When every generalist robot model scores 95%+ on a benchmark, the numbers become meaningless.
What if we built a photorealistic benchmark that never saturates and can generate new scenes and tasks with AI Workflows in minutes?
We introduce RoboLab! 🧵(1/6)
Lots of people advocate against doing a PhD these days and I can relate to the sentiments. This is largely driven by the race towards “AGI/RSI” and the irrelevance of academia in this race. I’d like to make a case for doing a PhD specifically in computer science today.
If you don’t believe we are approaching singularity in the next few years, then naturally you’d agree that fundamental research breakthroughs are necessary and academia is historically the best place for long-term curiosity-driven research.
If you do believe, on the other hand, singularity means the value of “thinking” as a production tool will diminish. But the value of “thinking” as a form of human agency and freedom as well as intellectual entertainment will increase. In other words, humans think to understand our existence, exercise intellectual freedom, explore the unknown, and experience immense pleasure in this process. A PhD has never been just about teaching students to think so that they can produce, and will be less and less so in the future. It’s more about teaching students to think deeply and philosophically and subsequently, enjoy thinking itself. As @phillip_isola put it, we do it “for the love of the game”
QED
All in all, if you wanna join us on this journey to unlock your intellectual curiosity and deep philosophical thinking, please consider applying for PhD program at @umdcs!
https://t.co/ZKCH0VJ60k
📢 We're hiring (interns)!
The Seattle robotics lab @NVIDIARobotics is looking for talented PhD research interns to join us in 2027.
Interns at SRL will have the chance (depending on their preference) to either contribute to one of our ongoing group efforts or work on independent projects/topics of interest on topics including robot foundation models, agentic robotics, and precise manipulation.
Reach out to me if interested in working on robot foundation models 🤖🤖
https://t.co/IPlKIOqRCi
🚀 Flexible test-time scaling with a 4B model and just 8K-token context windows!
One of my favorite things about 🪽 Hermes is how many simple but powerful ideas come together to tackle a much broader challenge:
How do we teach a small model an understudied agentic capability and inadequately measured by existing benchmarks?
In our case, that capability is contextual reasoning: knowing when and how to delegate, what to preserve in compact reasoning summaries, and how to continue reasoning from them.
🎓 On-policy distillation has a teacher problem. SCOUT fixes it.
Then irony? OPD lets students learn from reasoning paths they actually visit. But their teachers must provide feedback on paths they rarely take themselves.
As these paths grow longer, even strong teachers struggle to provide reliable supervision.
Our idea is simple: Why not let the teacher and student learn together?
Introducing MILO, a step toward recursively self-improving agents for long-horizon, reasoning-intensive problems! 🧬
TL;DR: MILO evolves harnesses, but also the search process that discovers better harnesses.
It beats Claude Code and Codex on Terminal-Bench, outperforms methods like GEPA and Meta-Harness, and that's not all! It also discovers tighter upper bounds on three open math problems.
🧵👇
Can and should a general-purpose model serve as a robot policy?
In Astranomy, we study GPT-6 Astra for general pick-and-place manipulation. Astra sets a new SOTA on RoboLab-120, but real-time robot control remains an open challenge.
Full tech report:
https://t.co/EtpeflewvK
Native VLA/WAMs have very weak language-grounding and reasoning capabilities. With ARC, we uplift existing VLA/WAMs to gain ~30pp performance with no new robot action data. The capability is already in these models; we just haven’t fully unlocked it yet.
https://t.co/u3WOH9ErzK
Today we are introducing ARC, a reasoning recipe for robot foundation models that sets a new state of the art on RoboLab-120 and MolmoSpaces.
The prevailing way to improve robot foundation models is brute force: larger models, more robot demonstrations, costly training at scale. ARC needs none of that. With no new robot data and no foundation-scale training, π0.5 and Cosmos3-Nano-Policy gain up to:
📈 +50.0 pp on RoboLab-Reasoning-50
📈 +29.8 pp on RoboLab-120
📈 +27.2 pp on MolmoSpaces
To our knowledge, these gains are unprecedented.
📄 Paper: https://t.co/AZJ7Zfu3Nh
🌐 Website: https://t.co/xfjBysk91a
🧵 (1/7)
Excited to share that VoLo: A Physical Orchestrator for Open-Vocabulary Long-Horizon Manipulation has been accepted to #CoRL2026! 🎉🤖
VoLo explores agentic robotics for long-horizon, open-vocabulary tasks, where a VLM acts as the robot’s brain with memory to reason, plan, monitor, detect failures, recover, and orchestrate tools—including VLA/WAMs, vision models, and action primitives.
We also introduce RoboVoLo, a high-fidelity simulation benchmark with 126 tasks spanning common sense, memory/state tracking, complex references, and world knowledge, together with detailed diagnosis of current system limitations.
This work was done during my internship at @nvidia@NVIDIAAI@NVIDIARobotics. Huge thanks to all collaborators @HugoHadfield1, Alexander Zook, @mikacuy, @luke_ch_song, @erwincoumans, @xuningy, Faisal Ladhak, @BirchfieldStan, Jonathan Tremblay, and @robovalts!
🔗 Project: https://t.co/Q2pEymnWim
💻 Code: https://t.co/AwuKKrS0U8
📊 RoboVoLo Benchmark: https://t.co/9IULOYNVOw
Looking forward to discussions at CoRL 2026 — hope to meet many of you in person!
#CoRL2026 #AgenticRobotics #Robotics #EmbodiedAI #RobotLearning #PhysicalAI #NVIDIA
CPT is an important bridge between academic and industrial research.
I have benefited from it on both sides: as a graduate student, learning how frontier research is done in industry, and now as an industry researcher, working with graduate students to advance fundamental research.
Weakening that bridge would hurt both academia and industry.
We have just open sourced Droid+, a really simple and easy way to drive the Droid hardware system using modern web technologies and very few dependencies! I hope people find it useful! https://t.co/wA8GYI9Uuu
We released DROID+, a new and improved DROID control stack from SRL. We believe robots (and hardware) should be easy to setup and easy to use, and so we updated the DROID stack to be the same. @HugoHadfield1
https://t.co/1irJSpM8Bd
Here's pi0.5 doing our "hello world" task:
@_varunnair@viktor_malstrom Real world eval for single task policy is ok, but fundamentally you cannot do real world eval for generalist policies. Sim scaling is the only way to go.
Our team at SRL is hiring!
We're looking for people who dive deep, think systematically, and care deeply about doing rigorous work. If you're excited about understanding the science behind building foundation models, please apply!
We're hiring! The Seattle Robotics Lab (SRL) at NVIDIA is hiring a research scientist to work on robot foundation models. The focus will be on investigating the science behind building VLAs, WAMs, and world models for robotics, and leveraging the findings to propose the next generation of models and training paradigms.
We're seeking a world-class researcher (PhD grad) or senior researcher (3+ years of post-PhD experience) with expertise in training generalist policies. The researcher should aim to question conventional wisdom, push beyond the status quo, and dive into the foundations of general-purpose robot intelligence.
NVIDIA SRL provides a rare environment to conduct open research on fundamental and applied robotics, with the freedom to publish, collaborate broadly, and tackle big scientific questions at scale.
If this sounds like a strong fit, please reach out to us and apply here!
https://t.co/F6XHzOnLDn
https://t.co/extfNzBMDe
There's a growing gap between how fast generalist robot policies are improving and how well we can actually evaluate that progress, and understand when and why they fail.
I wrote a blog on our take: https://t.co/0vfDuQ2M2v
Want to discuss? Find me at #RSS2026!
The next wave of robot foundation models is here. ⚡
World-Action Models are emerging as a powerful new approach for physical AI, combining pretrained scene dynamics with action generation.
Learn more ➡️ https://t.co/3TdIkOkpiq
A fantastic technical deep dive into what's happening now and next in robot foundation models from @moritz_reuss. We're thinking deeply about *how* to build an efficient task generalist policy. Check it out!
World-Action Models (WAMs) have become the second dominant recipe for robot foundation models, next to classical VLAs.
So where do they come from, and how do they compare vs VLAs?
I wrote an small overview of the WAM landscape, with some personal takes:
https://t.co/6S4gH9tWTt
VoLo is a really good way to address long horizon manipulation problems is to combine a variety of tools and capabilities in robotics into a single system. We call this a physical orchestrator.
We've come a long way since code-as-policies.
Wonderful to be back from #CVPR2026, and excited to share the release of our follow-up work:
VoLo: A Physical Orchestrator for Open-Vocabulary Long-Horizon Manipulation
VoLo introduces the idea of a physical orchestrator for open-vocabulary, long-horizon manipulation. Our goal is to move toward robots that can reason, plan, act, monitor, and recover by adaptively using VLA/WAMs, vision models, and action primitives as tools.
We introduce three main contributions:
🤖 VoLoAgent — a physical orchestrator that plans, monitors, and recovers by adaptively using, halting, and redirecting robot actions with tools.
📊 RoboVoLo — a high-fidelity benchmark with 126 open-vocabulary long-horizon manipulation tasks spanning common sense, memory/state tracking, complex references, and world knowledge.
📈 A large-scale empirical study comparing action models, code-as-policy systems, TAMP-style systems, and ablations of the VoLoAgent orchestrator, complemented by real-robot experiments.
This work was done during my internship at @NVIDIA and would not have been possible without my brilliant collaborators: Hugo Hadfield, Alexander Zook, @mikacuy, @luke_ch_song, @erwincoumans, @xuningy, Faisal Ladhak, @qu_1006, @BirchfieldStan, Jonathan Tremblay, and @robovalts. Huge thanks to everyone!
🔗 Project: https://t.co/Q2pEymou7U
🔗 Previous work, SpaceTools: https://t.co/xNLUjiNG4j
#Robotics #EmbodiedAI #VisionLanguageModels #VLAModels #RobotLearning #NVIDIA #CVPR2026 #LongHorizonManipulation #AI #ComputerVision