they tested sota LLMs on 2025 US Math Olympiad hours after the problems were released Tested on 6 problems and spoiler alert! They all suck -> 5%

My latest working paper in which we explore men's and women's job sorting behavior when *workplace hostility* is present - and how that may interact with the rise of increasingly widespread workplace arrangements: *remote work* (vs. in-person) and *teamwork* (vs. solo).

Question: Can Google AI overviews be considered a form of self-preferencing since the search engine disintermediates third-party web pages in favor of "its own" AI-generated content? (cc Andrey Fradkin)

Shreyas Sekar Calling this self-preferencing is tricky, since consumers likely prefer the overview to clicking on Chegg. This sort of issue will be ubiquitous for any intermediary and information provider. We will need a new economic model for creators of information (news outlets, etc...).

I suggest that the mathematicians in my twitter circle carefully follow this this is the first instance, in my career, of an AI tool being genuinely, non-trivially, surprisingly useful

Serious question: How do you folks use AI for anything critical? I gave Claude my previous year tax returns and asked it to add up the numbers in two specific rows and it still got the amounts wrong? Anthropic

Mildly Amusing/Frustrating: 1. Delta between CORS (Canadian OR Society) annual meeting registration fee for members and non-members is $150 2. CORS membership fee is $140

A reminder that R Ben Shapiro (of UW) and I run CERAMICS, a way for outsiders who want to learn about getting into CS Ed research. Read more and apply here! docs.google.com/document/d/e/2…

How sellers blur the line between real and fake reviews

Excited our paper "Learning Product Rankings Robust to Fake Users" (with Vahideh Manshadi, Jon Schneider & Shreyas Sekar Shreyas Sekar ) won the 2025 Best OM Paper Award in Operations Research! 🏆 We design ranking algorithms that resist manipulation by fake users.