Excited to start our LLM interpretability journey with COSMIC 🔮! If you're at ACL, make sure to go see our poster to chat with Vincent about this work and our future plans in the space (of which we have many)!

🚨 Excited to share our latest work at KnowFM 2025: MLAN: Language-Based Instruction Tuning Preserves and Transfers Knowledge in Multimodal Language Models! 🌟 #ACL2025 #KnowFM

I get it, I can't see files this big right now. But how many hours do I need to wait for GitHub to go to the then where I can see it and make it now? When will then be now?

Happy to announce that I have moved with Chenguang Wang (hiring) to UCSC to continue my PhD! Excited for all the new opportunities and to keep focusing on making impactful research together.

bit.ly/4nLwNaB Excited to share our recent work on #AgenticAI at #SFTechWeek, co-hosted by Baskin Engineering at UCSC and Silicon Valley AI Pioneer Club ! 🤖✨ 💡MassGen MassGen : Multi-agent AI for complex tasks → github.com/Leezekun/MassG… 💡 rLLM Agentica Project : RL for LLM agents →

Excited to share that my lab has moved to UC Santa Cruz Baskin Engineering at UCSC! 🌊 We’re recruiting the next generation of pioneers in #AgenticAI and #AISafety, with openings for postdocs and PhDs. Join us!👉cgraywang.github.io 🔬 Lab: cgraywang.github.io/people/ 💻 GitHub:

How might making an LLM safer in one aspect make it behave differently in others? To find out, we built SteeringSafety, a benchmark for representation steering methods spanning 7 safety perspectives. Crucially, we test both effectiveness (how well steering works on a target

Andrej just dropped LLM Council — super fun vibe, love it ❤️ If you want the leveled-up version with parallel agent orchestration, live token sharing, iterative refinement, and agent-to-agent/human broadcasting, join a community that's been pushing hard for 5 months (50+ releases

I am attending #NeurIPS2025 as well! Please reach out if you want to chat more about our recent papers or opportunities in the lab.

Excited to share our latest work, now on arXiv and at FoRLM @ NeurIPS'25! 🎉 Introducing **LLM Chess**: a benchmark for evaluating reasoning and instruction-following in LLMs through chess. LLMs now reach experts in math & coding, but can they *reason* in dynamic, multi-step

Excited to present our new Web UI for MassGen! If you found LLM Council interesting, but wanted a more sophisticated multi-agent system with new features constantly added, please check us out! Easy to start and with an active Discord community. What is MassGen? An open-source

🚀So excited to have just received a research gift from Google to support our work on AI agents! Huge thanks to Google. 🙌Come join us, let's build the future of agents together!

Claude Code Agent Teams vs MassGen Claude Code Agent Teams: code.claude.com/docs/en/agent-… - Task division across teammates - Built for engineering workflows - Great for parallel development Why check out MassGen: massgen.ai - Open source - Any model: Claude, Gemini, GPT,

We're excited to release MassGen v0.1.48, adding Decomposition Mode! 🚀 A new coordination mode that decomposes tasks into subtasks assigned to individual agents, with a presenter agent synthesizing the final result. Plus: Worktree isolation for safe file writes with review