Simons Institute for the Theory of Computing

@SimonsInstitute is inviting fellowship apps for the Spring 2027. The term includes a semester-long program on various facets of constraint satisfaction problems & their connections to proof complexity, quantum information, coding theory & beyond. cstheory-jobs.org/2026/02/06/fel…

Excited to give a talk at @SimonsInstitute Trust in Decentralized Systems Workshop on Tuesday at 11am! Title: "2026 Is the New 2016" — on federated memory, contextual privacy, and personalized agents. The privacy conversation has moved way past training data memorization. With persistent memory, tool use, and agents acting on your behalf (#clawdbot👀), the real risk is what models do with the data you feed them at inference time. Context window is the new attack surface! I'll talk about our new benchmark CIMemories, where we test whether models can actually make context-dependent decisions about what to share with whom from memory. Turns out they really can't, up to 69% violation rates, and it only gets worse the more you use them. Link to slides🔻🔻🔻

On Pi Day, #1stproof, a community experiment to measure the capability of AI to do research-level math, announced their forthcoming second batch and benchmark. 1stproof.org

The latest episode of our Polylogues web series explores what we're learning from #1stproof, a community experiment that asks how well AI can do research math. With Dan Spielman, Nikhil Srivastava, and host Venkat Guruswami. youtu.be/G0DqKlAcNxM?si…

Today is "Big Give", Berkeley's 24-hour online fundraiser Last year @Berkeley_EECS graduated 1,029 Computer Science majors. Next year it will be ~350. Enrollments were slashed primarily due to the high cost of instruction, with undergrad teaching assistants now costing the department $73-82/hr (equivalent pre-tax comp. rate ~$89-103/hr; see tinyurl.com/ugradtas) If you want to help us maintain our excellence in this world of high costs, you can donate to the Berkeley EECS Excellence Fund at givingday.berkeley.edu/amb/supporteecs

Great series of blog posts on quantum advantage by Dominik Hangleiter, one of the postdocs in our Quantum Research Pod. quantumfrontiers.com/2026/01/06/has… via @IQIM_Caltech #simonsquantum

Next week at the Simons Institute, a workshop on Trust in Decentralized Systems. Register to attend in person or view the livestream. Hope to see you there! simons.berkeley.edu/workshops/trus…

4/4 If you ask an LLM to categorize a movie review, you implicitly give it the task & the review and get the answer in one single sweep, said Matthew McDermott of @Columbia, at the Simons Institute workshop on Theory of Computing and Healthcare. Video: simons.berkeley.edu/talks/matthew-…

3/4 A foundation model "takes in both the data and the task and predicts a label specific for that task...This is exactly how we use the best examples of foundation models today, which are language models," said Matthew McDermott of @Columbia, at the Simons Institute.

1/4 Foundation Models: "We don't really know what we should expect [the phrase] to mean...and this causes real challenges," said Matthew McDermott of @Columbia, at the Simons Institute workshop on Theory of Computing and Healthcare. We need a formal definition, he said.

2/2 Ken Rothman spoke of the history of, and the right way to do epidemiology, saying that "computing technology is not really a fundamental necessity for epidemiology research," at the Simons Institute workshop on Theory of Computing and Healthcare. Link: simons.berkeley.edu/talks/ken-roth…

1/2 Ken Rothman of Boston Univ. began his Distinguished Lecture at the Simons Institute on "Epidemiology is Easy – Anyone Can Do It" with a caveat: “Anyone can attempt to do it but doesn’t always work out that well.” He spoke at the workshop on Theory of Computing and Healthcare.

2/2 "Most of the interesting computations that happen today are in some way or another computations that happen on individuals’ potentially sensitive data," said Katrina Ligett of HUJI at the Simons Institute workshop on Theory of Computing and Healthcare. simons.berkeley.edu/talks/katrina-…

1/2 "There is sort of an obligation on the part of somebody who spends a lot of time thinking about privacy to open with the bad news," said Katrina Ligett of HUJI, talking about Research on Sensitive Data at the Simons Institute workshop on Theory of Computing and Healthcare.

Two counterintuitive puzzles, from @alexbellos and @fortnow. blog.computationalcomplexity.org/2026/02/a-prob…

Join us at 3:30 p.m. PT. Register to attend or access the livestream. simons.berkeley.edu/events/respons…

3/3 "Current training recipe doesn't really support fragmented data," said Sewon Min of @UCBerkeley at the Simons Institute workshop on Federated and Collaborative Learning Boot Camp. Video: youtu.be/vmt2_LZ8zgI?t=…

2/3 One reason neural scaling laws might stall is because data is getting fragmented. "I’m talking about proprietary datasets, where datasets are owned by different owners...that cannot be gathered into a central location," said Sewon Min of @UCBerkeley at the Simons Institute.

1/3 Will increasing compute and training data continue to improve performance of foundation models? Maybe. "I’m interested in asking: if it’s not the case, then why?" said Sewon Min of @UCBerkeley at the Simons Institute workshop on Federated and Collaborative Learning Boot Camp