Jianyang is a fantastic collaborator. He led BioCLIP 2 to a huge success, and is also a great mentor to younger students and larger teams (FinerCAM, SST). 10/10 recommend working with/hiring Jianyang!

I’m also hiring researchers and engineers at #NeurIPS for NeoCognition. If you find true passion in agent plasticity and reliability or an alternative path to general intelligence, happy to chat! hiring@neocognition.io

So nice to see everyone! Wish whoever on the job market or taking new adventures BEST of LUCK!

@ysu_nlp is one of the most visionary people IMO. He was thinking about computer-use agents, world models, and continual learning 2-3 years ago, long before they became mainstream. I have always trusted his vision, and this time is no exception. Working at NeoCognition is one of

The GDM mechanistic interpretability team has pivoted to a new approach: pragmatic interpretability Our post details how we now do research, why now is the time to pivot, why we expect this way to have more impact and why we think other interp researchers should follow suit

@DakingRai is one of my best. Highly recommend for any team recruiting for LLM Interp/reasoning/controllability/etc. He plans to graduate in a year or so. Go meet him at the poster session or msg him for a chat directly! Exhibit Hall C,D,E #3815 Thu 4 Dec 4:30 p.m. PST — 7:30

Paper rejected from #NeurIPS on space constraints with 3/4 accepts. Resubmit to #ICLR: 2 rejects, 1 accept. Work hard, get miracle: responsive reviewers who all raise their scores after discussion. Then all scores revert due to a bug. What a ride @iclr_conf!"

I think that the ICLR response to leaking of identites is a bit nonsense.. as reviewer I was in the middle of a discussion with the authors on a couple of papers, and I really wanted to ask some final questions to finish understanding how to change the score

I respect that @iclr_conf had to respond to the OR leak, but I disagree with resetting scores. Many students worked hard on rebuttals and improved their papers in good faith. I hope the organizers reconsider and revert the reset. If you agree, feel free to retweet.

I was an ICLR PC two years ago. Woke up every morning to hundreds of unread conference-related emails. It’s simply not sustainable. And let’s not blame the committee—they’re already working nonstop putting out fires. We need structural fixes: smarter tooling, sane reviewer

Pros: Claude 4.5 Opus sets a new SoTA on OSWorld for computer-use capabilities. Cons: It also achieves the highest ASR (83%) on our new RedTeamCUA evaluation. Capability is rising fast, but trustworthiness depends on security keeping pace.

It's time to shift from LLM alignment to CUA alignment! I will be attending NeurIPS next week. Happy to chat more on LLM/Agent safety/security and build truly deployable and reliable agents together! I am also exploring better on-policy distillation. If you're doing similar

RedTeamCUA: https://t.co/Kb5DvRZgco we will update our draft with the new results soon! Demos to see what a successful attack looks like:

Computer-use agents are getting more capable, but are they getting more secure? No! We are actually observing the opposite trend. @AnthropicAI @claudeai Opus 4.5 released two days ago tops the OSWorld leaderboard, but is showing the highest Attack Success Rate (ASR) on our

@yoavgo Thanks Yoav! I’m not leaving academia and will certainly continue to publish. Meanwhile, agents have gotten to a point where learning from deployment is necessary. Besides business, I hope this will also be a new kind of research platform that shares back with the community.

Life update: I moved to silicon valley to tackle agents' biggest challenges: plasticity and reliability. Today's agents are smart but brittle. They lack plasticity (continual learning and adaptation) and reliability (stable, predictable behavior with bounded failures). These two

Come help Computer Science pay out the tech debt vibe coding is creating! @ohiostate CSE is hiring THREE faculty positions in Programming Languages and Software Engineering, including an endowed chair. https://t.co/sCYoRQMoNA

We are hiring @osunlp! - frontier agent/LLM research - hundreds of H100s/A100s + clouds - uncapped LLM APIs Join us if you have bold ideas about the future of AI

🔥🔥Ph.D. positions available at the OSU NLP group @osunlp (led by Yu Su @ysu_nlp, Sachin Kumar, @shocheen and myself). Collectively, we're interested in agents (everything to advance their capability and safety/security), multilingual models, continual learning, personalization,

Big congratulations to the Yutori team!!!Thanks for choosing our Online-Mind2Web for evaluation.