get it right and save it, you can always reference it easily
link to install to your agent setup below ⬇️

[Arsh Tangri @T_MUFC_A]

@notsheesh01 Legit?

[Cameron Smith @cameron_smith_]

A paradox in robot learning:
robot policies are spatially dumb and data-inefficient but their image features are spatially robust and semantically rich.

Image encoders are spatially smart — features embed semantics + geometry + are multiview-consistent ...
but robot policies we build upon them are spatially dumb — hundreds of demos to learn simple pick+place and a small camera bump breaks them

To better bridge this gap, check out BIND: a new action head that binds each candidate robot action to its projecting 2D image feature(s)
BIND basically offers the network the info of '‘choosing this candidate robot action would move the robot EEF to this image feature."
Result: policies much more data efficient and robust to OOD viewpoints and OOD object positions (same rgb input, no 3D sensors, dense EEF trajectories out) 🧵

[Arsh Tangri @T_MUFC_A]

@pupscub @fdotinc Congratulations brother!

[Arsh Tangri @T_MUFC_A]

🚀 Excited to share that I’ll be in San Diego for NeurIPS 2025 this week! 🌴

My work right now focuses on Robot Learning, especially VLA Models for robotic manipulation and RL for real-world manipulation and control

Happy to exchange ideas, and discuss research directions.

[Arsh Tangri @T_MUFC_A]

These motions are so smooth!!!

[Tangible @tangiblerobots]

Hello, Eggie.

The world was built around humans. Eggie doesn't just look human, Eggie interacts like us.

Dexterous. Mobile. Compliant.

We’re building Eggie to be the smartest robot to ever walk on Earth. Join us.

Built from scratch and with love in California. ����

[Arsh Tangri @T_MUFC_A]

@NotoriousMOY @nostalgiafkninc Bruh?

[Arsh Tangri @T_MUFC_A]

@pupscub @primal_brainer Hahahaha

[Arsh Tangri @T_MUFC_A]

@NotoriousMOY @Banks My boi @pupscub is back at shitposting. Lfg!

[Arsh Tangri @T_MUFC_A]

@pupscub @Dexerto @grok can you confirm if we are back?

[Arsh Tangri @T_MUFC_A]

@pupscub @RespectfulMemes @grok @AskPerplexity @grok is an @AskPerplexity wrapppperrrrr

[Arsh Tangri @T_MUFC_A]

@NotoriousMOY @TheNotoriousMMA @NotoriousMOY you goat of twitter trash talk for a reason

[Arsh Tangri @T_MUFC_A]

@pupscub

[Boce Hu @boce_hu]

Grasp detection is crucial for robotic manipulation but remains challenging in SE(3).

We introduce our #CoRL2024 paper: OrbitGrasp, an SE(3)-equivariant grasp learning framework using spherical harmonics for 6-DoF grasp detection.

🌐 https://t.co/HzWP5OghY3

[Arsh Tangri @T_MUFC_A]

@chris_j_paxton @YuXiang_IRVL Maybe using Equivariant methods is the answer? (especially for table-top settings)

https://t.co/47plyqL9mX

[Arsh Tangri @T_MUFC_A]

@_onionesque Sergey Levine’s Deep RL course. (CS 285 Berkeley)