
Asst. Prof.@HKUST, Postdoc.@ETH, Ph.D.@RiceUniversity, AReaL (https://teapaw.org/t.co/aBhA1AUC8u) ML Systems, fan of @LFC.
Hong Kong
Joined February 2014
- Tweets65
- Following418
- Followers241
- Likes147
Pinned Tweet
Excited to release AReaL 2.0🥳This is our biggest update so far, and the first big milestone after AReaL joined the @PyTorch Ecosystem.
Check out our work: github.com/areal-project/ARe…
Vision report for self-evolving RL system substrate: github.com/areal-project/ARe…
🙌Welcome to the AReaL community. First tweet.💥Big release.
We are excited to announce the release of AReaL 2.0: RL as Micro-Service! Thanks to the great efforts from the AReaL community and collaborators❤️
In AReaL 2.0, we rebuilt online agent RL around four decoupled components: Training Service, Inference Serivce, WeightUpdate Service and Agent Service.
Change one URL in your existing agent, start learning from production traffic.
#AIAgent #agenticRL #AReaL #LLM
💻 GitHub: github.com/areal-project/ARe…
🙌 Also huge thanks to our community collaborators. @Macaron0fficial is collaborating with us on LoRA RL and AReaL-Mint based APIs: github.com/areal-project/ARe…
AReaL 2.0 is an open invitation: let's build self-evolving agent infra together!
40
We put a lot of effort into making this release an exciting milestone:
We enable micro-service based agent training/inference worker, session routing, data proxy, online weight sync, CLI management.
We provide cool examples about Hermes Agent online RL and Claude Code Agent RL.
1
50
Binhang Yuan-HKUST-AReaL retweeted
🙌Welcome to the AReaL community. First tweet.💥Big release.
We are excited to announce the release of AReaL 2.0: RL as Micro-Service! Thanks to the great efforts from the AReaL community and collaborators❤️
In AReaL 2.0, we rebuilt online agent RL around four decoupled components: Training Service, Inference Serivce, WeightUpdate Service and Agent Service.
Change one URL in your existing agent, start learning from production traffic.
#AIAgent #agenticRL #AReaL #LLM
💻 GitHub: github.com/areal-project/ARe…
Binhang Yuan-HKUST-AReaL retweeted
Our research team has 9 papers at ICML next week!
Spanning the full stack from frontier agents to GPU kernels, we're excited to share what our researchers and collaborators have been working on.
If you're in Seoul for ICML come meet our team or catch a talk! 🧵
Binhang Yuan-HKUST-AReaL retweeted
🌀 Introducing Vortex — sparse attention designed by AI agents, efficient at scale.
📈 Same accuracy, way more throughput — across every model we tried 👇
🔹 GLM-4.7-Flash (MLA) → 4.7× faster
🔹 MiniMax-M2.7 (229B) → 1.37× faster
🔹 Qwen3-1.7B (agent-discovered!) → 3.46× faster
🤖 How? An agent writes a flow in a few lines of Python; Vortex compiles it into fused kernels in a real serving stack (SGLang) and benchmarks it end-to-end.
🏗️ The design: a Python frontend (vFlow) over a page-centric tensor abstraction (vTensor) + a serving-integrated backend.
📄 arxiv.org/abs/2606.06453
💻 github.com/Infini-AI-Lab/vor…
🌐 infini-ai-lab.github.io/vort…
📚 infini-ai-lab.github.io/vort…
AReaL v1.0 released: Effortless #RL to make your #OpenClaw self-evolve 🚀:
•🛠️ One-click agentic RL for any existing agent
•📈 Open-source SOTA on tau2-bench
•💎 A new PyTorch-native 5D-Parallel Engine Archon
•🤖A full #opencode recipe
GitHub: github.com/inclusionAI/AReaL
Binhang Yuan-HKUST-AReaL retweeted
📘 Holiday read! From Software Engineer to AI Environment Architect
🚀 Tldr of our blog: We see an exciting future where engineers 👩💻 won’t stop coding — but the highest leverage shifts to designing the environments 🛝 where AI can think, build, and evolve.
🎬 Demo: Inspired by opinions from @karpathy @RichardSSutton, our newly built framework Vortex shows this in a concrete action: by architecting the right environment in LLM serving systems, an agent from @OpenHandsDev can generate and implement new Sparse Attention algorithms on @sgl_project in a single run and deliver up to 4× ⏩ gains — work that normally takes an ML-systems engineer weeks.
- In the short term, these environments let AI agents contribute meaningfully to real engineering work today.
- In the long term, they become the playgrounds where future agents learn to surpass today’s limitations 💖.
Get too excited by the demo, write a blogpost before the holidays with my great students @chenzhuoming911 @IronSteveZhou who built it:
infini-ai-lab.github.io/ai-e…
Vortex code: github.com/Infini-AI-Lab/vor…
Vortex doc: infini-ai-lab.github.io/vort…
#MLOps #AIAgents #SystemsEngineering #AIInfrastructure #OpenSourceAI #AIOps #AIFrameworks #SparseAttention #AIResearch
Binhang Yuan-HKUST-AReaL retweeted
Sorry for a typo in my previous post. The correct arxiv paper link is: arxiv.org/pdf/2508.07976v1
🔍We introduce ASearcher, a search agent trained by end2end RL
Large-scale (up to 128 turns) RL with AReaL unlocks Long-Horizon Agentic Search (+20.8/+46.7% on GAIA/xBench)
💻Data, Code&Model: github.com/inclusionAI/ASear…
📄Paper: arxiv.org/abs/2508.07976v
#Agent #OpenSource #LLM #AGI
Binhang Yuan-HKUST-AReaL retweeted
Tired intricate system code for RL training? 🤯
We release AReaL-lite – A lightweight AReaL version for AI researchers! 🚀#opensource
✨ Algorithm-first design & APIs🎉
✨ 80% less code w. 90% AReaL's full efficiency 🎉
✨ Customizable agentic RL🎉
🔗 github.com/inclusionAI/AReaL…
Binhang Yuan-HKUST-AReaL retweeted
🥳
Huge thanks to @tinytitans_icml for an amazing workshop — see you next year!
Honored to receive a Best Paper Award 🏆
Let’s unlock the potential of sparsity!
Next up: scaling to hundreds/thousands of rollouts?
Or making powerful R1/K2-level LLMs (not just 8B 4-bit models) run on edge devices?
Big kudos to @RJ_Sadhukhan, @chenzhuoming911, @haizhong_zheng, @IronSteveZhou, collaborator Emma Strubell, and our advisor @BeidiChen!
Binhang Yuan-HKUST-AReaL retweeted
🔥 We introduce Multiverse, a new generative modeling framework for adaptive and lossless parallel generation.
🚀 Multiverse is the first open-source non-AR model to achieve AIME24 and AIME25 scores of 54% and 46%
🌐 Website: multiverse4fm.github.io/
🧵 1/n
Binhang Yuan-HKUST-AReaL retweeted
We release fully async RL system AReaL-boba² for LLM & SOTA code RL w. Qwen3-14B! @Alibaba_Qwen #opensource
🚀system&algorithm co-design → 2.77x faster
✅ 69.1 on LiveCodeBench
🔥 multi-turn RL ready
🔗 Project: github.com/inclusionAI/AReaL…
📄 Paper: arxiv.org/pdf/2505.24298
1/3👇
Binhang Yuan-HKUST-AReaL retweeted
Another angle: What I always encourage people is to believe there is no barrier on whatever you want to work on. Whatever training or inference, nothing is really hard if you spent time and focus. They are all paper tigers. The secret is to just ignore people who told you it’s too late/hard and only listen to people who’d like to help you achieve your goal.
Future ML specialization: Inference or Training?
Very soon training LLMs will become a domain of a few companies and there will be very little need in experts in LLM training. Especially when LLMs will be at the level of CV cats-vs-dogs quality.
Inference expertise on the other hand has a tiny barrier to entry, so it'll become commoditized in no time. It'd be very difficult to compete with others and differentiate one self doing inference work.
I suppose finetuning/RAG is going to be somewhat needing experts for some years to come - until foundational models will have all that provided out of the box. And then what?
Where do you feel one still could be an ML expert 5-10 years from now?
I worked though the dot com bubble and this ML bubble feels too too similar. Hence the asking.
Thank you for your insights.
Binhang Yuan-HKUST-AReaL retweeted
We are recruiting! Applications including 1) a cover letter, 2) a full curriculum vitae, 3) names and contact information of at least three referees, 4) a research statement, and 5) a teaching statement should be submitted via facrecruit.hkust.edu.hk/.
Binhang Yuan-HKUST-AReaL retweeted
1/ 🌟 Excited to announce #Model-#GLUE (#neurips2024 D&B), a new framework designed by an extensive team from UNC, UMD, UT Austin, HKUST, Google, and CMU to #scale pre-trained LLMs efficiently!
🚀 Tackling the challenge of #aggregating disparate pre-trained LLM, we introduce a holistic guideline and benchmarking if you have a large, diverse model zoo "in the wild"! #LLM #AIresearch
Binhang Yuan-HKUST-AReaL retweeted
🚀 Big news! We’re thrilled to announce the launch of Llama 3.2 Vision Models & Llama Stack on Together AI.
🎉 Free access to Llama 3.2 Vision Model for developers to build and innovate with open source AI. api.together.ai/playground/c…
➡️ Learn more in the blog together.ai/blog/llama-3-2-v…
Binhang Yuan-HKUST-AReaL retweeted
🚀 NVIDIA H200 and the Together Kernel Collection (TKC) are coming to Together GPU Clusters: delivering accelerated performance, efficiency, and scalability for AI training, fine-tuning, and inference workloads. ⚡
🔗 Read the blog post together.ai/blog/nvidia-h200…
Binhang Yuan-HKUST-AReaL retweeted
#MLSys2025 call for papers is out! The conference will be led by the general chair @matei_zaharia , PC chairs @CelineLinatGT, and Gauri Joshi. Consider submitting and bringing your latest works in AI and systems—more details at mlsys.org.

























