
1st-year PhD @naturewill @Tsinghua_Uni advised by @stingning and Bowen Zhou Prev: Research intern @taoyds @haopeng_uiuc
Joined April 2024
- Tweets349
- Following1.3K
- Followers632
- Likes627
Pinned Tweet
🚀 OpenRSI is a new open research series from @FrontisAI for concrete, testable progress toward recursive self-improvement (RSI).
As its first project—and also my first work as first author—I’m proud to present OpenMLE: an open full-stack AI4AI system for autoresearch, where evolutionary agents improve ML solutions through executable feedback.
OpenMLE has three components:
- OpenMLE-Gym: 5,758 executable tasks + evaluators
- OpenMLE-ERL: execution-grounded SFT + RL
- OpenMLE-Evo: experience-guided long-horizon search
🏆 The full system—our trained Frontis-MA1-35B model paired with OpenMLE-Evo-Max—reaches 71.21% Medal Average on MLE-Bench Lite: surpassing GPT-5.5 + Codex (68.18%) and just 1.52% from GPT-5.6 Sol + Codex and the 2.8T Kimi K3 + Claude Code (72.73%). Budget: 12 hours/task on one RTX 4090 capped at 12 GB VRAM.
🌍 On 10 held-out NatureBench Lite tasks, both components transfer:
• same framework, model swap: Match-SOTA 50% → 70%
• same base model, framework swap: Match-SOTA 20% → 50%
🔓 Paper, code, models, data, and analysis below. 🧵
Really cool work! I believe continual learning is undoubtedly the next big frontier for LLMs, and figuring out how to enhance this capability through training is at the very heart of it.
🤖 LLMs are already very capable, how do we make sure they can continually improve from their experience at deployment?
And how do we train agents to push the frontier of both capability and continual learning ability?
In our new technical post, we build an RL recipe for this: cmu-aire.github.io/RIME/
On Continual Learning Bench, our trained Qwen3.5-4B performs comparably to GPT-5.4 and Claude Opus 4.7, and gains more from its past experience than they do, or than Claude Code and Codex 📈
🧵👇
4
264
Junlin Yang retweeted
From understanding the physical world to acting in it. 🚀
A great first step toward scalable Robot-Use Agents powered by Qwen. More to come!
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
Excited to see this unified foundation for open models on robotic tasks! Try it out!
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
Open robot-use agents beat a published GPT-6 Astra baseline
uni-robot-agent is a unified harness for open-weight robot agents. Alibaba's Qwen3.8-Flash-Next scores 31.39 on RoboDojo, above the published Astra–RoboProbe 28.97, with no privileged inputs.
Junlin Yang retweeted
推荐清华和伯克利开源的 uni-robot-agent。他们用开源模型 Qwen3.8-Flash-Next 直接控制机器人,在 RoboDojo 上拿到 31.39 分,超过了 GPT-6 Astra 的公开成绩,真机任务上也表现不错,而且全程不依赖仿真器里的特权信息。
这个工作确实很有趣,以前总觉得机器人操控还得靠闭源大模型或者大量真机数据微调,现在开源模型加上一套合适的 harness 就能直接打到这个水平,说明这件事的门槛其实比想象中低,也让人更有信心去自己试着训练和部署。
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
So interesting! We are cooking too! Stay Tune.
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
Robot control has seemed reserved for frontier LLMs like Astra and Opus 5.5. Our results challenge that view: robot-use is a general capability emerged across many LLMs, including open-weight models like Qwen and DeepSeek, without RSI, distillation, fine-tuning, or robot-specific data.
These open-weight models are already competitive when deployed out of the box: zero-shot Qwen3.8-Flash-Next scores 31.39 on RoboDojo, surpassing GPT-6 Astra’s result from official RoboDojo team. These capabilities also extend to real-world, where Qwen scores 53.50/100 across 19 real-world tasks, outperforming WAMs/VLAs such as Dream-1 and fine-tuned π-0.5.
Why LLMs can work as a robot policy? They approach much like solving a math problem: first understand the scene, then triangulate feature points across multiple views to locate them in 3D, and finally compute grasp poses using coordinate transforms and inverse kinematics. This process draws on four general capabilities: visual understanding, spatial grounding, coding in geometry, and embodied reasoning. Robot performance appears to emerge once these capabilities are sufficiently strong.
This computation-heavy approach can yield sub-millimeter precision when feature points are correctly identified. This suggests that improvements in Robot-Use Agents should instead focus on high-level intents and visual grounding, including strategy, object affordances, 2D feature understanding and pinpointing, rather than on low-level actions.
With open-weight models, we can now train, test, and improve these capabilities directly.
Blog: openrua.ai/uni-robot-agent
Code: github.com/Open-RUA/uni-robo…
Paper: arxiv.org/abs/2611.01560
Junlin Yang retweeted
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Jiarui is an absolute wizard🧙♂️ at harness engineering.
We built uni-robot-agent to make robot-use research accessible with open models. Beyond stronger zero-shot performance, we see a path toward training visual understanding, spatial grounding, geometric coding, and embodied reasoning together through robot interaction.
4
233
👀
📺 We ran uni-robot-agent on a real Franka arm fitted with a Sharpa hand. Watch the index finger 👇
We asked the robot to pick up a toy bird without using it, then press the green button with that same finger.
📸 We also ran Qwen locally on 19 dual-arm tasks, including stacking Hanoi disks and moving test tubes between racks.
3
146
Junlin Yang retweeted
With uni-robot-agent, open-weight Qwen shows robot-use capabilities comparable to the published Astra–RoboProbe results on RoboDojo. The unified harness lays the groundwork for generalizable agentic training.
📄 modelscope.ai/papers/2611.01…
🏆 Stronger robot use: Qwen3.8-Flash-Next achieves 31.39 Score / 25.00% success without privileged simulator inputs, versus 28.97 / 22.48% for the published GPT-6 Astra–RoboProbe configuration.
🤖 Physical-robot validation: Across 19 tasks, zero-shot Qwen scores 49.50/100, outscoring both π₀.₅ (47.75, fine-tuned on 10 hours of demonstrations per task) and Dream-1 (35.50, evaluated zero-shot). The agent uses depth; both baselines use RGB only.
🛠️ Unified for generalizable agentic training: Shared motion, perception, skill, and episode interfaces make trajectories across environments and robot embodiments easier to combine for future training.
🧠 What to train next: Analysis identifies visual understanding, spatial grounding, geometric coding, and embodied reasoning as joint priorities for future training.
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
Qwen plays robots🔥!
Check out @zuo_yuxin @x_h_ucb ‘s brainchild, an open and highly-performant harness for robot-use agents! Glad to have played a small part in the project😤
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
We built uni-robot-agent to make robot-use research accessible with open models. Beyond stronger zero-shot performance, we see a path toward training visual understanding, spatial grounding, geometric coding, and embodied reasoning together through robot interaction.
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Excited to have worked on uni-robot-agent with this team!! 👊 Seeing it enable Qwen to perform well both in simulation and on real robots has been especially rewarding.
This also kicks off OpenRUA (openrua.ai), a research series we’ve been building to lay the foundations for agents that can act in the physical world.
The best is yet to come. 💫
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
On a more personal note, I also built our website, openrua.ai, with Opus 5.5. It was my first time working with it on a website, and seeing my visual ideas come to life was a lot of fun. Would love to hear what you think!
6
127
Junlin Yang retweeted
Well, turns out Qwen3.8-Flash-Next can do more than work with software tools, it can operate real robots too.
A group of Qwen developers from Tsinghua and UC Berkeley recently built uni-robot-agent, an open-source harness that connects general-purpose LLMs to robotic systems. It gives Qwen a way to interact with the physical world through tool calls without changing the model weights.
With this harness, Qwen3.8-Flash-Next reached the performance level of some previously reported closed-model robotic agent systems. The average API cost was $1.09 per task, compared with $21.29 for GPT-6 Astra. On 19 real-world robot manipulation tasks. Without any robot-specific fine-tuning, Qwen achieved a slightly higher score than the fine-tuned π₀.₅
The harness is open-source if anyone wants to tinker with it.
‼️ 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 RoboDojo, its score rose from 1.90 to 31.39, surpassing the RoboDojo Team’s score with GPT-6 Astra.
2️⃣ On real-world tasks, it outscored fine-tuned π0.5 by 5.75 points and zero-shot Dream-1 by 10 points.
🧵 [1/7]
Junlin Yang retweeted
General models with dexterous manipulation is coming into view like @naturewill 's Natura-1. We need benchmarks ready to measure progress as these capabilities develop.
We introduce HandsOn, a simulation platform and benchmark for generalizable dexterous manipulation.
It brings together 66 human-verified tasks and 18,000 curated human demonstrations, covering tool use, precise interaction, articulated objects, and bimanual coordination.
We evaluate diffusion policies, vision-language-action models, world-action models, and LLMs to study how they handle these interactions.
Alongside success rates, we examine where execution breaks down: establishing the right contacts, coordinating actions, and maintaining achieved goals.
The paper, code, and dataset are available below.


























