David Held

Associate Professor at Carnegie Mellon University | he/him

[David Held @davheld]

CMU Robotics Institute is hiring faculty this year!

RI is an amazing place to work. It's like Disneyland for robotics. And the students are exceptional! They have a "get it done" spirit that you won't find anywhere else.

We're hiring across all tracks.
https://t.co/ZUNhJowbkJ

[Thomas Weng @ CoRL 2026 @thomas_weng]

Excited to share that I’m now leading a research team at @rai_inst! We’re training dexterous policies in simulation to be deployed and refined on custom hardware.

We may be looking for interns in this area soon. I’ll be at #CoRL2026 in Austin — DM me if you’d like to chat!

[Kallol Saha ✈️ CoRL '26 @_ksaha]

🚨Introducing DISCO🪩: Distilled Tree Search over Continuous Actions (Best Paper Finalist🏆 at #CoRL2026)

DISCO combines learned policies with Monte-Carlo Tree Search (MCTS) over continuous actions for long-horizon manipulation tasks.

A thread 🧵👇

[Ying Yuan ✈️ CoRL @Ying_yyyyyyyy]

Can learning to grasp teach a robot to use tools? 🤖🖐️

Introducing our #CoRL2026 paper:
“From Grasps to Dexterity: Large-Scale Grasp Pretraining for Dexterous Manipulation”
We turn grasp data into pretraining for autonomous tool use.

🎥 Project & videos: https://t.co/KGkNkyheYP
🧵👇 (1/9)

[Yining Hong @CoRL @yining_hong]

Thrilled to share that our paper DISCO: Distilled Tree Search over Continuous Actions has been selected as a 𝐂𝐨𝐑𝐋 𝟐𝟎𝟐𝟔 𝐁𝐞𝐬𝐭 𝐏𝐚𝐩𝐞𝐫 𝐀𝐰𝐚𝐫𝐝 𝐅𝐢𝐧𝐚𝐥𝐢𝐬𝐭! 🏆 Congrats to @_ksaha 🎉

Can robots learn to plan for the long horizon without expensive search at deployment?

DISCO uses Monte Carlo Tree Search in simulation and iteratively distills its discoveries into a policy that can execute directly, without search at test time. We demonstrate strong performance on long-horizon object packing and stacking, including real-world transfer!

📍 Catch us in Austin!
🎤 Oral: Nov 10, 11:00–11:30 AM | Hilton Austin (Session O6)
📌 Poster: Nov 10, 5:00–6:30 PM | JW Marriott Austin (Session P2, Poster #176)

🔗 https://t.co/SXgvbkcQTj

[Yufei Wang @YufeiWang25]

Life update: I recently graduated from CMU RI, and last year I went through the academic faculty job market in robot learning. I was fortunate to receive a lot of help and support from many people throughout the process.

Compared to industry job applications, there are relatively few resources available online about the faculty application process. So I decided to open-source all of my application materials from last year here:

https://t.co/4bDycVLxjO

I hope these materials can be helpful to others who are considering pursuing an academic career!

[Yishu Li ✈️ CoRL 2026 @LisaYishu]

Robots can’t always see the physics they act under: hidden mass, friction, kinematic constraints. A scalar reward only says "failed." The outcome of an action says much more.

Introducing SCOUT: test-time policy adaptation from action-outcome feedback. Sim-trained, adapts in the real world in a few tries 🧵👇 #CORL2026

[David Held @davheld]

@sha2nk_t I think "human data" includes UMI-style data collection?

[David Held @davheld]

@leto__jean Economists often present ongoing projects at conferences to get feedback. The challenge in AI-related fields is the short time between “here is an idea” and “here is a working implementation.” Sharing an idea can risk being scooped unless you remain intentionally vague about it.

[Andrew Curran @AndrewCurran_]

A man in Australia asked his agent (Claude running on OpenClaw) to book him a spot in a popular gym class. The agent found a software vulnerability that let it book the class weeks further ahead than should have been possible. When the user then asked if it could move him up the waitlist, the agent discovered the API had no authorisation checks on cancelling other people’s reservations, so it cancelled the person in the first spot and moved him up the list.

Some people will call this misalignment, but his agent was perfectly aligned to him - it was only trying to help its user get what he wanted. The most important thing about this story, in my opinion, is that it gives you a window into what is about to start happening on a massive scale once millions of people have an agent trying to get their beloved users the best seats, bookings, appointments or reservations through absolutely any means necessary.

[Sriram Krishna ✈️ CoRL '26 @sriramsk99]

Boo! 👻 GHOST has been accepted to #RSS2026!

Can we learn manipulation skills from human video (without action retargeting)?

Yes! GHOST learns generalizable manipulation skills by training a hierarchical policy with an embodiment-agnostic sub-goal predictor + embodiment-specific controller.

🧵(1/7)

[Ellina Zhang @EllinaZhang]

🧩 #ICML2026 💥 How can a model discover the 3D objects in a scene—their shape, color, and position—without any labels? Introducing 3D-DLP, a self-supervised object-centric model that decomposes colored 3D scenes (RGB-D and voxels) into a set of 3D latent particles.

[Zohar Rimon 🛩️ CoRL @ZoharRimon]

Excited to present the first-ever 3D, generalizable, handheld tactile imaging system! 👇

Our system learns to visualize the inner structure of soft objects purely through touch -
"More with LESS: Local Scene Representation for Tactile Imaging," has been accepted to #RSS2026 🇦🇺

Curious how we leverage the locality of touch for large-scale tactile SSL? Read on! 🧵

[David Held @davheld]

@Philip_MIT A friendly tip - keep the colors in the legend consistent across videos

[David Held @davheld]

@jon_barron @alexandertmai @KostasPenn @GeorgiaChal @vincesitzmann @anand_bhattad Right but it's an intermediate rollout of tokens before the final "answer" is produced. Kind of like "planning".

[David Held @davheld]

@jon_barron @alexandertmai @KostasPenn @GeorgiaChal @vincesitzmann @anand_bhattad Regarding LLMs, is that really true? What about chain-of-thought reasoning?

[David Held @davheld]

@thomas_weng @lakshitadodeja @rai_inst Proud PhD advisor here, great work!

[Thomas Weng @ CoRL 2026 @thomas_weng]

My first last-author paper, with @lakshitadodeja and @rai_inst collaborators!

When deploying Behavior Cloning (BC) policies, we encounter performance gaps due to compounding error, subtle task shifts, etc.

We want to be able to post-train BC policies with a few online interactions to quickly improve performance.

Q2RL is what we hoped for in a robot post-training algorithm:
- ✅it starts with BC policy rollouts (no need for large, diverse datasets), and
- ✅it's sample-efficient enough for real-world online reinforcement learning!

See the thread and linked website for more details. Paper and code are released!

[Tal Daniel @TalDaniel8]

I’m attending #ICLR2026 in Rio this week to present LPWM! Friday April 24 Poster Session 3 10:30, Oral Session 4B 3:15p. Happy to chat about self-sup object-centric learning and world models. I’ll be on the job market soon and looking for exciting opportunities!

#ICLR @iclr_conf

[David Held @davheld]

@GlenBerseth Link to slides?