PhDing @UCBerkeley @berkeley_ai advised by Prof. @Ken_Goldberg. Robotics, AI | Intern @xdofai
Berkeley, CA
Joined June 2023
- Tweets128
- Following235
- Followers460
- Likes461
Justin Yu retweeted
Announcing: the XDOF Robotics Leaderboard
Rigorous evaluations on real robots are an important and overlooked ingredient in solving general purpose robotics.
Our goal is to provide trusted evaluations, with meaningful benchmark design and thorough rollout execution that researchers can rely on.
The Robotics Leaderboard is where we’ll publish how frontier models perform against our benchmarks. We’ll update the leaderboard with every model we evaluate.
Justin Yu retweeted
Last year I trained the advantage-conditioned VLA for π*0.6 during my internship @physical_int. Now, I have a much improved recipe.
Introducing ✨Advantage Guidance✨, a performant RL method with stable advantage-conditioned supervised training and contrastive guidance sampling
Justin Yu retweeted
Introducing Embodied First Exam:
One line of code to run coding agents in 1,090 embodied tasks across 31 simulation suites and 44 robots.
Can your agent go far beyond pick-and-place:
Puzzles with spatial constraints;
Olympic sports with multi-robot collab;
Real-time eval, where the world won't pause;
Dynamic, agile, whole-body control;
AutoResearch with sim2sim
Try your own agent:
harbor run -d embodied-first-exam -a -m
Justin Yu retweeted
My favorite takeaway from LBM paper is that scaling evaluations is also scaling. The ability to scale evaluations will be critical in the era of agentic robotics!
Justin Yu retweeted
We heard the hype around LLMs and robotics, so we put Astra and Fable to the test.
16 tasks testing 13 capabilities.
Our finding: both did surprisingly well on reasoning and planning and struggled with dexterity. But Astra did better overall.
Here’s what happened:
Justin Yu retweeted
Real2sim has made impressive progress with coding agents. But what about sim2real, especially the visual gap? Meet RoboRender: turning simulated trajectories into photorealistic videos for policy training, achieving 71% average success in zero-shot real-world robot deployment.
Justin Yu retweeted
Amazed by Astra controlling robots? It turns out other LLMs can do it too – even the text-only ones like DeepSeek. We also find Astra did best watching and correcting every move live; Fable did best building its own tools and policy in code first.
We introduce SEAR, a new benchmark and framework to rigorously evaluate and study how LLM agents self evolve on robots.
🔁📈The loop: within a controlled time budget, an agent and its library update a policy → the policy runs in simulation or on hardware → a judge returns feedback → the library carries the skills and lessons into the next task.
💻🚏 Self-evolving agents operate in myriad ways - we study all of them at once: the policy can be code the agent wrote (code-as-policy), the agent itself in the control loop (agent-as-policy), or a VLA it trained (autoresearch).
🧑🔬🤖We spent 4+ years of accumulated agent evolving time across simulation and real hardware and tested 7 frontier models (GPT-6-Astra, Fable 5.1, Opus 5, GPT-5.6-sol, Gemini-3.6-flash, Kimi-k3, DeepSeek-v4-flash) on 330 sim tasks, 2 physics engines, and real robots with different embodiments.
🌐Project overview: sear.bot
A thread 🧵1/5
Justin Yu retweeted
Practice makes perfect. A truly agentic robot shouldn’t just know what to do. It should also figure out how to get better.
What if a robot could watch a demonstration, figure out what it needs to practice, and learn how to get better through practice?
Introducing RPG: Guided Self-Improvement for Embodied Agents, a real-to-sim-to-real pipeline that turns real-world demonstrations into targeted practice.
RPG reconstructs practice tasks from demonstrations, uses demo videos to guide skill improvement, and continually updates a shared skill system from failures.
On real robots, RPG achieves 30/30 successes across 3 tasks. In simulation, it reaches 95.0% success across 22 manipulation tasks after 15 rounds of self-improvement.
Demo → Reconstruct → Guided Practice → Self-Improve → Go Real
Justin Yu retweeted
Very excited to release EyeRobot 2.0! We also built a simulation environment with 6 manipulation tasks and around 1,800 evaluation trials. Demos were collected in VR with the same GELLO-style setup we use on the real robot, and two tasks (pot and tape) mirror our real-world ones. Sim let us iterate much faster and test more ideas at scale, and the same overall trends carried over to the real robot.
Almost all bimanual robots include wrist cameras. They add bulk, cost, and complexity, but are critical for performance. Are they really needed?
Excited to finally share EyeRobot 2.0, where we use active gaze from an ego stereo camera instead 🧵 (with @kushtimusprime)
Justin Yu retweeted
Excited to release EyeRobot 2.0! We noticed it is better at dealing with distractors. When the gaze policy looks at the right target, the policy can recover
Almost all bimanual robots include wrist cameras. They add bulk, cost, and complexity, but are critical for performance. Are they really needed?
Excited to finally share EyeRobot 2.0, where we use active gaze from an ego stereo camera instead 🧵 (with @kushtimusprime)
Justin Yu retweeted
Almost all bimanual robots include wrist cameras. They add bulk, cost, and complexity, but are critical for performance. Are they really needed?
Excited to finally share EyeRobot 2.0, where we use active gaze from an ego stereo camera instead 🧵 (with @kushtimusprime)
Justin Yu retweeted
Our paper RATs has been selected as CoRL spotlight!
We apply "sleep time scaling" to agentic robotics learning. The agent aquires skills during play and then uses these skills at test time. Check out our project page at : playful-rats.github.io/ @junyi42
Excited to release our new work: Playful Agentic Robot Learning w/ @junyi42!
Instead of relying on test-time scaling, we found that a "pretraining stage" through curious play enables robots to discover general skills before any tasks are assigned.
🌐 playful-rats.github.io
Justin Yu retweeted
can all the eval companies start reporting throughput metrics
People watching GPT-6 Astra demos on X may not realise how slow it actually is in real time. Here’s a rollout of GPT-6 Astra controlling the MolmoAct2 YAMs to complete this task in real-time.
There is potential, but i don’t think it’s practical for real-world robotics deployment yet.
I applaud your patience if you make it through the whole video.
Justin Yu retweeted
Can robots help in data centers? It’s tricky when all the cables are the same color.
Introducing TRACE: bi-directional tracing + interactive perception for monochrome cables, @ #IROS2026 from @AUTOLab_Cal @UCBerkeley (1/8)
Justin Yu retweeted
Incredibly excited to return to ETH Zurich as Assistant Professor of Robotics and Artificial Intelligence next summer. I'll be recruiting PhD students to build a new lab on general-purpose embodied intelligence for humanoid robots and dexterous manipulation. Details coming soon!
Justin Yu retweeted
2/ WARP-RM
The first bottleneck was the quality of the training signal.
WARP-RM learns a dense relative-progress signal from successful demonstrations using time-warp augmentation.
The key idea: progress shouldn't depend on the exact timestamp.
The same behavior can happen faster, slower, or with different temporal dynamics, but it should still represent the same underlying progress toward the goal.
WARP-RM learns that signal, giving us a way to distinguish more useful vs. less useful parts of a demonstration instead of treating every frame equally.
@uynitsuj
Justin Yu retweeted
We may teach robots active perception through unified data–model–hardware co-design~ 🤖👀
Introducing ActiveScale: a small step toward a unified way to learn active perception with egocentric human data, single-operator teleop, and a pose-grounded VLA.
active-scale.github.io