Davis Brown

@davisbrownr

Research in science of {security, safety, DL}

nyc

davisrbrown.com

Joined June 2017

856 Following

698 Followers

Replies Reposts Media

Davis Brown
@davisbrownr
Jul 8
a famous AI puzzle: since GPT-3, models have consumed a significant fraction of human knowledge. However, only very recent models have made discoveries by connecting ideas across fields. Why did it take so long?
1
Davis Brown
@davisbrownr
May 19
Short note: we are overestimating model safeguards for structural reasons. As models become more capable, they are juicier targets for misuse. So stronger attacks may only appear later, once models are worth targeting.
3
Davis Brown
@davisbrownr
Apr 10
In new work, we find that cheating on model capability evaluations is rampant. For example, the top 3 Terminal-Bench 2 submissions all cheat, usually by sneaking the correct answer to the model. Blog linked below.
7
Davis Brown
@davisbrownr
Mar 18
Mathematicians use conjectures to point to important, open problems. We collect nearly a thousand (currently 890) recent conjectures from the math literature for a new dataset, OpenConjecture. On a subset, GPT-5.4 finds candidate proofs, and formalizes several in Lean.
10
Davis Brown
@davisbrownr
Jul 17, 2025
Our dataset of open algebraic combinatorics questions received an oral, and is being presented today at ICML! the talk is happening now, and poster presentation is from 11 am to 1:30 pm (West B2-B3 503).
Davis Brown
@davisbrownr
Mar 11, 2025
How useful are AI tools to working mathematicians? We are releasing a suite of challenging (many of them open!) research-level questions in algebraic combinatorics to test conjecturing ability in pure math.
1

Davis Brown

Davis Brown’s posts