Runhan Huang

@RunhanH

PhD student @Harvard | Prev. @Tsinghua_IIIS (Yao Class) | Robot Learning, Generative AI

Cambridge, MA

hrh6666.github.io

Joined September 2024

757 Following

384 Followers

Replies Reposts Media

Pinned
Runhan Huang
@RunhanH
Oct 16
Honored to co-lead RAX with Xiangcheng (@xczhangxx)! Excited to share what robots can learn through scaling exploration—no demos, no model finetuning! We’re open-sourcing the full framework and every experiment session. Please fell free to try out! Project + sessions:
Yilun Du
@du_yilun
Oct 16
Robotics has spent years scaling demonstration data and model size. What happens when we scale EXPLORATION for robot agents? We're incredibly excited to share RAX, an agentic robotics system that learns by scaling online exploration. A coding agent connects directly to the
5
Runhan Huang
@RunhanH
6h
Scaling up exploration is unlocking Astra’s dexterity. With RAX, Astra turns experience into reusable code and chains six skills to open a jar. Congrats to @jerryhuang01 @RoboticsCtr on bringing RAX to real dexterous hands! #EmbodiedAI #Deepseek #Claude #OpenAI
Jerry Huang
@jerryhuang01
12h
0% success on L3 tasks in a recent Astra benchmark, including bottle-cap unscrewing. Now RAX tackles jar opening with real dexterous hands. 6 chained skills. 4 twists. 140° rotation. The jar opens. @RunhanH @xczhangxx @du_yilun #RAX #Robotics
Runhan Huang
@RunhanH
6h
Scaling up exploration is unlocking Astra’s true potential in robotics. Now Astra with RAX is tackling human-level dexterous manipulation! Astra struggled to unscrew a bottle cap. With RAX, it explores on the robot, turns experience into reusable code, and chains six skills to
Robotics Center of Silicon Valley
@RoboticsCtr
12h
0% success on L3 tasks in a recent Astra benchmark, including bottle-cap unscrewing. Now RAX tackles jar opening with real dexterous hands. 6 chained skills. 4 twists. 140° rotation. The jar opens. @RunhanH @xczhangxx @du_yilun #RAX #Robotics
Runhan Huang
@RunhanH
Nov 4
Huge congrats to the OpenRUA team! Exciting to see open-source LLMs powering robot control 🤖 We’re also pushing in the same direction with RAX, using the fully open-source DeepSeek-V4.1-Flash as our base LLM—with support for self-hosted inference. RAX harness +
Yuxin Zuo
@zuo_yuxin
Nov 3
‼️ Can Qwen directly control the robot as well as GPT-6 Astra? 🚀 Introducing uni-robot-agent: a unified harness for Robot-Use Agents (RUAs), a first step toward RUA training and local deployment. We tested uni-robot-agent with Qwen3.8-Flash-Next running locally: 1️⃣ On
2
Runhan Huang
@RunhanH
Oct 27
🔥Great to see @RoboticsCtr @jerryhuang01 run RAX on a real robot and get such strong results! Really exciting to see others run RAX agent on their own robots and learn reusable skills through scaling exploration—no demonstrations, no teleoperation! We’ve open-sourced the
Robotics Center of Silicon Valley
@RoboticsCtr
Oct 27
An agent explores on a real robot for 2 hours, then runs the skills it learned as code. That’s RAX. No demonstrations. No teleoperation. Just exploration turned into reusable robot skills. @RunhanH @xczhangxx @du_yilun @pabbeel @wangyenjen #Robotics #EmbodiedAI

Runhan Huang

Runhan Huang’s posts