Twitxy Saved

Robin Wang

Verified account This account is verified. Learn more

@WangRobin58334

PhD student @princeton in Robotics,
prev @mondorobotics @umich
Princeton, NJ weiqianwang123.github.io Joined December 2024
357 Following 90 Followers 81 Posts

Latest posts 5

Robin Wang Verified account @WangRobin58334
严肃学习
Xuan (Billy) Zhang Verified account @xzhang_billy 3d
总结下简中区与PhD相关理论:
1. 大 Bench 理论:x.com/vclee8/status/…
2. 大实习理论:
x.com/vclee8/status/…
3. 大粉理论:
x.com/collinzrj2001/…
4. 大日结理论:x.com/persdre/status…
5. 大斜率理论:x.com/Selen700571791…
6. HR 理论:x.com/xzhang_billy/s…
颇有百家争鸣的氛围
View quoted post on X
Robin Wang Verified account @WangRobin58334
So cool
Octi Zhang Verified account @octi_zhang Oct 9
(1/5) Everyone says dexterous manipulation needs real-world data. These policies were trained with none. Our robots mesh gears, thread nuts, and insert pegs zero-shot from RGB, and the same recipe works for locomotion across terrains much harder than before.

The missing piece wasn't a better RL algorithm, but a small tweak to what the robot practices.

Today we're releasing Success-Guided Sampling (SGS), a simple change to the task sampler that unlocks new levels of dexterity and agility with large-scale, end-to-end sim2real RL.

No demonstrations, no per-task reward tuning. Just PPO at scale.

Check out the policies yourself at sgs-rl.github.io! All videos at 1x speed! 🧵 🔈
View quoted post on X