Our paper, "What's in My Human Feedback", was selected for an oral at ICLR! Our method automatically interpretably identifies preferences in human feedback data; we use this to improve personalization safety. Please reach out if you have data/use cases to apply this to!

great to see more work from @GoodfireAI on interp for data curation! our ICLR'26 paper, What's In My Human Feedback, also trains SAEs on chosen vs. rejected responses to forecast behaviors from pref data. the GF paper provides more evidence this can improve post-training.

When people strongly disagree on an issue, can they agree on what makes a good AI response? We find: yes, more than you might expect! We present PARETO, a large human study w >200k evals, measuring the Pareto frontier of approval btwn opposing groups on controversial issues 🧵

User simulators have emerged as promising tools for building interactive AI, but what makes a “good” simulator? We reframe the problem as what creates downstream value for humans Our new simulator test: how an LLM assistant trained with the simulator performs with human users🧵

I'm joining Carnegie Mellon's CS Department (and HCII by courtesy) as an assistant professor in Fall 2027! I'll be recruiting PhD students next cycle. If you're interested in AI systems or human-AI collaboration, list me in your application. Stay tuned for more about my new lab!

Congratulations to Vatsal, who did terrific work in our lab last summer!

🎉 Thrilled to have two papers accepted to ACL 2026 main! 1. Graph-based models match LLMs on close-ended human simulation tasks with far less compute & greater transparency 2. (oral) How to allocate human samples towards fine-tuning vs post-hoc rectification in simulation

New paper: What Do LLMs Know About Opinions? If we want LLMs to reflect diverse human views or simulate human responses well, we need to understand what they know about human opinions. Current evaluations mostly rely on next-token probs, but what if that misses a lot of what the model actually knows? 💡 In our ICLR 2026 paper, we find that models know much more about human opinions than their outputs reveal.

Our lab, within the Berkeley EECS department, is hiring a postdoc! More info and quick application form: forms.gle/41tTVesNqtz33R838 Apply by May 1! Please reshare :)

BREAKING: Pope Leo XIV on Trump’s warning to Iran of “civilization” destruction — “This is truly not acceptable. Here there are certainly questions of international law, but even more than this a question of morality for the good of people.” He adds the war is “continuing to escalate and is not resolving anything… is only provoking more hatred throughout the world.” “attacks on civilian infrastructure are against international law, but it is also against sign of the hatred and division that we are capable of.” Video @Reuters

New paper: "In Your Own Words"! We created a framework to identify themes in free-text survey data and showed its benefits on a new dataset of how people describe their own identities (available for research!) See @jennyshwang's thread below.

This would be a good time for: a.) STRATCOM commanders who implement orders to fire nuclear missiles to read up on Nuremberg and prosecutions for obeying orders to commit war crimes; and b.) for cabinet members to reread the 25th Amendment and have each other on speed dial.

This is superb. Among other things, I think a great read for anyone starting a PhD or other graduate study with an interest in AI and philosophy. I particularly agree with this quote (unsurprisingly). I'm hopeful that our new School of Government and Policy in DC will train a good number of top quality practitioners in the kind of philosophy and politics necessary to make powerful AI go well.

We have a new piece in Nature Health led by @dmshanmugam, @sidhikab1, and a wonderful team of coauthors on how to move towards a world in which race is not used in clinical algorithms!

Congratulations to @gsagostini, whose recent Nature Comms paper releasing a fine-grained migration dataset (nature.com/articles/s41467-0…) just won a student paper award at the American Association of Geographers Annual Meeting!

Horrifying. We need to know exactly how this happened. Again, I do not believe for a second that this was intentional, but a terrible terrible mistake was made. We very much need to know if any policy changes contributed to it.

📢 I'm recruiting a postdoc to start in summer 2026! My lab is part of @Berkeley_EECS, @UCJointCPH & @berkeley_ai. We're looking for candidates in AI & society, with projects on the societal impacts of gen AI (collaborating w/ real-world orgs) and modeling human behavior with AI!

.@OpenAI is nothing without its people -- many of whom are brilliant, ethical, and able to work anywhere. Please, guys -- is this empowerment of authoritarians really what you want to be striving towards? Your talents are better-used elsewhere.

200 Google and OpenAI staff have signed this petition to share Anthropic's red lines for the Pentagon's use of AI let's find out if this is a race to the top or the bottom notdivided.org/

AI is changing economics, and --- as we just saw in Dwarkesh's interview with Dario --- AI researchers need to start thinking about economics too! The Center for Applied AI at UChicago will be hosting an AI & Economics Summer Institute to explore exactly this. We will bring together leading researchers with advanced graduate students in economics/AI/ML/NLP for an in-person program between Aug 6 - 11.