Research @AppliedInt, MS @PortalCornell, Robot Learning
Joined May 2016
- Tweets255
- Following203
- Followers175
- Likes813
Prithwish Dan retweeted
🚨How much can we trust robot policy leaderboards?
Benchmarks play a key role in advancing robot learning, but the rankings we get from limited evaluations can be surprisingly misleading.
In AnyRank, we combine adaptive sequential testing with graphical hypothesis testing to reliably rank policies at any stopping time, using up to 50% fewer real-robot trials than rigorous baselines.
Check out @ApurvaBadithela's new paper! arxiv.org/abs/2610.14603
How many real robot trials do you need to reliably rank a set of policies?
Introducing AnyRank: reliable multi-policy ranking in as few trials as possible, with rankings that stay valid whenever you stop.
Excited to present this work at #CORL2026!
Paper: arxiv.org/abs/2610.14603v1
Website: abadithela.github.io/anyrank…
Prithwish Dan retweeted
SpatialDex learns 3D-grounded bimanual dexterity from 1M RGB-only human video clips. It reconstructs scene and hands in each video, aligns them at one scale, and trains the policy on both, nearly doubling the success of prior VLAs on real tasks. spatial-dex.github.io/
Prithwish Dan retweeted
Simulation is still our bread and butter and it’s only becoming clearer that it’ll unlock robotics at scale
Prithwish Dan retweeted
[8/n] By the way, for the simulation folks wondering how we simulated tight-clearance parts: the usual tools didn't work for us. SDF was so slow and hard to tune. Off-the-shelf convex decomposition like V-HACD or CoACD filled the clearance, so parts that should fit jam.
So we built mjfit. It decomposes common assembly parts (pegs and holes of any outline, threads, and gears) into convex pieces that keep the clearance exactly as designed, and it runs an order of magnitude faster than SDF.
It's open source: github.com/younghyopark/mjfi…
Prithwish Dan retweeted
Can a robot assemble tight-clearance parts using only its proprioception? No camera, no tactile sensor, no F/T sensor, no joint torque sensing, no priors over the object geometries or poses: just joint position readings.
It actually can! Blind policies trained in simulation insert pegs, mesh gears, and thread nuts, and the peg policy works zero-shot in the real world. Here's what we found 🧵
Joint work with @BronarsToni, my wonderful co-first author, who's on the job market right now! Advised by @kushalk_ and @pulkitology.
Prithwish Dan retweeted
Every robot already carries contact sensors: its own joints. When a robot touches something, its joints drift slightly off their commands, and the encoders pick it up. Legged robots and humanoids already leverage this signal to walk blind - but is it enough for blind precise assembly?
We find that it is! Our policies can insert pegs, mesh gears, and thread nuts using only the robot's joint positions - no camera, no tactile, and no object poses. And the peg policy works zero-shot in the real world! 🧵
w/ my amazing co-first author @younghyo_park, and advised by @kushalk_ and @pulkitology.
Prithwish Dan retweeted
Resets are critical for RL! For dex hands, human demos are a rich source of reset states. RL can learn the dynamics between the states.
Awesome work by @prithwish_dan !!
Prithwish Dan retweeted
Reinforcement Learning (RL) in simulation is a scalable strategy to learn dexterous skills, but naive exploration from scratch often fails to discover manipulation strategies for diverse objects.
At @AppliedInt, we propose X-Reset, a straightforward method to resolve the exploration problem for generalist RL policies by randomly sampling resets from human hand-object states.
- We train a single dexterous policy to manipulate a variety of objects to any goal pose, ranging from small markers to heavy powerdrills with zero teleoperation data
- The learned policy zero-shot generalizes well beyond the training set (DexYCB), and transfers from sim-to-real using object pose-tracking models
We are continuously expanding robotics research at @AppliedInt beyond X-Reset, with the bounds of simulation and synthetic data for dexterous manipulation pushed. Stay tuned and join us!
🌐 Project Page: xreset-applied.github.io
📄 Paper: arxiv.org/abs/2609.35715
Prithwish Dan retweeted
Dexterous manipulation in simulation has an exploration problem.
Here is how our team is fixing this problem.
Exploration is hard for RL when training high-DoF dex hands ✋
X-Reset makes generalist object-centric RL easier by resetting to human hand-object states 🙋♀️
We mitigate the exploration problem, and train a single policy to manipulate a variety of objects to any goal pose 🧵
X-Reset zero-shot transfers behaviors from sim→real using off-the-shelf pose-tracking models. Check out these recovery behaviors under @chaitanya1cha’s perturbations in the realworld, coordinating the arm along with in-hand reorientation skills!
This work was done at @AppliedInt in collaboration w/ Chenyang Ma, @Wei_ZHAN_ 🙏
Check out our work 👇
🌐 Project Page: xreset-applied.github.io
📄 Paper: arxiv.org/abs/2609.35715
Prithwish Dan retweeted
Human fingers move in coordinated patterns. Can those patterns help robots explore?
Excited to share EigenDEXplore! 🤖✋
Key idea: add exploration noise along directions learned from human hand motion, while leaving robots free to discover motions of their own
See @harshgupta's 🧵
Prithwish Dan retweeted
How can we train better behavior from imperfect robot demos?
In NEEDLEWORK, we study how we can get stitching to work on robot data, and more importantly, why these connections help.
Website: needle-work.github.io
Paper: arxiv.org/abs/2610.02339
Prithwish Dan retweeted
How can generalist policies learn new skills without losing their pretrained prior?
We introduce Proxy Policy Steering (PPS), which uses proxy policies to steer the frozen base policy at inference, adding task-specific behavior without touching its weights.
Prithwish Dan retweeted
Setting up new tasks with Play2Perfect is super easy!
Here are 2 additional envs + assets that Opus 5 generated from SCRATCH. The same RL training code works without any modifications. Go play with the awesome demo @tylerlum23 created :)
play2perfect.github.io/
Excited to share that Play2Perfect was accepted to #CoRL2026!
We also added an interactive demo to our project website 🖥️
Watch Play2Perfect perform precise, contact-rich assembly in MuJoCo, right in your browser!
Zero-shot sim-to-sim transfer from Isaac Sim to MuJoCo.