🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Thrilled to be in Atlanta for #ICRA2025—as a member of the organizing committee for the first time! Join us Tuesday for two IEEE RAS IEEE RAS Student Activities Committee events: Lunch with Leaders (12:15–1:45pm) & Student Social Hour (7–9pm). Always happy to chat research or life—come say hi or DM me!

In robotics, task success/failure has long been the default metric. But is binary success enough to capture a policy’s true capability? 🤖 No, as robotics models grow as capable as vision/language models, we need deeper diagnostics. Introducing RoboEval, A new paradigm of

RL for robotics has struggled to take advantage of compute scale for hyper param sweeps because we don’t have stable reward-independent success metrics that can be maximized. this work could help unlock the ability to do real automated parameter search, which is super exciting!

One of the most impotent indications to the converges coming into robotics is more and more benchmarking coming in hot They allow the community to better understand how and where to invest - a super important director to a high resource consumption field

Having reliable signals to compare different robot policies is sooo important. The more informative the metrics, the better we can compare, improve, and build smarter robot policies. Love to see work in this direction.

How do we accelerate progress in robotics? As Jitendra MALIK has noted, computer vision leapt ahead once robust, openly-shared benchmarks appeared. Robotics still lacks that foundation. Real-world benchmarking is logistically expensive and hard to standardize; simulation is

I love the idea of 'Behavior Metric'! Like we did in #RoboArena, we are trying to record more fine-grained, general and useful task descriptions Robotics has suffered enough for simple, binary score, we should involve metric like progress score and this, to further support the

In this generation of embodied intelligence, success isn't just about whether the task is completed. We care how it's done: – How fast – How well – How robust And when it fails—why it fails, and what happens next. These design questions are the key to bringing end-to-end models