1st-year PhD @Tsinghua_Uni advised by @stingning and Bowen Zhou Prev: Research intern @taoyds @haopeng_uiuc
Joined April 2024
- Tweets342
- Following1.3K
- Followers622
- Likes604
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. 🧵
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 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.
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!
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.
Junlin Yang retweeted
FrontierCS 2 is our new benchmark for CS research & development.
Each task is based on a paper: we delete the contribution from the released code, and ask agents to outdo the original method, as judged by the authors' chosen metrics. Agents don't have access to the paper or web.
🔥
✨ Introducing AutoRLBench — a benchmark built from real RL training failures.
AI training AI is no longer hypothetical. Agents already run post-training, read the result, and decide what to do next.
But a loop only turns if each round can be judged. When a run goes wrong, the agent has to find the root cause, and fix it — otherwise the loop runs on a broken signal.
AutoRLBench is the first benchmark that measures those abilities on real RL training failures. And we're recruiting more contributors 🧵
🌐 Project: thinking-space.github.io/Aut…
Junlin Yang retweeted
1/n ✨ In the early days of computing, people used computers by typing machine code ,just 0 and 1. Today, we use keyboards, mice, and GUIs. The convenience in between comes from the OS.
AI is now at a similar early stage: people write code to train models, run inference, and process data. Maybe in the near future, AI will be able to build AI recursively, so-called recursive self-improvement (RSI). At that point, what will the OS of RSI be?
As a first attempt, we built an OS layer called RSI Service. In this initial implementation, we focus on SFT data engineering. So we propose AutoSFT, which is an AI system that post-trains AI: a multi-agent research policy that iteratively designs the SFT data recipe, with a fixed base model, trainer, and budget.
📄 Paper: huggingface.co/papers/2610.2…
💻 Code: github.com/Thinking-Space/Au…
Junlin Yang retweeted
Full demos of six real-world tasks, all performed by Natura-1.
Everyday work involves more than picking and placing rigid objects.
Natura-1 demonstrates sustained autonomous execution across workflows that combine precise manipulation with multi-step coordination.
Junlin Yang retweeted
Introducing Natura-1, @naturewill's first general foundation model for the physical world.
It is a single model that follows plain-language instructions zero-shot and carries them out with dexterous hands.
- 50B MoE, pretrained on 1M hours of egocentric human experience.
- Drives 20-DoF dexterous hands from plain language, zero-shot, in a 100ms loop.
- Scaling human data from ~44K to 1M hours cut action MSE on held-out robot data by 67.7%.
- Built on in-house data and compute infra, it handles fine-grained long-horizon tasks and a ~30-minute zero-shot interaction in one continuous take.
The general model exists, and it can only get stronger.
naturewill.com/blog/natura-1…
Junlin Yang retweeted
30 minutes. One take. No cuts.
We put Natura-1’s zero-shot generalization to the test with complex spoken instructions and unfamiliar tasks and environments.
It responds to new requests, adjusts and retries when an action fails, and keeps going autonomously.
Junlin Yang retweeted
Some work seems to be what we’ve done before.
Building efficient infrastructure. Collecting a lot of data, then a lot more. Running ablation after ablation to understand what the model should actually look like. Debugging software and hardware, again and again. The unglamorous work of processing data. Training in parallel, at scale, on a lot of compute.
Yet somehow, they all seem so different.
This field hasn’t converged yet, and much of the infrastructure and many of the methods had to be figured out again.
Nothing was mature. Software and hardware were unstable in countless ways, and had to be made to work together. The architecture for a new kind of model had to be explored from scratch. There was no training recipe to refer to. And every bit of grunt work needed someone watching over it.
This time we get to watch the result work right in front of us.
