🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

2/ Why does this matter? MoAA makes high-quality AI more accessible & scalable—no need for massive proprietary models! 📝Check out our blog post: together.ai/blog/moaa 🔗 Full details in our paper: arxiv.org/abs/2505.03059 By Junlin Wang Roy Xie Shang Zhu Jue WANG

Our new #icml2025 paper w/Together AI shows how to use synthetic data from Mixture-of-Agents to boost LM fine-tuning + RL. Turns out a mixture of small agents is much more effective/cheaper than using a large LM as teacher 🌐together.ai/blog/moaa 📜arxiv.org/abs/2505.03059

Please check out our freshly accepted #ICML2025 paper on LLM post-training research using Mixture-of-Agents. We released our SFT data and fine-tuned models for the community to try out!

1/ We built an open-source AI agent that can reason like a data scientist. It loads data, writes Python code, retrains when models fail, and solves real Kaggle + DABStep tasks. Here’s how we did it (and how you can too): 👇

We built a data science agent from scratch and made it accessible to everyone (pip install & one command-line call). Open-source codebase: github.com/togethercomput… We look forward to your feedback! With Federico Bianchi Zain Ben Athiwaratkun James Zou

Check out our fully open source data science agent! It performs complex tasks like hyper-parameter optimization, runs experiments and can even complete kaggle competitions!

Excited to introduce Open Data Scientist: ✅outperforms Gemini data science agent ✅solves real Kaggle tasks ✅fully open source, easy to adapt ✅sandbox for safe exec Step-by-step tutorial on building our agent together.ai/blog/building-… Great job Federico Bianchi Shang Zhu

Shrinking the Generation-Verification Gap with Weak Verifiers "we introduce Weaver, a framework for designing a strong verifier by combining multiple weak, imperfect verifiers." "Weaver leverages weak supervision to estimate each verifier’s accuracy and combines their outputs

LLMs are getting more powerful, but they still struggle with super long documents. A common trick is "Divide and Conquer" - chop it up, process chunks, and combine. But... when does this actually work? And when does it fail catastrophically? We investigated. 🧵

How can we close the generation-verification gap when LLMs produce correct answers but fail to select them? 🧵 Introducing Weaver: a framework that combines multiple weak verifiers (reward models + LM judges) to achieve o3-mini-level accuracy with much cheaper non-reasoning

📝 Paper: arxiv.org/abs/2506.18203 ✍️ Blog: hazyresearch.stanford.edu/blog/2025-06-1… 🔧 Code: github.com/HazyResearch/s… 🤗 Datasets and Models: huggingface.co/collections/ha… Joint work with Kelly Buchanan, Mayee Chen, Tzu-Heng Huang, Brendan McLaughlin, Tanvir Bhathal, Shang Zhu, Ben Athiwaratkun, Fred Sala,

LLMs often generate correct answers but struggle to select them. Weaver tackles this by combining many weak verifiers (reward models, LM judges) into a stronger signal using statistical tools from Weak Supervision—matching o3-mini-level accuracy with much cheaper models! 📊

Introducing the Open Deep Research app! Generate detailed reports on any topic with open source LLMs. Free & fully open source. We’re releasing everything: evaluation dataset, code, app, and blog.🔥

Thrilled to see our deep research efforts turning to an app! Built by Hassan !

Announcing Open Deep Research! Generate detailed reports on any topic with open source LLMs. 100% free and open source. opendeepresearch​.dev

Open Deep Research app + fully open recipe ☺️

Introducing Weaver, a test time scaling method for verification! Weaver shrinks the generation-verification gap through a low-overhead weak-to-strong optimization of a mixture of verifiers (e.g., LM judges and reward models). The Weavered mixture can be distilled into a tiny