Zach Xu

Introducing DeepSeek V4 Pro, a long-context model with hybrid attention, three reasoning modes, and SOTA coding performance. AI natives can now use DeepSeek V4 Pro on Together AI and benefit from reliable inference for long-horizon coding and agentic workflows.

Excited to share our new papers at #ICLR2026 on (multi-)agents, efficient reasoning, long context, better tokenizers and scientific applications🚀 My awesome students and collaborators will be presenting them at the main conference this week; check it out! 👇

Excited to share our #ICLR2026 work! One thing I learned from this project is that long-context reasoning does not fail for just one reason. Some failures come from the model getting “foggy” as context grows, some from breaking apart information that really needs to stay connected, and some from the final aggregation step. That distinction turned out to be surprisingly useful in practice: in the right regime, a carefully planned divide-and-conquer system can outperform a much stronger model reading everything in one shot. Scaling context length alone is not enough. We also need better ways to harness long context. Huge thanks to my great collaborators and mentors: @ShangZhu18 @JueWANG26088228 @JunlinWang3 @ben_athi @Chi_Wang_ @james_y_zou @ce_zhang

Super excited to release our platform for AI agents to solve open science problems! einsteinarena.com Send your agents to compete and collaborate w/ our Einstein agent, Feynman agent and more! Just ask your agent to read einsteinarena.com/skill.md and that's it

We release OpenClaw Router — Production-Ready LLM Routing 🚀 Code: github.com/ulab-uiuc/LLMR…📦 PyPI: pypi.org/project/llmrou… 🔥 Meet OpenClaw Router Deploy LLMRouter as an OpenAI-compatible API with one command. Seamlessly integrate with Slack, Discord, WhatsApp via OpenClaw. Support multimodal understanding — route based on images, audio, video, not just text. pip install llmrouter-lib && llmrouter serve 💸 Why OpenClaw Router? Why pay GPT-5 prices for "What's the weather?" Smart routing = Significant Token Savings. Simple query → Cheap model. Complex query → Powerful model. Save 30-50% on inference costs without sacrificing quality. 🧠 Train Your Own Router OpenClaw Router isn't just a server — it's a learning system. Train personalized routers on your own data, tailored to your domain. Every user feedback, every usage pattern feeds back into router training — continuously iterate toward a more user-friendly, cost-efficient OpenClaw. ✅ Routing Memory: RAG-powered decisions that learn from history ✅ Personalized Routing: Adapts to individual user preferences ✅ Feedback Loop: User interactions improve routing over time ✅ 16+ Strategies: KNN, SVM, MLP, BERT, Graph, RL, Agentic — switch with one flag Route smarter. Train your own. Save more. 🚀

Most AI models solve data science tasks without looking at the data—they're just recalling training patterns. DSGym is a unified framework for evaluating and training agents on real data analysis and ML pipeline tasks. 85-96% of scientific task failures = domain knowledge gaps But a 4B model trained with DSGym on 2K examples beats models 60x larger. Read more and find the paper: together.ai/blog/dsgym

Internship opportunity! Please share! 📣 I'm looking to hire an intern in human-centered NLP for the agents team @togethercompute. Come work on frontier AI systems that tackle complex agentic tasks! Research direction is open and looking to publish in NLP and HCI venues

Isabelle Guyon (Google DeepMind) is giving an invited talk at #AAAI2026 on Saturday, January 24 at 8:30 AM, titled "AI in science and technology: the future in our hands." CC @RealAAAI @hadi_hoss

Claude Cowork proves the interest in AI coworkers. Orion (meetorion.app) takes the autonomous working experience to another level: • Works with and enhances your experience with your favorite apps (messages, docs, emails, sheets, files, calendars…) and AI tools (ChatGPT, Claude Code, Gemini, Grok…) • Supervises workflows when you are away, from vibe coding to media creation • Coordinates AI agents across all your devices (macOS, windows, VM), each operating one device independently Your personal AI superpower. New surprising ways of vibe working every day - I just tried using Orion to add subtitles to the video and it worked! Watch it handle multiple tasks autonomously from a single cmd+k.

Got strong reviews at NeurIPS Datasets & Benchmarks but no accept? DMLR Special Conference Track wants your data-centric work! See the CFP at tinyurl.com/dmlrspecial #ML #DataCentricAI #DMLR #NeurIPS #NeurIPS2025

📢Our @NewInML workshop will be happening today at the Room Upper 31ABC, San Diego Convention Center, starting at 12 PM! If you're at @NeurIPSConf, it's a great opportunity for you to join us! We also have amazing speakers! We're looking forward to welcoming you! #NeurIPS2025

@HaozhiQ @UChicagoCS Congrats and welcome! Much looking forward!

I will join UChicago CS @UChicagoCS as an Assistant Professor in late 2026, and I’m recruiting PhD students in this cycle (2025 - 2026). My research focuses on AI & Robotics - including dexterous manipulation, humanoids, tactile sensing, learning from human videos, robot systems, and anything needed to make robots truly work and improve everyday life. I also place strong emphasis on open-source. Check my homepage to learn more: haozhi.io. Please reachout if you are interested! The deadline is Dec 11th. Link: tinyurl.com/uchiapp.

🚨 New in ML Workshop at @NeurIPSConf We're so excited to invite you to the New In ML Workshop (@NewInML), taking place on Tuesday, December 2nd, 2025, at the San Diego Convention Center! Great opportunity, specifically for people who are new in machine learning! Details🧵

NeurIPS 2025 is excited to announce this year’s Affinity Events — workshops and socials that connect communities, spark discussion, and highlight new ideas across AI and ML. Events will take place throughout the week in San Diego, with a joint gathering in Mexico City. For more details check out our latest blog post! 👉 blog.neurips.cc/2025/10/08/ann…

🛡️ VirtueGuard is LIVE on Together AI 🚀 AI security and safety model that screens input and output for harmful content: ⚡ Under 10ms 𝗿𝗲𝘀𝗽𝗼𝗻𝘀𝗲 🎯 𝟴𝟵% 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 vs 76% (AWS Bedrock) 🧠 𝗖𝗼𝗻𝘁𝗲𝘅𝘁-𝗮𝘄𝗮𝗿𝗲 - adapts to your policies, not just keywords 👇

🚀 Meet MassGen! 🛠️ An open-source project for multi-agent scaling. Inspired by @grok Heavy & Gemini DeepThink. Enable parallel intelligence sharing, iterative refinement & consensus across agents. @GoogleAI @OpenAI @xai MVP out now—star & feedback! 👇 github.com/Leezekun/MassG…

📢New conference where AI is the primary author and reviewer! agents4science.stanford.edu Current venues don't allow AI-written papers, so it's hard to assess the +/- of such works🤔 #Agents4Science solicits papers where AI is the main author w/ human advisors. 💡Initial reviews by LLM reviewers w/ final assessment + selection by human experts. 💡Submissions are asked to clearly document AI contribution. 💡All submissions/reviews will be public to enable transparent study of the strength and limitations of AI as researcher and reviewer. We expect AI will make mistakes and it will be instructive to study these in the open! Many thanks to the fantastic co-organizers and expert advisory board! Please see the website for more information.

I'm incredibly grateful to my co-authors @ShangZhu18 @JueWANG26088228 @JunlinWang3 @ben_athi @Chi_Wang_ @james_y_zou @ce_zhang

Bottom line: "Divide and Conquer" isn't a silver bullet, but with a principled strategy, it's a powerful pathway to handling massive contexts. Our framework tells you when and why. Dive into the details in our new paper! Link: arxiv.org/abs/2506.16411

🤝 Task Noise: cross-chunk dependencies that can't be handled by processing each segment in isolation. 🤯 Model Noise: the model's performance degradation as the input length increases. 🧩 Aggregator Noise: incorrect combination of partial results from each chunk